During pregnancy, I was no longer able to run, so I sought out other things to do to find that zen-like outdoors feeling. I gravitated to birdwatching because it requires you to be fully present. To identify birds, you have to listen closely, pay visual attention to your environment, and synthesize a lot of different inputs at once - bird calls, colors and markings, behaviors, and more.

As a Type A person, I also like the competitive aspect of birding. Birders build a “Life List” using the free Merlin app (part of Cornell’s ornithology non-profit) and track how many unique species they have identified in their lifetime. You can also create checklists to count and log the types of birds you see on a given outing or attempt a Big Year - a year in which you see as many bird species as possible (we’ll have to wait until retirement for this one). It’s strangely addicting.

There is a paradox here that is worth exploring as we think about the future of tech and AI: many people are seeking out “IRL” hobbies to escape from the omnipresent screens, pings, and digital noise in their lives. At the same time, we are going online to connect with like-minded communities, to dive deep into the “lore” of our new hobbies, and to track and share our milestones. It seems that there is a subconscious divide in many people’s mind between “bad tech” that keeps you hooked to a screen, and “good tech” that enables healthier activities.

The digital tools I’ve used to learn and grow my birding hobby are prime examples of what I think “good tech” looks like: technology that makes you more engaged with the world, not less.

Bad tech pulls you inward - driving more time spent behind a screen via cheap dopamine hits and addictive engagement loops. Good tech points you outward — toward the world, toward other people, or toward becoming more capable yourself.

Good tech optimizes for the user’s outcome, not the user’s continued presence inside the product.

As tech companies fight to deliver the next wave of consumer tech products that could shape how we engage with the digital world into the next decades, it’s worth exploring what would make this device a benevolent force instead of a damaging one.

Common principles of “good tech”

Birders have Merlin, runners have Strava, hikers have AllTrails, foodies have Beli, and readers have Goodreads - all digital communities that bring people together around IRL hobbies and activities. Across these platforms, there are unwritten commonalities that “good tech” environments foster.

They optimize for outcomes, not engagement

Merlin doesn’t nag me if I haven’t used the app in awhile. It’s there when I’m ready to bird, but isn’t sending me notifications if I miss a few days. When revenue depends on time spent (e.g. engaging with ads on TikTok), every design decision drifts toward extending time and energy spent on the app, vs. a customer-first purpose.

They reward contribution and expertise, not attention

Good tech platforms intrinsically reward quality contributions with community status, kudos, or expert badges. In contrast, Instagram and TikTok monetize engagement and views, which can incentivize attention-grabby content that may not be the most valuable to users

They build human capability, rather than replacing it

Merlin explains bird field marks and habitat context rather than just telling me “that bird is a Goldfinch.” This approach builds my confidence as I bird in the real world and leaves me with skills that can translate outside the app.

They reconnect you with the world

Strava suggests local races that you can join IRL. It helps users discover routes and running clubs near them. Ultimately, good tech platforms span both digital and physical spheres, continuing to put emphasis on humans connecting as a worthy objective.

Ultimately, it’s about agency. The best consumer technology should increase your ability to do things rather than increase your dependence on the technology itself.

Reviving a dead internet

Dead internet theory posits that much of the web’s traffic and engagement, including comments, likes, and entire accounts, is already generated by bots and agents, and that the share will only grow. The logic is hard to argue with: when a platform rewards views and engagement, anything that can manufacture views and engagement can win.

In contrast, “good tech” platforms will be more resilient to the bot takeover. A bot can’t verify a rare bird sighting or show up to a Saturday run club. The things that make these communities feel good for humans also make them hard to fake. They’re anchored to the physical world and to reputation among real people. Their value is ultimately validated offline.

Products designed around real-world outcomes may actually be structurally better positioned for an AI-generated internet.

Good tech toward the future

Two developments make me hopeful about the next wave of consumer tech. The first is the growing role of voice as a way to interact with technology. The second, which voice makes possible, is the move toward screen-free devices, including the hardware OpenAI and Meta are building. Both could help us unlearn some of the bad habits we picked up from earlier eras of the internet.

But screen-free doesn’t guarantee that tech is “good.” If designed to incentivize usage and engagement above all else, this next wave of devices could simply replicate the dark patterns inherent in earlier digital eras. For example, voice can be another vector to fake human presence and potentially game platforms in the attention economy.

The question shouldn’t be whether an AI device can keep us entertained. We should demand more and ask whether it can make us better at living without it.

Last Saturday in Brooklyn Bridge Park, I pulled out Merlin to confirm an unfamiliar bird call. The app did its job, and then it got out of the way. That’s what I want from the next generation of tech: to help when called upon, but then to let me live my life.

Stay Curious, Melina

Disclaimer: The views and opinions expressed in this blog are solely my own and do not reflect the official policy or position of my employer or any organization with which I am affiliated. All content is written in a personal capacity.

Leave a comment

I’m Melina, a consumer tech enthusiast and AI & data executive in the beauty space. All the nerdy opinions expressed here are my own.

This week, I’m at the RetailClub AI festival in Huntington Beach to speak about the future of AI in consumer discovery. I will be letting loose since it’s my first trip away since having my son! If you’re planning to be there, I would love to meet you! Look for me wherever the margaritas are.

This back-to-school season, with my toddler starting school for the first time, the amount of family admin on my plate has increased exponentially. Costco runs, holiday travel scheduling, and incredibly verbose school emails (that seem to bury the most important information at the bottom) are taking over my life. With a toddler and a newborn to raise, and a household to run, it seemed like the perfect time to test out the new slate of AI agents offering to automate the mundane from my life.

As I discussed last time, I’m dubious that the messaging of “automating everything” really resonates with most consumers. This recent tweet sums this up well - many shopping missions are about the journey, not the destination. The polling of the group chat for opinions on an outfit. The family FaceTime to decide on hotels for an upcoming trip. These are valuable parts of transacting online that many don’t want to lose. But there remains a large chunk of tasks that are purely frictionful. It’s with these tasks in mind that I put Instinct and Muse through the paces.

The AI agent explosion: Why Now?

Muse and Instinct have been winning the battle of headlines lately, but they are far from the only personal AI agents popping up. XAI launched Grok Bot in August, giving persistent agents their own cloud computers that work around the clock. Town, founded by Plaid's former CTO and an ex-Google applied AI lead, gives every user a "Townie" with its own email address that learns their routines and starts proactively taking tasks off their plate. Apple is finally rebuilding Siri into something closer to a true agent, while OpenAI is widely expected to follow suit. Instinct went from a $50M to a $2.5B valuation in four months and is reportedly now in talks at $10B, illustrating just how “hot” this space is. Why are these all appearing now?

Owning the personal AI agent interface is incredibly powerful. It enables these platforms to orchestrate complex activities on the user’s behalf, gaining tons of proprietary insights into us as consumers across all types of tasks and activities and becoming a sticky surface on which to layer other digital experiences. The end goal for many is to become the “front page of the internet” - the first place we go when taking action online. To drive adoption, many of these tools offer free usage. But they will monetize, and the potential to make money after they prove their value is significant.

We have also seen an explosion of brand- and retailer-owned AI agents. Their biggest limitation is context. Muse knows my Meta social graph, can access my Google workspace, and sees my preferences across task types. In contrast, Target’s chatbot only knows about my Target search and purchase history. It will be hard for it to deliver results that are as personalized or useful. But retailers hold cards the general agents don't: live inventory, pricing, and loyalty data. Whether these agents compete or cooperate will shape where commerce goes next (more on that in a future post). But back to Muse and Instinct.

When I had the idea for this post, I thought I would be comparing Instinct, Muse, and other tools. But I quickly found that this isn’t really necessary. The UI and capabilities of these tools are very similar. The winner will be decided by who can get user adoption first, who can make it really easy to onboard, and who can layer on other “apps” and experiences that add consumer value. Based on Meta’s scale and the types of data they already have access to, I am bullish on Muse.

What’s a personal agent?

Both Instinct and Muse are chat interfaces where you can text it questions and tasks. By enabling access to things like your Google accounts, payments, and retailer accounts, the agents can execute tasks (like purchasing things end to end on your behalf, sending emails, curating files).

I tested out Muse and Instinct on a few tasks, all of which they could both accomplish seamlessly.

Schedule pickups for my dry cleaning and donations

• The friction: These tasks currently take just a few minutes for me to accomplish. But it’s annoying to navigate to the site, log in, click around their calendar for availability that matches mine, and fill out a form to get an appointment.

• The result: Using it’s own cloud-based browser, Muse navigated the site for me and booked pickup slots for my chosen availability, triggering a confirmation email and adding the time to my calendar.

• Leveling Up: I have now asked Muse to check in monthly on dry cleaning pickup, removing the cognitive load of me remembering to do it.

Generate returns labels for recent purchases

• The friction: Generating a return label is annoying because brands and retailers use different providers with different flows. Some offer returns only at certain locations, others charge for certain types of returns (printed label vs. QR code), and they have varying timelines and policies around accepting returns. Submitting a return often involves combing your inbox for order confirmation numbers and filling out forms online.

• The result: Instinct was able to generate a return label and drop the PDF in my inbox. Notably, in a multi-product Zara order, it was able to return some of the items while keeping others based on a simple description of what I wanted to keep.

• Leveling up: If I wanted to, I could have Instinct/Muse run a weekly digest of what I bought and ask me what I wanted to return (along with return policy). So far, I don’t think I have enough returns volume to justify this, but it would be an easy way to make sure I don’t lose track.

Research toy options for my son, build an Amazon cart, and send to me

• My toddler is obsessed with ice cream. My husband heard from a friend that they have an “ice cream toy” their kid loves. I asked Instinct to find me the best option for his age, with a focus on longevity (something that will keep his interest for a long time). Instinct summarized the options, added it to my Amazon cart, along with a few other items, applied my rewards dollars, and sent to me.

• Notably, I tried this use case with ChatGPT about a year ago, and the agent got tripped up on shipping fees. This time, Instinct worked inside my own Amazon account, leveraging Prime shipping, saved address, and preferred payment methods. In the past year, the payments plumbing has matured: agents can now pay with tokenized virtual cards capped by spend and merchant, so you can hand over the wallet without handing over your card number.

• Also notable - as of Sunday, Amazon is blocking Muse from accessing it. This is just the first shot in a larger battle about who gets to own shopping interfaces and experiences in the agentic world.

From my little experiment, I have a few observations about what these tools do well, vs. where I see opportunity.

What Muse and Instinct do well

• Instinct and Muse are VERY easy to use out of the box. You can download the apps and have a task completed in minutes, which very quickly illustrates value

• The ongoing engagement, like how Instinct reacts to messages with emojis, is a lightweight way of demonstrating that the tool is progressing a task, without needing to generate lots of follow ups

• I like how Muse renders a browser in the chat so you can easily follow along and take control. It reduces the concern that these tools are operating as a “black box.” It’s essentially mimicking the steps I would take as a human on the site.

What’s not working: Onboarding

To win, these AI agents will need a broad base of users - from the techy to the non-techy. While the tools are super easy to use, there is still a big opportunity to improve onboarding and educate users about what they can actually do. Muse did a better job at this than Instinct, with a clear onboarding flow suggesting popular use cases. It was somewhat jarring how Instinct immediately asks for data access without really explaining why it needs the access and how the access adds value for the users.

To improve onboarding, I imagine the ideal analog scenario. If I were hiring an executive assistant, I would schedule an hour of dedicated time with them, tell them all my preferences, and talk to them about what’s important to me. I would understand the basics of how they do their job and what info I can provide to make it easier. The ease of starting to use these tools is a double-edged sword because we miss out on the in-depth onboarding that can create real value and stickiness over the long term.

As a result, I suspect we’ll see a lot of people use the tools a few times for novelty, then peter off because it’s not clear how they really fit into their lives.

What’s not working: Cognitive load

Muse and Instinct are sometimes slow at completing the tasks, which is ok with me - they function in the background. However, I find it disruptive when I am constantly getting pings from these tools asking me follow ups or updating me on status when I have mentally moved on to work on something else. Similar to how I would reference real-life executive assistant onboarding in the digital flow for these tools, I would recommend playing with update cadence with this in mind. High-performing EAs don’t constantly ping their team with follow-ups through the day. They grab 5 minutes to get clarity on a suite of issues in one swoop.

Similarly, these tools are looking to keep you active and are offering proactive ideas for how to use them. I love this in theory, but the proactive pushes haven’t been very helpful to me so far. Often, the emails I haven’t answered yet aren’t answered because they aren’t they important. So I don’t need multiple AI agents reminding me about them. This is an area that will likely improve as I get deeper into a working cadence with these tools.

Fighting Bureaucracy: Awesome for consumers, tough for brands

In past years, the digital world has become more hostile to consumers. I’ve observed some brands that have stopped offering phone numbers or emails for support, and require you to engage with a slow chatbot (annoying). Brand return policies are becoming more restrictive (uncool!). Our relationship with many brands is becoming more antagonistic and impersonal. This is backed up with data - Carey School of Business conducted a “rage survey” in 2025 measuring just how pissed off consumers are with brands. 77% had a product problem, more than double the 1976 rate, and of this group, two-thirds felt feelings of rage when trying to resolve it.

With tools like Muse and Instinct, I can cut through these bureaucratic barriers and get what I want more easily. But this is also terrifying for brands, who will have to field all of this outreach. One user on X tasked Muse with getting compensation for a delayed flight. Within minutes, Delta awarded the customer $250. Consumers obviously deserve compensation in some cases, but it raises questions about how brands are supposed to sustainably handle a barrage of outreach.

The brands that win will quickly figure out how to triage and support their valuable consumers with high-touch support, automate basic inquiries, and install guardrails to prevent people from taking advantage.

If you’ve played around with Muse and Instinct, especially for shopping, I’d love to hear your take in the comments!

Leave a comment

Stay Curious,

Melina

This is my first time writing after a long hiatus. My most notable update is that I recently had a baby! We are now a two-boy household (our other son is 2.5) and it is as chaotic as you might imagine.

As a quick refresh - I’m Melina, and I lead data & AI for digital and marketing at a large beauty company. I love all things consumer tech and data and thinking about how tech shapes how we live and how we shop.

I took some time off from writing here because truthfully, pregnancy really zapped my creativity. I write because I love to explore topics in consumer and tech with you all. During my pregnancy, all I could really handle was showing up to work and keeping my toddler alive…and even that was often a struggle. Now that my precious little parasite has transitioned to life outside of my body, we are SO BACK, baby!

Since I last published, it’s been notable to see the rise and acceleration of consumer backlash against AI. While AI-driven traffic to US retail sites grew 393% year-over-year in Q1 ‘26, the trust gap is widening. Pew Research Center found that 52% of Americans are “more concerned than excited” about AI’s role in everyday life, up from 37% in 2021.

I’ve always been interested in how technology can shape the shopping experience - reducing friction, making discovery more seamless and personalized - both online and in retail stores. Since the internet and rise of e-commerce, LLMs have the biggest potential to shape this industry. Yet many consumers are unhappy with how the technology is manifesting in their lives.

Rhetoric from the leaders of OpenAI and Anthropic highlights how using consumer AI tools like LLM answer engines and coding assistants can drive efficiency through automation. As a corporate leader in tech, efficiency and automation are well and good - but as a human, this value proposition rings hollow. Most normal people don’t view their personal to-do lists - things like planning meals for family, buying birthday gifts for kid’s parties, or planning a vacation - as scopes of work to be optimized. Rather, doing these things thoughtfully drives a feeling of satisfaction for many, myself included.

Recently, Sam Altman tweeted about how people can use AI to generate podcasts about their kids’ interests. Predictably, this received tons of backlash. For many people, talking to their kids about their interests is one of the most special, human moments in their day. The concept of outsourcing that to AI feels off-base and out of touch. As one commenter put it: “what if you just talked to your children”

Like Altman’s tweet, many of the early popular use cases for LLMs addressed the most fun, human aspects of tasks rather than the boring parts. It’s easy to use an LLM to brainstorm ideas for weekly toddler meals. But it’s still fairly difficult to successfully automate online grocery delivery end-to-end. For me, the creative parts are the ones I want to preserve, while the time-consuming process of tallying exactly what ingredients I need, comparing prices, adding to cart, and scheduling delivery is the stuff I would be happy to avoid. This creates a bind for the teams marketing Claude and ChatGPT (and consumer AI tools more broadly) - they want to drive adoption by presenting accessible use cases for the tech, but the ones that may resonate most with consumers are actually harder to unlock, requiring more data-sharing (a concern for many), integrations across tools and systems, and initial setup.

I think this gap is behind a portion of the backlash. People don’t see real value yet unless they are closer to the technology and willing to invest time to build something cool.

The killer consumer use case for LLMs is still to come.

A quick note on terminology: I’m specifically talking about consumer AI here - the tools people use directly for everyday tasks like shopping, planning, and research. That’s distinct from the AI already embedded in the back-end tech stack of most businesses, which is already mature in many areas. Consumer-facing AI, especially in shopping, is still evolving fast.

The divide between hype and lived reality applies to brands and retailers, too. I wrote last year about how consumers’ adoption of AI will vary based on their shopping mission, e.g. buying a TV will involve deep AI-moderated research, while buying a new $10 lipstick will be facilitated by hyper-targeted social content. This sounds obvious, but it highlights how the current wave of tech adoption isn’t one-size-fits-all; rather, each brand or retailer needs a clear understanding of their consumer and their needs and how their product fits within that context. For some companies, it may be investment in an AI chatbot or AI-enabled search experience; for others, it may be a focus on customer service, while others will leverage agentic workflows to automate and scale creator content.

Amazon has been an early winner in AI transformation in part because they have huge amounts of data to make tech-driven experiences relevant and the types of purchases people make from them - everyday essentials - are ripe for automation. As of Q1 2026, Amazon held 9.3% of US consumer retail spending, up from 8.6% the prior year, highlighting how these investments are starting to pay off. In fact, Amazon unseated Walmart as the top US retailer in this period. As a consumer, I’ve noticed search has become more relevant while customer service has become increasingly automated and streamlined.

The luxury segment is much tougher to crack. The jury is still out on how top luxury players should be approaching consumer AI investments, and while there have been a lot of buzzy feature launches (lots of chatbots), it’s unclear that the technology has really changed how we shop.

I predict that the next wave of AI investments in luxury will be focused on a return to “analog” life - in other words, leveraging technology to reduce time spent in front of screens. This is the area of consumer tech I’m most excited about.

How this will look:

• Personalized search results that make it easier to find what you need quickly, with fewer clicks (already well underway, especially as shopping sites leverage natural language and visual search more)

• Wearables that enable us to seamlessly execute tasks without interacting with a screen

• Automation and streamlining of “annoying” parts of shopping, like returns, issue management, sizing, filtering

• Seamless experiences across channels, including handoffs from AI chat to real humans depending on consumer need

• In luxury hospitality and real estate, the move to biometric engagement with appliances, vs. app-based, screen interactions

• In wellness, passive data collection that is presented to the user as insights, vs. tons of discrete data points

This shift is changing the way we measure success as digital leaders. We don’t necessarily want our customers spending more time on site, visiting lots of pages, and scrolling “below the fold.” Rather, we want quality traffic to convert quickly, already armed with knowledge about what we offer.

—

While I empathize with the burgeoning resistance to AI, I remain a tech optimist. The backlash isn’t really a rejection of the technology itself - it’s a reasonable reaction to use cases that optimize for the wrong things.

Working in this space, it’s up to us to shape the future of consumer tech in a positive direction, investing in the use cases that actually matter to people, and freeing them up to spend more quality time living their lives. The winners in this next chapter won’t be the companies chasing efficiency at the expense of their customers. The companies that win will deeply understand consumers and exercise discernment about how technology should show up in their shopping journey.

—

Melina leads AI & Data for digital and marketing at a large beauty company. All opinions expressed here are my own.

Leave a comment

While we managed to avoid politics at the Thanksgiving table, we did discuss AI. This got me thinking about the critical role of this holiday season in shaping AI adoption and sentiment.

Holidays are a catalyst for consumer tech adoption

Multi-generational sharing

The holiday season is unique in that multigenerational families typically spend lots of time together. This is a natural opportunity for different generations to learn from each other. While I learned about “67” and the latest fashion trends from my teenage cousins, my parents and older relatives were eager to pick our brains on the latest tech headlines.

A few years ago, early in the crypto boom, my husband explained to his parents and young cousins over Christmas how Bitcoin worked and gifted everyone a small share of cryptocurrency. Prior to this, the majority of the family had heard about crypto but hadn’t purchased it.

Similarly, with AI, my parents and family members had all read about AI in the news and several had dabbled with ChatGPT and noticed AI Overviews on Google. But beyond this, there was little experience with using AI tools for more complex and multifaceted tasks. I worked with my dad to shop for a new coffee machine using various specifications and preferences. He was surprised how well the options presented matched his needs!

Easy to use

AI has one major advantage over every tech wave that came before it: zero barriers to entry. Crypto required wallets, exchanges, seed phrases, and a totally new mental model for how money works (I still don’t know what happened to the NFT I bought a few years ago). Even early social media required entirely new behaviors and philosophies around privacy. Anyone who has searched on Google can transition to AI search with zero friction. There’s no new interface to master, no new vocabulary to memorize, no need to invest hours before you see progress. For new users, it feels magical to ask Gemini for a cocktail recipe using ingredients in the bar and immediately see fun, easy options.

For the record - our holiday cocktail (courtesy of AI) was the apple cider margarita: Apple cider, mezcal, Aperol, and lime. Delicious!

Built-in adoption via gifting

Gifting is another vector that is naturally helping AI scale this season. The majority of Americans are purchasing tech as gifts this year. Many new consumer tech products - from the newest iPhone to Amazon Alexa to even the fancy Nanit baby monitor comes with AI built in. Whether I am looking to “adopt” or not, these features are usually pre-programmed and part of the default feature flows - I would have to try really hard not to use them.

In her recent piece on AI holiday predictions, Kiri Masters likens AI’s rapid expansion to the quick adoption of mobile payments. Mobile payments are just so convenient that it’s obvious that everyone should use them. Perhaps in the early days, there were concerns around connecting your payments to your phone - but as soon as people realized the ease and benefits, these doubts were easily quashed.

In fact, AI is becoming so integrated into e-commerce already - via pricing algorithms, automated customer care tools, and personalized merchandising - that shoppers are benefiting from it without even realizing they are “using” it. This holiday season, consumers were exposed to more AI-enabled features than ever before.

Retailers and AI companies are releasing lots of new features aimed at holiday shoppers

The Amazons, Walmarts, and Targets of the world make 50% or more of their US annual revenue in November and December. Their teams have undoubtedly been working around the clock to release as many snazzy AI features as possible ahead of the holiday season to make their online experiences as frictionless as possible.

Their goals include:

• Increasing share of wallet - Walmart wants to get people to shop with them instead of going to Target, for example. An easy search and discovery experience, and price transparency, will make customers “stickier” and prevent them from looking elsewhere.

• Cost savings - automating super common customer service queries like “has my order shipped yet,” or “can I change the shipping address on the order I just placed” or “can I get that Black Friday discount applied that expired a few hours ago” can save brands upwards of $5/contact - a big boon to margins, especially as profitability is already challenged with higher costs.

• Conversion - customers are deal-hunting on Black Friday. Capabilities that improve pricing transparency and automatically match competitive offers will drive more sales.

A few recent examples:

• ChatGPT now supports Instant Checkout for major retailers including Walmart and Target — shoppers can add to cart and complete purchases directly inside ChatGPT without opening a website.

• Amazon continues to enhance Rufus, an AI shopping assistant built into the Amazon app and site — shoppers can ask natural-language questions (“gifts for my sister who loves hiking,” “best space heater for a drafty apartment”) and get curated product recommendations.

• Target rolled out an AI-powered Holiday Gift Finder — users describe the recipient (“mother-in-law who cooks, under $100”) and get context-based picks instead of manually browsing categories.

• Walmart’s Sparky AI assistant now builds entire shopping lists or carts from a single prompt — e.g., “ingredients for Christmas brunch for 8” or “gifts for a 5-year-old who loves dinosaurs.”

• Best Buy added AI-driven product comparison tools — customers can ask for side-by-side recs (“best laptops for college under $800”) instead of filtering hundreds of SKUs.

• In-store: Walmart introduced an AI Store Navigator in its mobile app — it shows real-time aisle locations, stock availability, and promotions tied to the exact store a customer is in, reducing time spent hunting the shelves during holiday rush.

Many new features target a budget-conscious demographic - changing the face of AI users

A throughline in many of these recent launches is their focus on the budget-conscious shopper. That reflects the reality of the consumer economy today: spending power is strained, and shoppers are working harder than ever to make every dollar count. U.S. households have become more price-sensitive amid persistent inflation and higher borrowing costs, and more than half of consumers say they are prioritizing value and bargain-hunting this holiday season, according to Deloitte’s 2025 Holiday Retail Survey.

This is a notable pivot from AI’s early audience. Like most tech waves, AI adoption initially skewed toward younger, higher-income, tech-savvy consumers, and the use cases that received the most attention — saving time at corporate jobs, automating emails, planning travel, booking restaurants — mostly resonated with that group. A June study found that high-income households were nearly twice as likely to use AI for financial decisions as lower-income households.

But the vast majority of Americans don’t live in that world. For most of us, saving money trumps bumps in convenience - it’s cool that Amazon will assemble a grocery cart for you, but it will probably be overpriced. For these Americans, the promise of AI isn’t email automation - it’s price transparency, deal discovery, comparison shopping, and not getting ripped off. And that’s exactly where AI innovation is now accelerating: Walmart’s Sparky assistant and Amazon’s Rufus increasingly emphasize finding the best bargains and comparing options, not just convenience. Also, shout out to Kiri who touched on some of these same trends in her recent piece - Retail Media Breakfast Club is a must-read if you haven’t yet subscribed.

What’s especially interesting is the return of the store as a tech touchpoint. More than 70% of U.S. retail spending still happens in physical locations, even after the pandemic e-commerce spike, according to McKinsey. So if AI meaningfully improves in-store shopping, it has the potential to be transformational. Walmart’s new AI Store Navigator, which shows real-time aisle locations, stock availability, and local deals tied to the exact store a customer is in, is a clear signal that retailers see AI not just as a web feature, but as an on-premise customer-experience upgrade.

This holiday season feels like a shift in who AI is for. These launches aren’t aimed at a tech elite — they’re aimed at everyone. It echoes what we saw with Uber, Spotify, and the iPhone: the moment when a frontier technology stops being a novelty and starts solving normal people’s daily problems.

And the usage data is already reflecting that shift. A nationally representative 2025 survey found that 54.6% of U.S. adults now use generative AI, up nearly 10 percentage points in a year — a sharp broadening beyond the early-adopter crowd.

AI’s Turning Point

This holiday season represents an inflection point for AI adoption — not because people suddenly decided to “start using AI,” but because AI is now woven into the shopping behaviors people already do without thinking. In 2026, AI tools won’t feel like a novelty or a tech demo. They’ll be the unseen plumbing behind everyday decisions: comparison shopping across retailers, getting personalized deal recommendations, building gift lists, checking store inventory before driving over, and even automating purchases within guardrails like price caps and preferred retailers.

If you squint, you can already see that shift happening.

Black Friday confirmed it. U.S. consumers spent $11.8 billion online on Black Friday alone, up 9.1% year-over-year, and analysts directly attributed a major share of that growth to AI-assisted shopping and deal discovery.

We’re not talking about niche usage. AI-driven traffic to retail sites surged more than 800% year-over-year, with shoppers increasingly relying on chat assistants and agent-style tools to compare prices, spot deals, and check stock in real time

And zooming out across the full season, Adobe expects 2025 holiday e-commerce to reach a quarter-trillion dollars - $253.4 billion - the largest in history

These outcomes matter because they are indicative of a real, sustainable shift in how we shop. AI no longer sits on the sidelines as something tech-forward consumers choose to experiment with. The choice is disappearing. AI is being embedded into the funnels of the world’s largest retailers — search bars, navigation, recommendations, price matching, cart building, checkout — and increasingly, the in-store experience.

For consumers, this will feel like magic (at least when it’s executed correctly): less clicking, less hunting, less overpaying, less guesswork. There are still “bad AI” experiences out there - things that feel gimmicky and slow sites down. I’ll have to revisit my e-commerce icks with this in mind!

Thanks for reading, and Stay Curious!

Melina

Leave a comment

Voice-to-text isn’t new. But lately it’s felt like less of a novelty and more like an emerging status quo. This prompted to ask why - and what’s next.

Voice - rise, fall, and rise again

Voice-to-text capabilities emerged en masse over a decade ago, with Siri debuting in 2010 and Amazon’s Alexa in 2014. While these initially gained a lot of traction, usage flatlined as the voice-to-text capabilities weren’t always accurate. The widely-discussed voice app ecosystem - in which commerce would be conducted via conversations with at-home devices - did not materialize. I remember being approached by vendors trying to get my company at the time - a jewelry brand - to invest in creating a voice-enabled app. Needless to say, the business case was nowhere to be found. But perhaps, things are changing.

The resurgence of voice-to-text is not just in my imagination. The data shows that indeed, people are using voice-to-text more, and a key driver of this is using voice for search. Tools are also driving this behavior by adding more prominent microphone icons that signal the ability to use voice.

• 81% of Americans now use voice tech daily or weekly, with 68% increasing their usage over the past year.

• 32% of consumers say they use voice daily to perform searches they’d otherwise type — pointing to normalization of voice input as a primary way to search

LLMs have improved the quality of voice-to-text

Back in the early days of voice tech, using speech-to-text was like playing a high-stakes game of telephone. You’d say, “Hey Siri, remind me to pick up dry cleaning,” and she’d cheerfully start playing a random song on Spotify. Dictating an email while driving felt efficient—until your boss received a note riddled with typos.

The issue wasn’t us users, it was the technology. Early voice tools relied on rule-based speech recognition - enormous dictionaries of possible words and pronunciations. This meant the systems lacked context. So if you said, “book a flight,” the system didn’t actually know whether “book” was a verb or a noun. It just guessed.

Then large language models entered the chat. Modern voice-to-text systems (like OpenAI’s Whisper or Google’s Gemini Voice) don’t just transcribe words, they predict meaning. They use the same kind of neural networks that power ChatGPT to model the relationships between words, grammar, and intent. Instead of recognizing isolated sounds, they understand patterns and probabilities, mimicking the way humans talk a lot more closely. The result? Amazon has reported that consumers talk to Alexa+ (their enhanced AI version of the OG Alexa) 2x as much.

This leap from hearing to contextualizing is what finally made voice-to-text feel human, and the increased accuracy is a big piece of why utilization is increasing.

Subscribe now

Natural language search lends itself to voice

Unlike the keyword-based searches of the past, it’s now possible to ask a question naturally and get robust answers. I often ask ChatGPT for parenting guidance - like “what are some signs that my son is teething and what are some ways to treat the symptoms.” In old-school Google search, I would have made several keyword-based searches to get these answers, first searching for “teething symptoms” and clicking through a few links and then searching separately to research potential treatments and their pros and cons, perhaps reading through Reddit threads, WebMD and other sources. In this situation, there is little efficacy of using voice to search. In contrast, with natural language search, a single query takes the place of this longer process, reducing steps. Further, I rarely need to click through from the answer, meaning I can read it directly in the chat, or even use Voice Mode to have the answer dictated back to me.

As voice tech improves, companies are likely thinking of the best way to capitalize on the medium. We are already seeing companies like Friend, a voice-enabled AI necklace - pop up. If you haven’t checked out Friend and their totally cuckoo marketing (which recently went viral for taking over the NYC Subway), it’s worth a look.

When and how will we use voice for commerce?

A few months ago, I published my Value x Vibes framework positing how AI adoption would be different for different categories. This framework can also be used to understand what is next for voice tech.

Voice tech will play the biggest role in the “Automate” quadrant. These are the products we aren’t excited to buy (and may actually dread shopping for), but need to keep life running. From compiling shopping lists with Alexa to asking your agent via voice to reorder paper towels, voice tech will be commonly used at all stages in the funnel - search, discovery, validation, and purchase.

Voice thrives when:

• The action is routine or low-stakes

• The user already knows what they want

• The friction of opening an app outweighs the mental load of just asking out loud

• The platform can act autonomously (no need for decisions or comparison shopping)

In other words, typing is for considered decisions; talking is for delegation. And these are the types of shopping tasks we want to delegate.

As search becomes conversational (“What’s a shampoo that won’t strip color?”) and checkout becomes agentic (“Order the one with the best reviews under $20”), voice becomes the natural bridge between thought and transaction. The less you need to think, tap, or compare, the more likely you are to hand tasks off to an AI that’s always listening. This is why Amazon, Walmart, and other big retailers are investing in voice tooling to capture consumers in an easy to use ecosystem.

I can also see voice commerce taking root in the “research” quadrant - perhaps focused less on conversion, and more on product comparison and background information. On the drive to Best Buy, you could activate voice mode to share a brief overview of the best-selling TV options and their pros/cons. You could even have your AI tool check inventory at a specific location.

In contrast, I have a hard time believing that more visually-rich, subjective product categories - like fashion, art, or other luxury goods - will see significant purchases driven via voice alone.

What’s next for voice and commerce?

Home assistants - but smarter

We’re all using Alexa to set our kitchen timer and play the “song of the summer” - but what if it could do A LOT more? Amazon recently released early access to Alexa+, their AI-enabled version. The marketing video shows some pretty meaty capabilities, though I think some feel a little bit invasive. We see a family interacting with Alexa+ over the course of the day, doing things like booking a vacation and scheduling an Uber. At one point, the father asks Alexa “when did the dog last go out?” and Alexa pulls up timestamped camera footage of the pup’s last bathroom break. As a formerly rebellious teen, I would not want my parents to have this level of surveillance around the house - but perhaps this is where the future is going!

It also raises concerns about balancing privacy and personalization across shared devices. If I’m researching birthday gifts for my husband, perhaps I want to keep that interaction separate. Or (as suggested in the Alexa+ video), if I’m confiding about mental health concerns, I wouldn’t want other family members to have access. There is scant documentation about how Alexa+ approaches these challenges - but permissions that enable collaboration while respecting privacy are an opportunity beyond home assistants that I hope to explore further.

Customer care taken to the next level

I recently shared a great customer care experience I had with Oura ring. They struck the perfect balance between automating pieces of the experience and handing off to a real human agent seamlessly when needed.

Voice capabilities have the potential to further enrich these experiences. I chose chat for my Oura ring needs, but calling is another options that many default to when seeking support. Imagine how much better airline customer service would be if they could staff virtual agents to help with the easier tasks alongside humans for the more nuanced tasks. If I had to guess, we’ll see significant adoption in 2026.

Fin is one example of a next-gen care agent that has voice capabilities.

Diagnostic experiences

Imagine a diagnostic tool that helps you build your ideal skincare routine based on a conversation instead of a static quiz interface. This would be an opportunity for brands to share their experience digitally in an inviting, brand-right way - scaling the type of expertise you would otherwise only have access to in a physical store.

In my experience, some of these customer-facing experiences end up as gimmicks that fail to drive sales or efficiencies. However, with a deep understanding of one’s customer, I think it will be possible to integrate voice thoughtfully into processes like product recommendations, upsell/cross-sell, etc.

Voice Engine Optimization?

Looking ahead, this means that in addition to SEO and GEO (Generative Engine Optimization), brands will also need to think about VEO — Voice Engine Optimization. I know what you’re thinking…even more acronyms.

As we search more with voice and utilization of Voice Mode and other tools increase, brands need to think about how discovery and conversion will operate via spoken language.

First, discovery. Unlike a Google search or Amazon results page - which show multiple options - a voice assistant will recommend just one preferred choice - maybe two. Brands need to consider how their product data translates to voice search - how the product descriptions sound being read aloud, and how customers’ follow up questions can best be answered with FAQs and detailed use case information. Brands also need to specify who their product is for - and who it isn’t for. Precision about target customer and their needs will help brands tell their story in a way that LLMs can parse to provide super-targeted and high-converting recommendations - which will keep a brand “top of mind.”

In these early innings, consumers are not buying new-to-them products with voice. These will index towards “order it again” experiences where customers ask Alexa+ or Siri to look up information on a past order and re-buy it. But winning the initial sale may not be enough - voice assistants may proactively share better deals. And if the originally-ordered item is no longer available, a brand’s product data needs to be good enough for LLMs to find and suggest the best alternative.

For voice to actually convert - not just answer questions - brands need some technical basics set up: identity (who’s buying), product data (what they want - including navigation of lots of variants and price options), and fulfillment (how to pay and where to ship). Luckily, brands don’t have to build every piece from scratch - platforms like Amazon and Walmart already have voice-enabled checkout baked in, so plugging into those ecosystems can unlock “order it again” purchases long before a brand has its own voice tech.

Holiday mode activated

It’s my favorite time of year - the holidays! To celebrate, I’ll be writing about some festive themes in the coming weeks. Stay tuned for tech-enabled gift guides and more!

ICYMI

Last time on Curious Commerce…

Stay Curious!

Melina

Would love to hear from you! Leave a comment or DM and let’s chat!

Leave a comment

On the personal front, I’ve been dealing with the seemingly universal phenomenon (at least among toddler parents) of a consistently sick kiddo. He is all recovered, but the last several weeks have been firing on all cylinders - long work hours, broken sleep, and a lot of time with the snot sucker (oddly satisfying). In between we did our first trip to the pumpkin patch. Check in on the toddler parents in your life - we are not okay!

I’m now writing to over 1,000 of you, which is super exciting. Thank you for tuning in!

On to Atlas. I’ll start with some discussion of strategy, then dive into a product teardown/mini demo.

First a quick poll…

What is a browser?

I’m starting super basic here. As a millennial, I never really knew life without browsing the internet. I never gave much thought to what a browser is or the value it provides.

Browsers enable us internet users to look at and interact with the internet’s content easily. They are the first layer (and gatekeeper) between us and everything that lives online. All the internets key players - consumers, brands/retailers, and advertisers - interact with the browser.

As it turns out, a browser occupies a valuable place in the internet’s value chain. As the “window to the web,” the browser sees valuable data like:

• internet history - what you looked at online

• personal details - your name, address, location

• your interaction history - did you click, scroll, abandon, come back to a page later?

Technically, we as users own our personal data - but we constantly grant access to other parties to make our experience online more personalized and efficient. The browser controls access to those data pipes, shaping our experience on the internet.

• Browsers rent your attention out to search engines: Browsers choose a default search engine, selling access to tons of valuable monetize-able traffic. It’s estimated that Google pays Apple $20B+ per year to be the default on Safari. Google then uses all the data from the queries they capture to power their own $$$ ads business.

• Browsers channel data and set the rules for web analytics: Google Chrome turns all the data its collects on our browsing habits into valuable inputs to the Google Ads behemoth. Similarly, other browsers partner with advertisers and search engines to sell insights around your data.

• Browsers take a cut of internet activities: Google Pay and other payment methods take a cut of transactions.

There are other benefits of being the browser:

• Ecosystem lock-in: As the first point of interaction with the internet, the browser can prompt you to interact with more services in its ecosystem. For example, Google’s Chrome will push to to Google Drive, YouTube, Google Search, and Google Shopping.

Google and Chrome: A parallel for OpenAI and Atlas

In the mid-2000s, Google was just a search engine living in someone else’s browser (mostly Internet Explorer and Firefox…remember them?). Google was dependent on these browsers for access to their user, and they didn’t like that.

Chrome was launched in 2008 to allow Google to own that critical piece of real estate - essentially taking over as the web’s operating system.

• Chrome made Google Search the default search engine

• It integrated with Gmail, Maps, and other Google tools

• It collected data directly

Today, OpenAI is aiming to do the same thing for the AI future.

Why did OpenAI want to launch Atlas?

As we can learn from Google and Chrome, OpenAI’s goal with Atlas is clear: to own how we interface with AI. OpenAI wants to be

• your search engine

• your shopping assistant

• your productivity tool

• …and your interface to all digital interaction

Diving a bit deeper, we can infer how the Atlas browser will fuel a flywheel to make it easier and easier to use ChatGPT, Instant Checkout, and other OpenAI service offerings, solidifying ecosystem lock-in.

More first-party data = More personalization

Atlas enables OpenAI to learn more about us then ever before, powering more personalized advice, search results, and suggested actions

ChatGPT everywhere = Less friction

Instead of going to ChatGPT’s app or website, it is a persistent sidekick as I browse.

Integrated Commerce = Control

Commerce will be seamlessly built into the experience from research through transaction.

Today when I use ChatGPT to plan my toddler’s meals, OpenAI has no idea whether I end up making those recipes or where I get the groceries from. With Atlas, it will have the full funnel view from starting to plan meals to building a grocery list to evaluating options and pricing to checking out. They are meeting me and understanding my needs before I need to open a tab in Google or search in Amazon.

Overall, this timing alongside other commerce-related announcements is strategic. Agents are just starting to be used for actual, practical purposes. At this inflection point, OpenAI’s priority launch signals it knows that whoever owns the browser owns the future interface for AI commerce, search, and action. It also positions OpenAI extremely well to launch paid marketing (as speculated based on recent job postings), becoming a serious challenger to Google.

For a more detailed dive here, watch the official Atlas product announcement - the way they talk about the product closely mirrors these strategic goals.

Testing Atlas out

Now for the fun part! I want to see just how easy it is to go from research/ideation through purchase with Atlas. I am starting fresh out of the box, with only my password manager integrated.

My objective: create an awesome Halloween-themed charcuterie board for my party this Friday.

Step 1: Scoping

Time to complete this step: ~1 minute

Step 2: Drilling down to a list of items

From the inspiration images, I asked for a shopping list. This is pretty spot-on. I then culled the list based on my needs (e.g., removing stuff I already have)

At this stage, I have a clear punch list of what I need to buy.

Time to complete this step: ~3 minutes

Step 3: Shopping for me

I logged into Instacart on a separate tab, enabling Agent to start shopping for me.

I like the UI showing how Agent Mode is working in the background, it’s nice to transparently see what is chosen.

Overall, it summarized its thinking and built a robust cart of items.

I decided not to checkout because there are a few issues around availability and pricing - I’m not sure why certain brands were selected over others. As I continue to use this flow, I will provide more training instructions so it can better understand my preferences.

Time elapsed: 4-5 minutes

Observations

Overall, I like the balance between agent and human actor in this flow. There are still some kinks to work out.

• It asks permission A LOT before taking actions - I imagine it’s better to over-ask and then correct later as people get more comfortable with agentic flows.

• It’s not very fast - seems pretty much the same speed as a human shopper searching for each item. But I can let this run in the background as I do other things.

• I had to start a new conversation to enable Agent Mode for the first time, and it forgot some of the details we had aligned on in the other conversation about what I need/didn’t need.

• It doesn’t know my preferences. While this is understandable (and I can train it to know my preferences moving forward) this seems like a barrier to adoption for less AI-knowledgable people. I think Atlas should prompt users to enter preferences up front. Overall, there is a need for an “onboarding” flow for Atlas (beyond what little is there) to show how to enable key use cases, like integration with Google suite (calendar, email, etc), syncing password manager, doing your first agentic task…this will be critical to get newbies comfortable.

• I can interact with the Instacart site fully within the Agent workflow, which is very cool. This allows me to edit the cart, add/remove items, and checkout on my own at any point.

• There were issues with SKU availability - I’m not sure why it would select items that aren’t available in my zip code, or wouldn’t check that before finalizing the cart for me.

• Within one conversation, you can use Agent for some pieces (like assembling a list and buying) and Chat for other (researching), but the handoff between is a little bit clunky. In future, I imagine they’ll default to a “selected mode” based on the type of query.

• The only option for online grocery delivery is Instacart. This makes sense because OpenAI hired Fidji Simo from Instacart recently to be CEO of Applications. Instacart is the first grocery provider fully integrated with ChatGPT and Atlas - though I imagine many others are evaluating this as well.

This is week 1 of the Atlas browser, and these Agent flows are still in beta. With that in mind, I see a ton of potential here to change how we shop.

How have you used Atlas or other agentic browsers? Hot takes? Let us know!

Leave a comment

Next time, I’ll revisit my Value x Vibes framework with some of these new agentic commerce capabilities in mind.

Until then…stay curious!

-Melina

One thing I’ve noticed: I am more excited about work these days. I’ve always liked my job - good coworkers, fun problems to solve, and free lipstick doesn’t hurt. And recently, the business problems we’re tackling have become a lot more interesting - evolving from should we turn the checkout button from gray to a darker gray? or what performs better - 15% off + free shipping, or 20% off? to what does the future of the brand website look like? Based on the pace of AI evolution, I don’t think things will slow down any time soon.

Friends outside of tech tend to see AI as a threat—especially to “white collar” jobs. Meanwhile my friends in tech are usually more upbeat—though you could argue that we’ve just drunk the Koolaid.

For now, I land firmly in the “optimist” camp. AI isn’t a miracle cure for corporate drudgery, but I think the potential is there to make many of our roles more fun and to free up avenues for career growth.

In this piece, I’m reflecting on corporate marketing roles specifically - as I have lots of experience here as an employee and manager. While I have my own predictions for other areas of the job market, I’ll leave those hot takes to others.

Escaping the grind

Anyone who has worked in corporate life knows the soul-crushing repetition that Mike Judge captured in Office Space. The protagonist is stuck in traffic, hassled by middle managers, and constantly berated about mis-filed TPS reports.

But this film is a time capsule, reflecting the technology of its day. Back in 1999, when the film came out, most companies weren’t using email, internet tools, or automated workflows. Within a decade a lot of those pain points had faded. Reports went digital. Remote work became possible. Whole categories of office drudgery disappeared.

Of course, tech also gave us new kinds of digital B.S. (logging endless Jira tickets, who can relate?). But when applied well, it tends to free us up from the most mind-numbing tasks. AI will create another shift in what it means to “work”.

We’ve Been Here Before

Back in 1930, the economist John Maynard Keynes made a famous prediction: thanks to technology and productivity gains, his grandchildren’s generation would only need to work 15 hours a week. By 2030, he thought, the “economic problem” of scarcity would be solved, and the challenge would shift to how we’d use all that free time.

Keynes wasn’t totally wrong—we got the productivity boom he imagined from industrialization (and later the internet). Where he overshot was thinking we’d harness that productivity to buy ourselves more leisure. Instead, we raised the bar for what a “normal” standard of living looks like and just kept grinding. Between housing costs, ambition, and the treadmill of capitalism, the 40+ hour workweek is still very much alive.

But the lesson is important: every big wave of technology (electricity, the internet, now AI) has sparked both optimism and fear. Optimism that drudgery will disappear. Fear that jobs will vanish. The truth usually lands somewhere in the middle.

AI feels like another Keynes moment. It could eliminate whole categories of tedious tasks—just like the internet killed TPS reports. Work evolves, but it doesn’t go away - it’s up to us to ensure that what replaces today’s digital drudgery is better for us as workers and ideally, is more fun as well.

The return of the generalist

Over the past few years, I noticed something interesting about marketing job descriptions - they were getting more and more specific.

Titles like Paid Media Specialist – Amazon or Community Manager – TikTok became the norm. The thinking was: if you haven’t worked in this channel before, you won’t succeed. This mindset makes it harder for managers to hire promising people from different backgrounds, and harder for employees to grow beyond a single niche.

This was driven by increasingly complex technology stacks that required more expertise to effectively manage. Each walled garden developed its own playbook, with a whole glossary of acronyms. Running Google Ads is genuinely different from running Amazon Ads. Google requires Quality Scores, feed management, and value-based bidding. Amazon, by contrast, is all about the buy-box, SKU-level contribution margin, AMC queries, DSP setups. And TikTok - that’s a whole other story.

Now I’m seeing the pendulum swing back. AI makes it easier to pick up new tools fast and fill in gaps where jargon or complex process would have been hard to learn. That creates room again for generalists—people who can connect dots, think creatively, and move across channels. At its core, marketing is about understanding our consumers and telling them resonant stories about our products. I believe a wider, more multidisciplinary approach will result in better outcomes and will offer more intellectually stimulating roles in our industry.

Why breadth beats specialization in the AI era

David Epstein makes this case beautifully in his book Range. His argument: in “wicked” environments—messy, unpredictable domains like business, science, or culture—breadth outperforms depth. Specialists thrive in “kind” environments like golf or chess (clear rules, fast feedback). But in messy systems, it is generalists who draw from diverse experiences, connect distant dots, and make analogical leaps are more innovative and adaptable.

Marketing lives squarely in the “wicked” camp. Consumer behavior shifts constantly, and channels rise and fall. The teams that win aren’t necessarily the ones with the deepest specialists, but the ones who can remix ideas across domains: apply a TikTok creative testing loop to email, take a retail readiness framework and map it to PDP copy, borrow an idea from brand strategy to improve feed architecture.

AI accelerates this shift. By compressing the learning curve of new tools, it lets people sample more broadly, faster. Suddenly, the skill of moving between contexts—what Range calls developing “match quality”—is more valuable than grinding deeper into one channel forever.

6 Ways to Thrive in the AI Era of Marketing

A few tactics we can all try to enjoy our inevitable transition into the AI era:

1. Get your hands dirty. Play with AI tools (lots of them are free) in your personal life and for simple work tasks.

2. Stay tool-agnostic. Anchor your skills in problem-solving, not in mastering one platform’s buttons.

3. Spin up side projects. This newsletter is mine. For you, it could be a Shopify store, a podcast, or a Substack. Starting something has never been easier—so just start.

4. Raise your hand. Volunteer for work slightly outside your lane. AI lowers the cost of entry; trust yourself to figure out the rest.

5. Apply ideas from other fields. Borrow lessons from other industries—or even your hobbies—and map them back to your marketing priorities.

6. Think bold, not incremental. We’ve been trained to optimize every detail and minimize risk. Use AI as a nudge to experiment bigger, faster, with less fear of being “wrong.”

AI won’t solve the “future of work” for us. It just gives us the tools. What we do with them—whether we make our jobs smaller or more interesting—is still a choice.

Leave a comment

I write a lot about how the e-commerce stack will evolve as shopping becomes agentic. I’ve really enjoyed ’s Retailgentic podcast to keep me up to speed and thinking about this space during my commute (not sponsored - just love it!). The recent episode with Karl Haller from IBM highlighted the prediction that ~25% of e-commerce will be agentic by 2030. This implies ~4% of total retail, as e-commerce penetration has hovered around 20% of total retail for the past few years. While not huge, agentic shopping will become a real (and growing) sliver of how we shop in the near future.

This forecast is in keeping with my Value x Vibes framework - agentic commerce is already starting to penetrate certain categories (like grocery, travel) and will expand from there. Some of these early categories may see significantly more than 4% agentic share by 2030.

This raises a lot of questions about the role of brand. “Agentic” shopping is distinct from getting recommendations on ChatGPT. To be agentic, the tool acts on the consumer’s behalf to make a purchase. Today, agents suggest a few products that fit our parameters and then take guidance from us on those options before purchasing. In future, we can provide more guardrails upfront (or these will be learned), and we’ll need an ever-lighter touch from need to purchase. In this world, brands lose a lot of control.

A few questions on my mind:

• What is the role of top-of-funnel marketing in the agentic age?

• Will loyalty programs still be relevant, or will we see a race to the bottom on price?

• Will we still need brand websites?

• How will brands “market” to agents?

Today we’ll talk about loyalty.

Let me know which of these (or others!) are top of mind for you in the comments, so I can cover in the future.

Leave a comment

Loyalty programs are everywhere

Loyalty programs aren’t new - they’ve been around since the barter economy. In the village marketplace, the baker might slip the butcher an extra roll with the promise, “Come back next week, and I’ll make it worth your while.” Whether it’s bread, punch cards, or points apps, the core idea has always been the same: reward repeat business and keep customers coming back.

A few key reasons brands maintain loyalty programs:

• Build a relationship with customers and drive brand love

• Steal share of wallet from competitors and drive incremental purchases

• Make the customer relationship “stickier”, and increase switching costs

• Gather customer data that can be used to personalize the shopping experience

Loyalty programs buffer brands from needing to compete solely on price and convenience. In NYC, there are coffee shops on every corner. Yet, I choose to walk further to Starbucks and spend more because I love their loyalty program.

Subscribe now

Loyalty at risk

In the agentic era, when a shopping agent decides what to buy on your behalf, the parameters driving that choice may never be visible to the consumer. Price and convenience—usually measured in speed of delivery—are almost certain to be among the top inputs.

For less differentiated categories - paper towels, batteries, cleaning products, or plain white T-shirts—brand may simply disappear. The shopper won’t be scrolling through a site, weighing brand names, and adding a familiar product to their cart. They’ll just say, “I need more paper towels,” and the AI will select whatever meets the criteria at the lowest cost and fastest shipping speed.

This shift puts traditional loyalty programs in a precarious position. For decades, these programs have relied on brand visibility and direct customer engagement to drive repeat purchases. If customers no longer visit your website, open your app, or see your promotions in-store, the levers that once powered loyalty—points, perks, and personalized offers—may never make it into the AI’s decision-making framework. In this world, the benefits of loyalty programs could shrink dramatically, especially for brands that live in categories where function outweighs identity.

Future-proofing loyalty programs

Some brands are already leveraging GenAI to make their loyalty programs more compelling. Starbucks - my personal favorite - has started using an AI-powered analytics tool called DeepBrew to personalize offers more precisely than the generic segments many of us are familiar with. This looks like:

• A push notification a few minutes before a customer’s usual order time with a special offer on a new drink

• A compelling offer aimed at infrequent shoppers to draw them back in-store after a hiatus

• A food/drink pairing offer informed by the customer’s order history

These tactics seem to be working. “In the quarter, Starbucks saw its rewards program 90-day active member base in the U.S. increase to 34.3 million, up 13% year over year, amounting to about 4 million new members.”

I’m betting Starbucks is already thinking about preparing their loyalty program for agents - and I look forward to following along.

What does an AI-native loyalty program look like?

The new wave of loyalty programs will be optimized for the agentic era. To do this, loyalty programs need to evolve in 3 key ways.

Show up in agentic workflows

How does it work? Brands must expose their loyalty program data - points balance, perks, tier benefits - through APIs. Shopping agents can pull this data in real time when evaluating purchase options, just like they check price and shipping speed. This ensures that the value of loyalty perks - not just sticker price - will be used to determine an agentic recommendation. In our coffee example, an agent will factor in Starbucks’s free pastry offer when comparing against ordering my daily cold brew from Dunkin.

While still in it’s early days, Loyalty-as-a-Service, or LaaS, will start to show up in more conversations as agentic conversations mature.

Dynamically offer rewards tailored to each agent interaction

How? Building on LaaS, loyalty programs must themselves be agentic - permissioned to offer different rewards in real-time while “negotiating” with a shopper’s agent. This sort of logic will be akin to how many predict that branded agents will offer personalized pricing and promotions. While guardrails must be established, truly personalized offers in real-time will delight the end consumer while converting a sale and building the relationship.

Dunkin will counter Starbucks’s offer of a free pastry with an offer of the customer’s favorite donut.

Use predictive capabilities to pre-empt agentic bartering

What’s better than winning a negotiation? Never needing to negotiate at all. Best-in-class AI-native loyalty programs will pre-empt a consumer ask by pushing recommendations to the shopper’s agent proactively. If I get Starbucks every weekday, why wait for my agent to make a request? Starbucks’s loyalty agent can lock in a week of coffee orders on Sunday.

I’m excited to see early versions of AI-native loyalty. A tough thing about writing about agentic commerce is that many of the capabilities are too early to find real-life examples. While I’m convicted that these shifts are happening already, it can feel “pie in the sky” to talk about capabilities that don’t yet exist. I suspect a company like Starbucks - which has long been a digital leader, and has products which are frequently consumed by a loyal customer base (meaning they have rich customer data profiles) - will be early to the agentic game. I’m sure lots of SaaS startups are also jumping on this idea. Time will tell!

If you’ve seen anything like this, or if you’re building in this space, please reach out - I would love to chat!

Stay Curious,

Melina

Many people warned me that traveling with a toddler is exhausting. They are 100% correct. The flights to and from Italy tested us (and probably all the poor passengers in our vicinity - sorry!). But it was well worth it. My son loved being in the mountains, trying Italian food, and exploring the playgrounds that are much nicer and much more numerous than I’m used to in NYC. I was also in my happy place with no cell service, not thinking at all about anything digital for a full week. Upon returning, I’ve been playing catch up. But I’m back!

Back in February, I asked AI tools to complete some shopping tasks on my behalf. What followed was a deep dive into the early capabilities of agentic commerce. I would love to revisit these capabilities in the next few weeks. Today, I wanted to take a similar approach and catalog my thoughts on AI answer engines for product search. I have written before about how the multimodal (especially visual) search capabilities inherent in AI answer engines will unlock richer product discovery for shopping. Today, I’ll put this to the test with some queries that are meant to challenge these capabilities.

I’m particularly interested in the fashion category. Fashion choices are highly subjective and nuanced, and traditional search has struggled to provide the curated, personalized results that consumers seek. Fabric quality, silhouette, or specific style preferences are hard to distill to type into a search bar. Daydream raised $50M to solve this problem, and Google has invested billions in Shopping capabilities. Unlocking a search experience that feels “magical” could turn traditional search on its head and revive the struggling luxury fashion e-commerce category.

What are AI answer engines?

Now that all e-comm platforms from Pinterest to ThredUp are injecting AI into search, which platforms “count” as AI answer engines? AI answer engines have the following features:

1. Provide direct, targeted answers vs. a list of links, like old-school Google

2. Are conversational, allowing the user to refine the response

3. Demonstrate multi-source reasoning - collecting inputs from multiple sources to provide a response

4. (Emerging) Answer engines like ChatGPT “remember” things about the user from previous engagements and can use these inputs to further personalize results. Not all answer engines have this ability today, but I believe this will soon become part of the standard as the tech advances.

By this definition, Google is a hybrid. Traditional googling still yields link-based answers, but more and more frequently, users see AI Overviews and other AI-driven content, which resembles answer engines. Google also offer Gemini, a traditional answer engine.

According to Orbit Media as of May 2025, ChatGPT and Google Gemini are used the most.

Interestingly, the starting point for all of these tools looks very similar. And nothing points users to a shopping use case. I’m curious how these interfaces will evolve as more users turn to these tools to shop.

In contrast, Daydream is fully shopping focused and provides prompting around how to search.

Shopping prompt challenge

To test out the top 3 answer engines (ChatGPT, Gemini, and Copilot) plus the new Daydream tool (which is specifically focused on shopping) I wanted to come up with shopping challenges that I think could stump the tools. Here are some actual things I’m searching for right now that I haven’t had much luck with (yet).

Prompt 1: A wedding challenge

I have a wedding in Telluride in early August. The attire is semi-formal with western-chic attire preferred. Please suggest some wedding guest dresses that are my style. Midi length preferred.

Why I think this could be challenging: Parsing wedding guest dress codes is notoriously difficult for humans. I’m curious how these tools will fare in curating appropriate dresses that also have a “western-chic” element.

Results:

❌ ChatGPT

Several of the recommendations were way too casual for a wedding. The first recommended option was white!

🟡 Copilot

The dresses seem reasonably appropriate, though not my style. I don’t like that the results don’t include images. Also, upon clicking the link, I am directed to a PLP instead of a PDP - so I need to search again on the retailer site to find the recommended style. Lulu’s was recommended multiple times - perhaps they are doing something right with structured data and/or piping product data directly to Microsoft?

🟡 Daydream

They did a good job showing midi dresses that are appropriate for a wedding (nothing white!) but the black options aren’t really appropriate for August. Their tool allows me to follow up and ask about other colors, which helps. I expected more unique options from the “western-chic” ask, since Daydream is all about apparel e-commerce.

❌ Gemini

Gemini described what an ideal dress would look like (fabrics, colors) but didn’t suggest any actual products. Gemini did this for all of the searches, which was surprising since it is Google’s product - I had hoped it would be the best shopping experience because Google Shopping is already very advanced.

Prompt 2: Fit criteria

I am looking for a cocktail dress that can be worn with a normal bra. Can you find some options in black?

Why I think this could be challenging: In traditional e-commerce, you have to use filters to search for items with specific product attributes. Many shoppers have strong preferences for clothes that can be worn with certain undergarments, have a specific shape (“gathered at the waist”) or similar, but traditional search doesn’t do a good job isolating these items.

🟡 Copilot

The options they suggested are most consistent with wearing a normal bra, but are all over the place in terms of color and formality, not checking several of my search boxes. They cite their sources, and I can see that they referenced lists and landing pages focused on the “bra friendly” aspect and didn’t necessarily triangulate beyond this. Lulu’s is again showing up frequently - the “Bra Friendly Dresses” landing page was clearly created to be SEO-friendly and it seems to be working well here, too.

❌ Daydream

The third option is strapless! But these are cocktail appropriate.

❌ ChatGPT

The second one is not normal bra friendly! ChatGPT shared lots of context on bra-friendly shapes, but then selected products that didn’t fit its own criteria. I wonder why.

Prompt 3: Photo inputs - with a twist

A top like this, but in red, blue or a different color (not white)

Why I think this could be challenging: This top has been following me around the internet since I clicked an Instagram ad a few days ago. I love the shape but I don’t like that it only comes in white. I’m curious how well the tools will do with multi-modal input.

🟡 ChatGPT

The suggestions were decent. I like how it talked me through the “thought” process. It is clearly interpreting my prompt accurately. The actual product matches still need work, though.

🟡 Copilot

Several of the options were in line with the ask. The second option was Sold Out.

❌ Daydream

If there was a top that was the polar opposite of the top in the prompt image, it would be one of these.

❌ Gemini

…Did not realize this was a shopping query and got confused.

My observations from testing out LLM search tools

Overall, these tools have a ways to go to crack the fashion code. It is cool that they can handle asks like “midi dress” or color preferences without the need for manual filtering.

Some things I noticed consistently:

• Playing telephone: The tools seemed to interpret my prompts well, but lost the thread when they went to seek out options from retailers. Several of the tools walked me through their sound reasoning for their recommendations, then suggested products that violated their own criteria. I suspect that overly broad or missing product data could be a factor. In a similar vein, these tools surfaced Sold Out items. Perhaps this is part of the rationale to push brands to directly share product data.

• Inability to handle multidimensional asks: Most tools could identify midi dresses, or tops that are bra-friendly. But the ability to find options that met multiple criteria was challenging. The tools seem to be relying on curated lists (LPs, “best of” lists) and then “trusting” that the things on that list are relevant.

• Results get better with more inputs and refinement: For the most part, re-prompting did improve results, e.g. “I like the first option, show me more like that”

• Tools need more inputs to understand my style: There was no magical moment where I thought “wow, that dress is so me!” The dresses checked some objective boxes, but the tools need more inputs on my personal style. Perhaps in future they will connect to social feeds and/or pull from my email to see what I’ve bought in the past.

• Lack of serendipity: In the store when you’re shopping with a skilled sales associate, they might suggest something beyond your specific ask if they feel it would be flattering or work well for your needs. I hope that in the future, these tools will be able to integrate that kind of experience, perhaps by relying on their memory and detecting patterns over time.

This exercise will be fun to revisit in a few months, as these tools will likely improve significantly in a relatively short time.

Have you all tried shopping with AI answer engines? What has been your experience so far?

Leave a comment

ICYMI

Last time, I wrote about my predicted secondary effects of AI in e-commerce. Of these predictions, I got the most feedback around my take that human content - models in ads, high-touch customer service, in-store support - will become a high-touch differentiator for luxury brands.

Disclaimer: The views and opinions expressed in this blog are solely my own and do not reflect the official policy or position of my employer or any organization with which I am affiliated. All content is written in a personal capacity.

Stay Curious,

Melina

It’s been a busy few weeks for the Flabiano Fam - we celebrated my son’s first birthday, juggled a few works trips, co-hosted the annual gala for Brooklyn Greenway Initiative (a fantastic local nonprofit for which I serve on the board - check them out!) and are prepping for our first family vacation (TBD on how restful this will be). We will be headed to the Dolomites in early July for a week of hiking and birdwatching, baby in tow. Send any recommendations - and tips for flying with a baby - my way. I hope you all are enjoying your summers too.

Now on to the topic of the hour/year/decade…AI.

I’ve written a lot about how AI will shape e-commerce, introducing an adoption framework and sharing how I tested AI Operator and Perplexity’s shopping capabilities to buy things on my behalf. One thing I love about e-commerce is the dynamic nature of the industry. Technology shifts always keep things interesting - and the past year has been super interesting.

A few near-certainties are already emerging about how shopping online will change.

These have been well-documented by pretty much everyone in the space. I especially loved this recent piece from , which speaks to several of these developments, and this one from .

• We will use AI answer engines - like ChatGPT, Gemini, Perplexity, Amazon’s Rufus, etc - to search for and discover products. We will be able to shop through these interfaces, and may never have to visit a brand website (unless we want to). Search is already moving to answer engines - 4% of queries are starting here, and that will grow fast - with one estimate suggesting 75% of revenue will come from LLMs by 2028 (!!!)

• AI will be a brand gatekeeper - just as Google ranks search results, AI tools will sift through options and only recommend a selection. With AI, brands are up against a tougher challenge - customers can’t just scroll for more results. They may only see one or two recommendations that are “vetted” by the answer engine.

• Agentic commerce will take over utilitarian shopping tasks like booking flights, making reservations, and ordering groceries. These synthetic customers will have their own preferences and decision-making frameworks. Over time, they may take over more nuanced, complex purchasing decisions as well.

• All digital experiences will become much more personalized, in the interest of driving us to convert. This means unique emails, online imagery, product descriptions, marketing assets, and more - if not tailored individually, then on the segment level for highly detailed customer groups.

• Websites will change how they are structured to better appeal to agentic traffic (SEO is being reinvented). It’s unclear how that will look to human users (and may be largely on the backend)

Given these shifts are already happening (albeit in the early days) I challenged myself to think more deeply about what the next order impacts will be for shoppers and for brands. This is the stuff I think is really cool - check out my predictions below and let me know your thoughts in the comments!

Let’s get into the predictions.

Real-time product development

Imagine my AI Agent is buying me a dress for an event. It “says” to the brand “hey, I would buy this dress for Melina, but she prefers black trim instead of white. Do you offer that?” The brand can decide to say yes or no. To capture the sale, they could say yes, and facilitate a custom order. If they say no, they can capture the data and, if multiple agents request the same thing, inform future product creation. They could also put the agent on a “notify” list - pushing out updates when relevant products do become available.

While complex, the ability to increase product flexibility to convert a lot more agentic sales could be really valuable.

In the supply chain world, postponement (waiting as long as possible to assemble a finished good from components) has been a big focus for years, as it reduces waste (costs) and enables more responsiveness to demand. I suspect we will see more focus here, as the influx of data from agents will inform assortments in real-time. This has implications for global supply chains - perhaps it encourages more near-shoring and local production to be closer to the user.

Shadow SKUs

There will be a long-tail of SKUs available to AI agents that normal (human) customers never see. They will remain “agent-only” and will be set up as “behind the scenes” products with a lower level of investment behind them first. If they are popular with agents (acting on behalf of customers) this signals early proof that they will work at-scale - so brands can buy in more and do a wider-scale launch and marketing campaign to attract human shoppers.

For example - in the shoe category, there may be tons of variations of heel height and color available to agents, but human customers see a more narrow assortment.

“Real humans” as a mark of luxury

My husband travels a lot for work and has status with Delta. When he needs to change a flight, he is immediately connected to a human who helps him complete the request. On the other hand, as a no-status nobody, I sit on hold for hours and interact with machines.

As AI voice and chatbot tools get better at replacing customer service tasks and automate lots of simple questions (where is my order?, can I get a refund?) we will come to enjoy the simplicity of those flows. But I suspect there will be high-touch cases where brands will invest in real humans - especially in the luxury space - to differentiate. The future-state will likely be a blend of both - injecting humans at critical moments for high-touch service, then automating flows using tech to provide status updates and other programmatic aspects to drive consistency.

On the marketing side, we are already seeing luxury brands leaning into “real” and authentic marketing, in contrast to increasingly AI-generated content. Burberry’s 2024 Holiday campaign was emblematic of this trend - it featured real VIP customers modeling their clothes. In the future, we will see “Real Human” badging or designations akin to “Made in the USA” and other quality indicators.

Personalized pricing - and prompt engineers who hack the system

Brands will do more with personalized pricing - charging different customers different prices - as agentic commerce accelerates. In traditional e-commerce, it’s a bad look when customers see different prices on your website for the same thing. Still, brands action this via shipping fees, promotions, early-access pricing, and other techniques. With agentic commerce, it is easier to personalize pricing given the individual nature of each query. Perhaps one customer gets a deal on a bundle. Another is charged more for an item because it’s the last one available. Yet another is given a special discount because they are on a strict budget. Sophisticated brands will experiment with ways to add margin to orders, especially if the customer’s agent isn’t price-sensitive.

This will lead to a new wave of prompt engineers who hack the system, experimenting with prompts and strategies via their agent to solicit the best deals from brands. For example, they may identify certain periods when brands are trying to hit plan and are more willing to sell goods at a discount - and inundate brands with queries during this time trying to seek the best deal. Then, they’ll turn around and share the tactics on Reddit - or behind a paywall. This will make offer stackers and deal aggregation sites look like child’s play.

The inversion of discovery

Instead of consumers searching for products, products will seek consumers. My AI agent knows I’m planning a vacation to Italy. I will no longer need to proactively search for hiking backpacks; instead, my agent will present me with a packing list and suggested products.

Today, a lot of marketing is built around the response to signals that consumers put out into the universe - actively searching things in Google, pinning on Pinterest, or watching Tiktok videos about a topic. With agents taking on the role of information curator, how will traditional marketing know to seek me out? Ads will become less effective.

Instead, marketing will reach out to my agent. My agent will put feelers out into the ether, seeking inputs for my vacation. How brands set themselves up to recognize, process, and respond to those “feelers” will determine who will win or lose in the next e-commerce era.

Evolution of creators and influencers

In a world of infinite choice, curation will be more valuable than ever (as I’ve written about previously). Consumers will prompt AI tools with queries like “I’d like some wear-to-work options in the style of Influencer X.”

Currently, influencers compile recommendations for us because they monetize the listings via affiliate links - a lucrative operation. If AI tools can simply absorb their style aesthetic and recommend similar options without sharing any affiliate revenue, it will have a dulling effect on content creation. I am curious to see how the creator ecosystem will evolve in this new era. Brands will still be willing to pay influencers. But the mechanism to attribute the sale when it is passed through LLMs is TBD.

Being offline is the ultimate luxury

One consistent theme across all these predictions: lots of research and admin tasks we do today will be automated.

I expect to see the glorification and growth of “offline” hobbies and activities, both as a response to the increased reach of technology and as a result of automation freeing up our time.

I love hiking, birding, and reading - and I already see increasing interest in all 3 among younger generations. Just as COVID spawned bread-making, knitting, and puzzling booms, this era will generate lots of offline activities. Smart brands will step in to capitalize.

Personally, I can’t wait!

Thanks for reading! Do you agree with these predictions? What am I missing? Let me know in the comments.

Leave a comment

Stay Curious,

Melina

What’s your favorite store experience? Let me know in the comments.

Leave a comment

But too often, going in store is an awkward encounter where I do a quick lap of the near-empty store under the silent gaze of a bored associate. On the slim chance he or she tries to engage with me, it’s a cursory “can I help you with anything,” and nothing more. Not magical!

E-commerce is growing at 13% annually, according to eMarketer, slowly eroding retail sales. When many of us imagine our ideal community, we picture mom and pop shops mixed in with our homes, schools, and workplaces. We picture popping into a specialty grocery after work or grabbing a bottle of wine from a local place where we know the owner. We want to spend our rainy weekends browsing fun shops – like thrift stores, fancy department stores, or cute home goods shops. In reality, it’s becoming harder for these places to stay in business. Many neighborhoods are seeing retail closures accelerating with those that stick around morphing into the same cast of characters – Starbucks, CVS, another CVS.

There are principles brands and retailers can follow to effectively plan their physical footprint for success in today’s environment – and strategically pair e-commerce tools with their physical presence to meet their goals.

I see 4 distinct in-store shopper missions - and each aligns with an assortment strategy.

For each mission, brands can use e-commerce in different ways to complement the physical store experience.

This framework can help brands decide how to invest in physical retail and supporting omnichannel efforts - and where to avoid spending.

Let’s break it down:

The Retail Positioning Matrix

The Hobbyist

This is one of my favorite categories of retail and one that has persisted (when done right) because the core user visits frequently.

The hobbyist is looking for a trusted destination for everything related to their activity of choice - running stores, knitting shops, ski shops, and more fit in this category.

The stores must have deep inventory of a wide set of products that are all geared toward a specialist activity. For a running store, they want shoes, apparel, cross-training supplies, nutrition options, sun care and more - to keep the customer coming back.

How these stores should use e-commerce:

• Publicize events that are hosted in-store – use online touchpoint to collect consumer data and build the relationship

• Build authentic community - for example, a running store can encourage its audience to sign up for local races and share content online.

• Offer auto-replenishment and/or buy online, pickup in store for convenient repeat purchases

The Research-First Shopper

I am researching an expensive or custom purchase, and I want to see/touch/feel it before buying. A quintessential example of this is your mattress showroom.

This store needs a wide option set for customers to research, but can get away with little or no inventory per SKU.

I don’t expect to leave the store with my item. But I do expect free, fast shipping.

Examples in this category: baby supply showroom, furniture showroom. An interesting example that crossed my desk recently – a custom jeans consultant/fit store!

How these stores should use e-commerce:

• Online appointment booking

• Pre-appointment education via text, email, and online content

• Add-on services – for a furniture showroom, can they provide easy-to-schedule dropoff services?

The inspiration-seeker

I have a case of the “shops” and I want to bop around and buy something cool – inspire me!

These stores should offer a tailored, curated assortment and can benefit from frequently changing inventory to keep the shopping experience fresh. In some cases - like thrift stores - items may be one of a kind!

Examples in this category: thrift and antique stores, boutiques/ “shoppy shops”, gift shops, luxury stores

How these stores should use e-commerce

• Consider live-shopping solutions to bring the physical store experience into the digital world

• Engage in brand-building via social media

• Showcase newness and tell the story behind the products

The Need-it-Now Shopper

OMG! It’s my kid’s birthday party this weekend and I haven’t bought any decorations yet. It’s too late to order online. What can I do?

The ideal store experience offers deep inventory of each SKU, but may have a relatively narrow selection. I can confidently shop for my last-minute decorations knowing that they will have what I need, but my options won’t be as vast as Amazon. However, by maintaining a smaller footprint, these “need it now” shops can be located conveniently where we live and work, for easy access.

In this case, I’m desperate. The store can probably mark up the items a bit – and in fact, I expect them to.

Examples in this category: hardware stores, party supply stores, pharmacies

How Need-it-Now stores should use e-commerce:

• Invest in rich store locator functionalities with inventory visibility that allow customers to browse what’s in store so they don’t waste their visit and leave disappointed

• Offer pickup and same-day delivery

• Use AI-powered tools online to answer customer questions on products that best address their needs, allowing them to build a targeted shopping list

• Offer complementary services to enhance ease and convenience – for example, can party supply stores connect me to caterers or photographers for my event? Can they charge a bit more to assemble the balloon arch for me prior to pickup?

• Leverage data from online shopping behavior to inform assortment at hyper-local small retail boxes

Why department stores are failing

Department stores are one model of physical retail that has been particularly troubled. In the past two decades, department stores have lost serious ground:

• U.S. department store sales dropped by over 40% from 2000 to 2023 (U.S. Census Bureau).

• In 2020 alone, J.C. Penney, Neiman Marcus, and Lord & Taylor all filed for bankruptcy.

• Macy’s has closed more than 300 stores since 2015 and plans to close 150 more by 2026, pivoting toward smaller-format concepts.

• According to Coresight Research, the number of U.S. department stores declined by nearly 50% from 2010 to 2023.

• Saks is about to go bankrupt

Department stores are failing because they are attempting to serve each of these shopper types under one roof - and that’s impossible.

• Hobbyist: Department stores cannot compete with the breadth and depth of specialized product that hobby-focused stores offer. They have a selection for a variety of needs, but dedicated hobbyists will shop elsewhere.

• Inspiration-seeker: Department stores are theoretically designed for browsing, but often they are segmented by brand in “shop-in-shops” that are not truly driving inspiration and fun

• Research-first: Economic challenges have resulted in lower staffing levels, making it difficult to replicate the high-touch service model of the showroom

• Need it now: Department stores can address this need in some cases, but may not match the hyper-local convenience of a “need it now” store.

Taken together, department stores are saddled with huge footprints, but aren’t really serving any customer type entirely.

It’s not surprising when viewed through this lens that this model has failed to survive in the digital era.

Some success stories using the framework

• Nike leaning into hobby oriented shopping with events done right

• Whole Foods experimenting with new formats to better serve the “need it now” customer

• The RealReal boutiques maintain a “treasure hunting” experience supplemented by online

Smart retailers are rethinking their physical presence around specific customer missions, not generic foot traffic goals. Whether it’s last-minute needs, guided research, inspirational browsing, or community-driven hobbies, each model thrives when paired with the right e-commerce tools and expectations.

The winners won’t be the ones trying to serve every customer in one space—but those who double down on doing one thing exceptionally well. For brands, this means getting clear on your store’s purpose and aligning everything—from layout to staffing to your online experience—to support it. When done right, physical retail can still feel magical.

Stay Curious,

Melina

I’m thinking I should do what comes most naturally - hate.

As someone who works in e-commerce, I am quite familiar with the minutiae of the online journey. So when I shop in my spare time and see one of these things, it makes my blood boil. These are my icks - things that won’t make me file a report with the Better Business Bureau, but will make me think twice about shopping again. Let’s dive in.

Premature asks for product reviews

There’s nothing worse than getting one of these emails before your product has even arrived. I’d love to tell you what I think about my order, but unfortunately it’s still in a distribution center in Pennsylvania!

All this does is make it very obvious when a brand has slow shipping.

Dolce Vita wasn’t really an offender here, though they sent me this just ~1 week after my shoes arrived. Since these were heels, I hadn’t worn them yet - I’m normally a sneakers girl. Brands should play around with the triggers for these emails as some of the best reviews come after the customer has really gotten to know and use the product.

Popups that are hard to click out of

Nothing will make me bounce faster than one of those “Enter your email for 15% off” popups that takes up the entire screen on mobile and is impossible to exit.

Cuts Clothing is unfortunately an offender. While they have a visible “X,” it falls so close to edge of my phone screen that I can barely click it. The timing of this popup is almost immediately on page load, so I can’t even tell if I like the product before I’m asked for my information.

I have started to see brands backing away from these invasive popups, especially as >70% of sessions are on mobile (clicking out of these on a phone is more annoying than on a desktop). Bonobos has their sign-up offer nested and in the top rotating banner - very visible, but not as disruptive.

and a related ick…

Popups that ask for too much information

First brands wanted my email. For 20% off, okay, it’s a fair trade. Next they wanted my phone number. And the table stakes offer has moved down to 15% off. Now, they’re asking additional survey questions, too.

What will be next? Will they want me to donate a kidney for 10% off?

I fully understand the value of this data for brands - the ability to reach customers directly is invaluable. And brands are investing in first party data for many good reasons.

But asking for all of this up front is too much - as new visitors to a site, we haven’t had a chance to become interested in what a brand is offering yet.

In this example, Lizzie Fortunato is one of my favorite jewelry brands. If SMS signup were positioned as a loyalty initiative in a post-purchase email, I likely would sign up for texts. Alternatively, if I’d browsed for awhile and then saw this offer in a PLP tout, I might sign up. I’ve also seen many brands tie Free Shipping offers to email and/or text sign up - this feels like a fair exchange.

Product pages that don’t make it easy to buy the model’s entire outfit

Sezane styled this chic, Parisian trench coat with pieces that seems to be available on the website. But these items aren’t made shop-able anywhere on the page. This is a huge missed opportunity, and so annoying for shoppers!

Tuckernuck, on the other hand, does a great job showcasing the shop-able look and making it easy to see more details on any of the items without leaving the page. Bravo!

Too much packaging

The conventional wisdom in the e-commerce world is that having a fancy “unboxing experience” is a special moment for customers to get to know a brand. This might be controversial - but skip the packaging, and just give me what I ordered!

When I open the box and see all this stuff, I feel wasteful. I also think that the brand’s margins are too high, and that they should lower the price by a few dollars instead!

One brand whose packaging I love is Ghia. It is beautiful, but functional. While they include printed collateral, it has a clear purpose - recipes to educate customers on the product and how to use it. Packaging with a purpose is fine in my book.

Aggressively high shipping minimums

More brands are charging for shipping as fulfillment costs continue to rise. At the same time, my ick is shared by the masses - high delivery costs is the top-cited frustration among shoppers worldwide, according to eMarketer.

I’m all for brands who want to protect their margins on small orders. I don’t expect brands to pay $10 to ship me a $20 item. But some of these shipping minimums feel like highway robbery.

Rachel Antonoff has a $300 minimum on her site, despite having an average unit price in the ~$150-200 range. I believe that a customer purchasing a single item at that price should have the shipping comped by the brand. In this case, I would advise the brand to raise prices slightly to compensate for shipping costs.

What is your biggest e-commerce ick? Do you agree with these? Let me know!

Leave a comment

Stay Curious,

Melina

I spend my days optimizing how we shop for beauty products online. Yet when I am shopping online myself, it often feels like a chore. In the e-commerce world, we use words like “discovery” and “inspiration” - but clicking through filters, popups, and product variants to buy something feels more like drudgery. I love writing Curious Commerce because it lets me explore how we can make shopping online fun. Today, I’m going to break down the key parts of the e-commerce shopping journey and share some of my favorite examples of brands and retailers who are doing it right.

Shopping Journeys: No one-size-fits-all

When imagining the typical online shopping journey, you may picture a very linear experience: a customer goes to a Brand’s homepage, clicks to the category they want to shop, views a few product options in this category, adds to bag, and checks out. In reality, this is just one of many paths a customer might take.

I’ll summarize a few of the common paths here:

• Common for new shoppers

• Common for direct traffic - typing a brand’s URL into the browser

• Often involves filtering in category pages and using the navigation

• Increasingly popular with paid acquisition and influencer campaigns.

• Stronger intent; often mobile-driven.

• Conversion can be high - especially for lower-priced (impulse) products

• PDP content is critical - ratings and reviews, UGC, and other “proof” that product is legit

• Skips browsing in favor of direct purchase.

• Reflects high-intent search behavior.

• SEO and site search performance are key here.

• Shopper may utilize PDP modules to view and compare alternatives

• Retargeting or re-engagement flow.

• Often for returning users or promotions.

These shopping journeys probably aren’t surprising to seasoned online shoppers like yourselves. We rarely sit down in front of our computer and say “it’s time for me to shop now” and type in a brand’s URL. Rather, shopping is integrated into our day-to-day, with an ad spurring an impulse purchase or a conversation with your spouse resulting in a quick Google search to replenish a household item.

Across all these journeys, we observe that some parts of the site are viewed more than others:

View frequency by type of e-commerce webpage

Why does this matter?

Across these journeys, customers expect to use different parts of the site in different ways. Sometimes, these expectations seem to conflict. Customers in the classic “browse first” path get to the product page after reading a lot about the brand on homepage. They want to make an “add to cart” decision quickly. Customers landing immediately on a product page from an Instagram ad may want to learn about the brand before deciding to add to cart - and the product page should also address this goal. Brands are constantly balancing these objectives. Here are a few insights from how customers are shopping today:

PDP is King

As the most-viewed page on most brand websites, balancing conversion-driving content to encourage exploration on the product detail page is critical.

J.Crew’s product pages do a great job balancing shopability and other objectives - like driving upsell/cross-sell.

A few things I like here that drive conversion

• the prominence of the customer rating

• the prominent Add to Bag button

• the clearly messaged sales price

• the model’s height and size called out

• details like fabric composition are collapsed - this can be overwhelming on mobile, and may be more relevant to some shoppers than others

At the same time, this page is positioned to do more than just convert on this particular product

• ability to “shop the look” with multiple styles shown - drives a bigger basket

• relevant alternatives presented, so shopper will continue their journey vs. exiting

• clear navigation in the header that leads to other cateogories

One thing I might invest more in for J.Crew, especially for hero products, is user-generated content. I’d love to see customer images in the clothes.

Custom landing pages are increasingly used to make shopping journeys more personalized

Shoppers may see a custom landing page when they click through the website from a paid ad or an email. This allows the brand to tailor content specifically to a certain customer segment (recipients of the email or ad) without altering the entire site. I’ve discussed in past newsletters how advances in AI will further automate the ability to create and manage lots of custom experiences, including landing pages. In the not-to-distant future, we could all be looking at our own unique version of a brand’s website.

A custom landing page just converted me yesterday. I saw a version of this very cute ad on Instagram.

When I clicked through, I landed on a page dedicated to NYC-focused baby gear from this brand. Adorable! Pages like this are easy to spin up and ensure that users who appreciate the ad are directed to relevant products upon click. However, the brand avoids cluttering their navigation with really niche pages.

Another way to customize is based on occasion. With the Coachella festival this weekend, lots of brands could curate an edit of festival-inspired styles and spin up a landing page to highlight these. They could then direct traffic here to take advantage of the moment.

Customers are bypassing the homepage

With better targeting and social commerce, shoppers increasingly bypass the homepage and landing on a category page or product page right away.

This means brands - especially lesser-known small brands - need to tell their story on these pages, in addition to the homepage. They must answer questions like “what does this brand stand for?” and “is this brand a good fit for me?”

Brightland, the olive oil brand, does a great job of integrating it’s story and benefits on the product page.

A few things I like:

• Brand-building information, like details on their farms, is included in the product details (but collapsed so it’s not too much for a conversion-oriented shopper)

• Inclusion of recipes and high-quality food images lend the product a feeling of quality and healthiness

As the lines blur between discovery, inspiration, and purchase, the brands that win will be the ones who meet shoppers wherever they are—whether that’s on a PDP, on TikTok Shop, or a Coachella-inspired landing page.

Online shopping doesn’t need to feel like a chore—and with the right strategy, it won’t. I’ll be back soon with more examples of brands who are reimagining the e-commerce experience, one step of the journey at a time.

If you loved this piece, share with a friend!

Share Curious Commerce

Stay Curious,

Melina

I initially created this as a workshop I led with colleagues at Pandora Jewelry. I was on the strategy team and created sales reporting that broke out our retail performance using this framework. I then introduced this framework across the company as a standard way of evaluating retail performance, so we could all share a common language. My goal was to keep it super simple, so even the less math-inclined (and I count myself in this group) could gain a deep understanding of the concepts. I still find this super helpful in my current role. Even the most technical, complicated-seeming digital marketing tactics can be boiled down to fundamental retail levers. This one is a little dorky, I but I know you guys are too!

I hope you enjoy!

Introducing: The Retail Tree.

The goal of the Retail Tree is to decompose sales into its most basic parts, allowing you to zero in on specific tactics that you can use to grow your business.

If you sell anything to a customer, you should understand these metrics and how they work.

Let’s dive in.

Let’s start at the top of the tree with Sales. I’m not claiming to be an economics whiz, but I passed the class and do remember one equation from the first day:

Revenue=Price*Quantity

Taking this universal truth and translating it into “retail speak”:

Sales=Average Order Value (Price)*Transactions (Quantity of Sales)

There are two ways to increase your sales. You can:

a. Make more sales (increase number of Transactions) AND/OR

b. Get more money per sale (increase average order value, or AOV)

This isn’t very helpful. “Make more sales” sounds great but isn’t actionable. We need to dive deeper into the framework.

Part A: Increase Transactions

Transactions=Number of customers*Purchase Frequency

Number of customers

One way to make more sales is to sell to more people. Awesome, let’s do it. How do you get customers? First, you need to get in front of an audience. Then, you need to convert people in that audience into buyers:

Number of customers=Traffic*Conversion Rate

This bring us to the root of the Retail Tree. Welcome! Now that we’re here, we are much better positioned to think tactically about ways to drive our business:

Traffic drivers (illustrative examples):

-Hold events in your store to draw people in

-Put ads in the neighborhood to tell people about your store

-Work with influencers in your subject area link to your site in their content or talk about your store

-Pay for Google ads to get people to click through to your site

-Improve online organic traffic through SEO

Conversion drivers (illustrative examples):

-Train store associates to recommend relevant products to customers

-Set up your store so customers can interact with the products and connect with them

-Invest in detailed and visually appealing product pages on your site that answer customer questions about the product (e.g. sizing, usage)

-Offer financing options to reduce upfront barrier to purchase

Purchase Frequency

The other side of the “transactions” branch is purchase frequency. How often does the customer buy? It is typically much easier and cheaper to get a returning customer to buy from you again than to convert a new customer. Brands will do everything they can to push one-time buyers toward their second sale (and beyond). So, we want customers to buy more often (duh!). But how? Here are some strategies you have probably seen in action:

Subscription models: I recently signed up for a free month trial with Classpass, anticipating canceling after a month or two. Now I’m hooked. By dangling the carrot of one “free” month, I have ended up buying 6+ months — and am now a loyal customer. While subscription models aren’t as “hot” as they were during the Birchbox era, they aren’t going away.

Loyalty programs: These programs are engineered to drive more frequent purchases (and can also target other Retail Tree metrics, like Basket Size and Average Selling Price, which we will discuss later). One of the best loyalty programs is Sephora’s VIB program, which uses a tiered rewards system to keep customers coming back.

Offers to drive repeat purchase: I recently tried a new DTC beverage brand and got an email when I was nearing the end of my supply. The email offered $10 off my next purchase, which prompted me to try another flavor.

Planned obsolescence and seasonality: These tactics are so common we don’t even notice them. I remember saving up to buy a pair of Airpods only to see a newer model released a few months later. This encouraged me to replace my pair with the “latest and greatest.” This is called planned obsolescence - when companies intentionally make their products obsolete so you’ll buy something new. Another strategy is how brands take advantage of seasonality and the trend cycle to encourage consumption. This is one reason why the average American throws away 80 pounds of clothes per year (but that’s the topic of another article!)

Part B: Increase Average Order Value

Average Order Value=Average Price*Basket Size

On this side of the tree, we want people to spend more money when they shop with you — buying more, expensive items at with fewer discounts each time they shop.

Average selling price

Average selling price (ASP) is the average price of the goods you sell to the customer. We also refer to this as Average Order Value (AOV). Here are some ways you can keep ASP high:

Upsell: Upselling is getting your customer to buy a more expensive version of your product. For example, when I was a barista at Starbucks in college, I would ask customers if they wanted an extra shot in their drink. My VIP customer would visit daily and order 8 shots of espresso on ice - an expensive and memorable drink that drove our AOV higher.

Limit promotions: A major way retailers increase sales in a hurry is by slashing prices. ASP is the main metric affected by this. Maintaining integrity of full-priced sales whenever possible is the best way to preserve ASP.

Add higher-priced items to your product mix: This is a longer-term strategy for most brands. As you increase the products you offer, keep an eye on the average price points, as this will impact ASP and AOV. Related to this, brands can bundle existing products together to increase ASP - for example, creating a matching set SKU with a shirt and pants that go together and allowing customer to add both to their bag at once.

It’s also worth asking whether you can raise your price, especially when you are first starting out.

Basket Size

Basket size (otherwise known as units per transaction, or UPT) comes from the traditional retail concept of trying to get customers to check out with more items in their shopping basket. The classic example of this is the gum and magazines you see in the checkout aisle of the grocery store, known as “impulse purchases.”

Cross-selling: Cross-selling is offering your customer a different, but perhaps related, product. Amazon has attributed over 35% of its sales to cross-selling, via the “Recommendations for You” section at the bottom of the page. As a Starbucks barista, I was required to ask all customers if they wanted a pastry with their coffee. This is a classic cross-selling strategy.

Add-on items: Similar to impulse purchases in the grocery store, many online retailers are offering add-on items to increase basket size. Shoe companies may offer socks; candle brands may branch into wick cutters and matchbooks.

How to Use the Retail Tree Framework

Know your data

Learn which of these drivers are weak and which are strong for your business. You can compare against historical trends and against competitors. This can help you figure out where to focus your efforts. In one of my previous companies, we found that we were really good at converting people shopping for gifts, but less good at converting people shopping for themselves. This helped us learn to (A) increase traffic of gift shoppers through targeted campaigns to drive traffic to our site around holidays and (B) learn more about self-purchase shoppers to get better at converting them.

Understand the tradeoffs

Driving one lever can (and probably will) reduce performance on others. For instance, if you add inexpensive add-on items to your site, you will see a drop in ASP, but may increase basket size and conversion. Use the tree framework to model out what that means for your total sales.

Test, measure, and learn

You can use this framework to create hypotheses about your business and then go out and test them. For instance, in my women’s clothes company I saw that people were visiting my site and not buying as often as I expected. I suspected it was because my site needed more detailed information on sizing and fit. I experimented with a more detailed sizing chart and saw some improvement in my key metric, conversion. Based on these early results, I feel more comfortable making larger investments in my site.

Observe what your favorite brands are doing

Once you start thinking about the Retail Tree, each e-mail you receive from a brand, every subway ad, and every new product launch is a clue into their strategy.

I hope you enjoyed this tutorial on the most significant Retail KPIs and how they are related. Thanks for reading!

Stay Curious,

Melina

I’ll admit, I’ve been writing “ecommerce” this entire time, in hopes that the time and energy saved from typing “-” repeatedly would lead to improve wrist health and allow me to achieve self-actualization. Unfortunately, I have been outvoted.

Google is still in my camp, finding both ways acceptable. Because I believe in listening to my audience, I will add a dash moving forward. Thank you for your feedback!

Today, we are moving on to the final quadrant of the Value x Vibes matrix for AI adoption in e-commerce. I predict that AI will be used differently for different categories of products, based on what we’re looking for when we shop for these products.

Part 1, covering “Boring” categories, is here. I covered how AI tools will enable automation and personalized research and product comparison to improve the customer shopping experience. These tools will be leveraged by brands and retailers to differentiate and steal share.

Part 2, covering Fun/Expensive categories, is here. I focused on how AI is less penetrated in these niche, expensive categories, but early movers are promising to democratize tools once only available to the wealthy - like the ability to source rare vintage luxury products, access a personal stylist, or even curate an art collection.

Some readers have taken issue with the products grouped into “Fun” and “Boring.” Several people seem to feel strongly that strollers belong in the “Fun” category. As someone who watched hundreds of Youtube and TikTok videos trying to understand what a “travel system” is and how that is different from a car seat, I beg to differ. But this feedback points out something important about the framework. Brands selling a product must understand how consumers shop for that product. Perhaps some customers approach a product as a “Fun/Expensive” purchase and are looking for the Brand to serve the role of Curator. Perhaps others categorize the purchase as “Boring” and look at the purchase journey as a way to maximize utility - seeking product comparisons and research. Many brands may have segments of customers inclined to shop differently - and must offer a suite of tools that enable multiple types of shopping journeys.

Fun/Cheap: Serendipitous moments

In the Fun/Cheap category, we’re looking for inspiration. We’re scrolling on our phones after work and shopping for a cute hair clip, a new book for our upcoming vacation, or a hostess gift. We’re open to buying right now…if we find something we like. In this category, we may not even “shop” in the conscious sense. The shopping comes to us via ads that seem to pre-empt our desires.

Brands are looking to present us with something that feels unique, special, and personal – a sweatshirt with the inside joke from my favorite podcast, or the candle that all my coworkers recommended, but I couldn’t recall the name of. Paid marketing has already advanced to the point where putting these things in front of us feels magical – 55% of us believe our phones are listening to us! And as AI sophistication continues to increase, this will feel even freakier.

AI in Paid Marketing

The serendipitous feeling of discovering the perfect product while browsing online is often powered by paid ads. Brands allocate budget to channels like Google, Meta, and TikTok to show relevant users paid “sponsored” content - in all, roughly 1 in 4 visitors to ecommerce sites come via paid traffic, with search ad spending alone topping $132B in the US.

In the heyday of paid media (2015-2019), brands used third-party cookies - small pieces of data stored on your browser when you visit a website - to track you across different sites and build a profile of your browsing activity. For those of us on Facebook during these wild times, we might remember the perennial print-on-demand ads for hyper-specific tees and mugs that would pull in your birthday and details about you from your Facebook profile. There was a sense of novelty in the early days of data-driven marketing that made products like this feel fun.

In a similar fashion, third-party cookies would compile a profile of you and data brokers and ad tech companies would sell it to Brands so they could serve you hyper-specific ads. In these peak years, paid media spend was growing at 49%, with 88% of ads dollars being spent programmatically.

In recent years, regulations have targeted third-party cookies, and Apple and Google are phasing them out. This reflects a value shift toward privacy-focused marketing. The idea that random internet brands could know everything about us no longer feels harmless - it feels creepy. For many Brands, paid ads have become less effective, and therefore more expensive. Increasingly advanced AI tools are promising to fill this gap with ever more sophisticated solutions.

Predictive Analytics and Targeting

In the world of third party cookies, a shopping experience might look something like this:

• I visit Nike’s website and browse running shoes.

• A third party cookie tracks my visit.

• Later, I see Nike ads following me around while browsing the New York Times. This is called retargeting.

There are a few issues with this approach. Consumers might find it intrusive to have products they’d previously Googled show up all over the internet. I remember a colleague getting flustered when Victoria’s Secret ads showed up on a totally unrelated website while he was screen sharing during a meeting. He quickly protested that he was just buying pajamas for his wife - but we got ammunition for plenty of jokes.

Another issue is that I will keep seeing these retargeting ads, even if I already bought the shoes elsewhere. This is a waste of money for brands.

In today’s world, AI-driven targeting could look more like this:

• I visit Nike’s website and browse running shoes.

• Nike collects first-party data on my visit.

• AI-driven targeting software uses Nike’s data about me and broader shopper data to predict what my future purchase journey will look like and inject the shoes at relevant touchpoints.

  • If I Google anything else related to running shoes, I might see the shoes I previously browsed (targeting based on relevant content)

• Whether I buy the shoes from Nike or not, the prediction engine will phase out their marketing of those specific shoes and evolve to show me products that may be more relevant, based on my ongoing interactions

In short, old-school targeting was focused on what the customer did in the past. AI-based targeting is focused on predicting what customers will do in the future.

Everyone from data science startups to the Googles and Metas of the world are investing in capabilities for AI-driven targeting, as finding a unique edge to reach customers at the time of purchase is extremely valuable. While all brands are already employing some of this technology by virtue of running ads on major platforms, the level of sophistication will continue to increase.

Creative, Copywriting, and Personalization

Imagine you have a small healthy snack brand. You might have a few different groups of people who you want to advertise to:

• Diabetics who need to control sugar intake for medical reasons

• Health-conscious men who listen to the Huberman podcast

• Crunchy moms who want healthy snacks for their kids

These groups will all need VERY different marketing materials when you create ads to speak to them. You and your team will need to create different visual “assets” for each group and different advertising text (“copy”). Now imagine you are running these ads across multiple channels - Google, Instagram, Podcasts - the amount of content you need to create will keep your team busy for weeks!

The promise of AI creative tools, like Adobe GenStudio or Google Performance Max for performance marketing, Canva, Synesthesia for video, and Jasper for content marketing, is that they can adapt your ad creative and copy dynamically for different audiences.

Google’s Performance Max (part of Google Ads) adapts a single set of ad inputs - like assets, copy, keywords, and headlines - into different ad formats best-suited for each Google channel. This saves work for creatives, and from Google’s point of view, makes it easier for Brands to spend money on Google.

Tools like Performance Max aren’t new - budget optimization and targeting capabilities have been around since 2021. Some of the creative updates have come more recently since 2023. Paid marketing has been deeply data-driven for years and this naturally evolved into AI-driven capabilities as the technology advanced. Many of these updates were largely “behind the scenes,” powering things like targeting (who sees your ads) and how spend is budgeted across a digital ad portfolio. Many of us are seeing these capabilities more now that they are maturing on the creative side - affecting the day-to-day work of creative and design teams.

What could the future of paid ads look like?

Since the majority of us already feel like our phones are listening to us - what’s next? This scenario illustrates how advances in predictive analytics, targeting, and personalization on the creative side could come together to power the next phase of digital marketing.

Returning to the Nike example - perhaps Nike has done a great job through their loyalty program, first-party site data, and community-building initiatives and has a bunch of data points on me:

• I’m a runner

• I’ll be running the Berlin marathon in the fall

• I run in NYC and have preferred routes in Brooklyn - in the past, Nike had a partnership with Strava, though I don’t know the details of how data was collected and shared

• Based on my purchase history, they know my general size and style preferences

Knowing all this, Nike could target me with personalized ads leading up to the Berlin marathon encouraging me to buy new shoes. None of this is novel, though few brands really nail occasion-based triggers.

Taking it to the next level, Nike could generate personalized copy speaking to the upcoming Berlin race. It could generate unique video assets to show a female runner in NYC and Berlin highlighting the product in my environment. Without crossing a line into “too creepy,” it could add a level of personalization to the ad that could help me envision myself in the shoes.

Going further, these ads could include chatbot-type functionalities that could let me learn more about the shoes and other related styles within the ad itself. It could let me swap colors on the model or see the key specs summarized.

Taken all together, an ad like this should drive significantly more conversion than a static picture of a shoe.

The bottom line - it’s all about data

One question underscores this scenario: How will Nike get the data to power this ad experience? These AI-driven marketing experiences are only as good as the data powering them. While Nike could theoretically get all of the data above via real, proven touch points - an online quiz, conversations with a store associate, a Nike-sponsored run club - the likelihood of piecing this all together under one clean customer profile is slim. Large marketplaces - like Amazon and Walmart - with millions of transactions per day across categories - are at an advantage. Smaller brands may struggle to glean detailed customer insights from infrequent visits and site interactions with a smaller product catalog.

I expect we will see a lot of innovation in this space to expand access to quality first-party data for brands, perhaps through partnerships. Brands are already focusing on loyalty programs and other sources of valuable customer data. Connecting real-life touchpoints in stores where possible could add another level of richness to the picture brands can paint of customers - but many brands operate stores entirely separately from their online businesses. Streamlining data collection efforts to create a “single view of the customer” will be more important than ever.

It’s ironic that in the “Fun/Cheap” category - the space where we shop on a whim for fun little treats - the situation behind the scenes is anything but whimsical. The seemingly magical ads that pop up on our phones are powered by cold, hard data - and getting that data will be a major strategic priority for brands hoping to optimize their approach to AI-driven marketing in the coming years.

Stay Curious,

Melina

Before we go any further…we have a serious matter to address.

Now, a brief reintroduction for my new readers before diving in this week. I am an ecommerce enthusiast with over a decade of experience selling things online at companies big and small. In that decade, I’ve touched most key functions in the ecommerce world - from digital marketing to supply chain and fulfillment to site operations to strategy & data analysis. Today, I lead Product Marketing at Estée Lauder. This newsletter is an outlet for me to explore why we buy and to get hands-on with new tools and products that I believe can shape the future of shopping.

I also had a baby nine months ago, so there’s never a dull moment in this household.

The Value-Vibes matrix groups products by how expensive they are and how fun they are to shop for.

Last week, I introduced the Value-Vibes framework for AI Adoption in ecommerce. I predict that AI will be used differently for different categories of products, based on what we’re looking for when we shop for these products.

I covered the “Boring” categories, and how AI tools will enable automation and personalized research and product comparison to improve the customer shopping experience. These tools will be leveraged by brands and retailers to differentiate and steal share.

Today, I am turning our attention to the “Fun/Expensive” quadrant - as my husband knows, my expertise as a shopper really lies in this area.

Fun/Expensive: AI for discovery and curation

Of all the quadrants in this matrix, the Fun/Expensive category is the least mature in terms of AI adoption in the shopping journey. This mirrors the lower online penetration for luxury categories vs. mass. In this quadrant, customers are often seeking a more traditional, white-glove shopping experience. I believe that technology can - and will - surpass the experience of shopping in a luxury boutique (minus the complimentary champagne). We will see a lot of innovation in this space as AI tools become increasingly sophisticated and trusted.

A common thesis in venture capital investing posits that successful startups democratize access to products or services that were previously only available to the wealthy. Companies like Robinhood (personal investing) and Uber (hiring a driver) exemplify this. Tools that enable access to and curation of luxury products, when positioned properly, could be similarly transformative.

I need a nice pair of heels for an upcoming wedding. I Google “best comfortable heels” and get a result for the ugliest orthopedic shoes that ever existed. So I look at “best sellers” on Nordstrom and the ones I like aren’t available in my size. I start checking other retailers. Pretty soon an hour has passed and I have a PhD in shoe theory. I still haven’t bought anything. Sound familiar?

I wrote previously about how AI-driven search and visual search tools hold promise to improve this “endless scroll” phenomenon. For Fun/Expensive items, AI tools that serve as a stylist, curator, or concierge promise to cut through the monotony of online search and replace the role of a hands-on sales associate in a luxury boutique. Let’s see how a few companies are approaching this problem.

AI as curator

Daydream

I have been excitedly following Daydream, an AI-powered search engine for ecommerce from Julie Bornstein, who previously founded The Yes and held leadership roles at Nordstrom, Sephora, and Stitch Fix.

Daydream has onboarded over 2,000 retailers to its platform, which allows users to search using conversational prompts and images - much like ChatGPT and Perplexity - but with a commerce-first orientation. As we are increasingly adapting to this way of searching with AI search engines, I can see a clear path to adoption of a new tool that focuses on matching consumers with relevant products - especially a tool that maintains a vetted assortment of brands.

While I don’t have access to the platform’s beta yet, their site gives a preview of the experience:

To succeed, Daydream needs to convince customers to start their searches in their tool, vs. on Google, Tiktok, ChatGPT, or another luxury site. Having a wide enough assortment to return interesting and relevant search results is a key factor - with over 2,000 brands on board, including popular ones like Alo Yoga and Doen, Daydream is clearly waiting to launch until they can check this box with confidence. Luxury ecommerce has been challenged in recent years - these challenges stem from the tough economics of ecommerce, including high return rates. Daydream could be entering a space where there is room to steal share while avoiding some of these pitfalls - as a search tool, they are not fulfilling sales.

Will you use Daydream or another ecomm-focused AI search tool?

Leave a comment

I’d love to explore further if I can get a beta invite 😊

AI as personal stylist

Doji

I previously wrote about Doji, an app that lets you try outfits on an AI avatar. Currently in beta, the app is a creative outlet for building designer looks. The ability to see an outfit styled on “you” could drive conversion and basket size, and in the future, this type of functionality could be built into retailer and Brand sites to enable a seamless try-on experience.

Luxury ecommerce sites

Multi-brand online luxury sites are obvious contenders to offer AI-driven curation and styling. This category has struggled in recent years, likely slowing investment in new technologies. While companies like Farfetch and YNAP tout AI in their press releases as a driver of personalization in search and merchandising, there aren’t any personal styling tools available just yet.

Your digitized closet: Whering & Indyx

Whering is a digital wardrobe and styling tool that allows you to mix and match items in your closet with things you might be considering purchasing. It enables you to see gaps in your wardrobe and get more wear out of pieces you might be unsure how to style.

Currently, it is a big lift to digitize one’s closet properly and ensure the data is clean - e.g., that shirts are labeled with the right color and style. This is a manual process, involving searching for each item or photographing it and adding details. After this is done, you can create outfits.

I can see Whering adding more AI-driven capabilities, which would make it easier to use. For instance, they could automate the cataloguing of items in the your closet - every time you purchase something and get an order confirmation email, that item is automatically added to your closet. Advances in visual search continuously make the cataloguing process faster. Also, they could add AI generation for outfit ideas.

For shoppers looking to optimize their closets, this is a useful tool. I enjoy the creative chaos of digging through my closet for forgotten items and probably would not invest the time to digitize my closet. By staking an early claim in this space and seeking to own the closet database, Whering could be well-positioned as AI capabilities advance.

Indyx offers similar capabilities to Whering, with an added layer of human personal stylists and resale capabilities. I really like the addition of a resale layer here, because it provides a further incentive to digitize one’s closet beyond wardrobe optimization. Viewing one’s closet as a stable of assets that I can choose to hold or sell is quite appealing to me, and would help me further justify my shopping addiction.

I’ll certainly be watching this space.

AI as concierge

The promise of ecommerce is infinite options at our fingertips, yet we often find something we like only to struggle to buy it. In the luxury space, this could be a rare handbag, a designer collaboration that has sold out, a piece of art, or a fragrance that was made in very small batches. The ultra-wealthy may have interior designers to track down vintage furniture or art, or sommeliers to track down rare wines. There are now AI-driven tools democratizing sourcing of rare goods for the masses.

Gab Waller - Sourced By

Gab Waller is a famed fashion buyer and sourcer to the stars - including the Kardashians! - so when she shared in Retail Brew that she’s launching a platform that improves sourcing using AI and visual search, I was curious to follow along.

From the profile:

Speaking specifically for my industry, it’s all about finding more. The issue at hand at present [is] we can only find about 30% of our requests. But I truly believe that every item that I receive a request for exists somewhere in the world. So, creating that technology to find more pieces to service more clients—that’s what gets me really excited. At present, we’re not fulfilling 100% of our requests. I know that we can, but we need the technology to do so.

Interestingly, the way the tool is framed indicates that this may be a proprietary solution for Gab’s team to leverage, at least initially. This speaks to the push-pull with new technologies like this - with Gab’s expertise in sourcing, she holds a unique place in this value chain. How can she leverage this technology to expand her business without losing that edge by granting access to everyone? Will she license this technology to other businesses?

Takeaways - Fun/Expensive

I had fun testing out AI products that democratize luxury via curation, styling services, and concierge services. Many of these products are still in the early days - in beta or pre-launch. Several of these tools include AI in a fairly limited capacity (ex - Whering). Some of the tools don’t yet have a clear path to commercialization on a large scale with brands or retailers (ex - Doji). As the challenges faced by luxury ecommerce illustrate, there is a big opportunity to improve the shopping experience for expensive, “fun” products. I would like to revisit this category in a year, because I expect we will see a massively different picture of the landscape - and I’m curious to see how these trends evolve.

Thanks for reading, and stay tuned for the final installment of Value x Vibes next week - Fun/Cheap products!

Stay Curious,

Melina

This familiarity breeds confidence in what was, only a few years ago, a new way of shopping. Ecommerce penetration of retail categories ranges from 12.4% for groceries to 41.6% for categories like apparel and beauty – most purchases are still made in stores. It’s comforting for many to see a consistent, trustworthy flow online from product details through purchase.

The uniformity of ecommerce can make online shopping feel tedious. Giants like Amazon and Walmart offer shortcuts such as Subscribe & Save—which has become increasingly popular—to reduce clicks from discovery to purchase. Yet, when browsing for inspiration, many consumers find these streamlined sites too limiting. They often turn to user-generated content and third-party channels for product research, rather than visiting a brand’s website directly. Just as we work to make our sites more efficient, we also seek to inspire discovery and embrace authentic storytelling.

For years, brands have been exploring new technologies to make our online experiences more fun and engaging on one hand, and faster and easier on the other. Everything from chatbots to express payments has become table stakes. AI tools represent the next wave of opportunities that promise a new level of personalization. In this piece I share a framework for how I believe brands will embrace AI tools, based on how consumers want to shop for different types of products.

Value x Vibes: Making online shopping fun again

I predict that AI will be used differently for different categories of products, based on what we’re looking for when we shop for these products.

I call this the Value-Vibes framework for AI Adoption in ecommerce.

Today, I’ll discuss the left half of this matrix – what I’m deeming boring products – and how we’ll be using AI to buy these.

The Value-Vibes matrix groups products by how expensive they are and how fun they are to shop for.

More expensive products typically require more upfront research, consideration, and comparison and have a longer cycle from initial consideration to purchase. Less expensive products can be a quick trigger – you see something affordable that you like or need, and you may hit “buy” quickly. For a brand, it’s about being in the right place at the right time.

All of us may have different preferences about which products fall where in this matrix. But all of us have categories for which shopping feels like a chore, and others that excite us and inspire us. We will use different tools and engage in different behaviors for different kinds of purchases – and brands should understand how we want to engage when considering how to invest in AI.

Boring/Cheap: Automating every day purchases

In the Boring-Cheap space, consumers are buying everyday purchases. They are seeking convenience and value. Many of these purchases are repeated on a weekly/monthly cadence – think groceries, toiletries, and home products. You may already have subscribe and save set up for some of these products. In this space, AI will primarily be used by customers to automate purchasing - offering convenience, time savings, and price transparency.

AI Shopping Assistants will start to replace subscribe and save and existing grocery delivery services: Boring/Cheap products will be the first purchases that shift to AI Shopping Assistants. As I wrote about in February, there are already AI Operator tools available that can purchase products on our behalf. As these tools become more advanced, we will “outsource” these low-stakes purchases first, gaining confidence in using these tools and building an understanding of the value they deliver.

ChatGPT’s Operator announced that two of their first partnerships are with Instacart and Uber – two platforms that traffic in Boring/Cheap items like groceries and everyday essentials. This is early evidence that platforms see capturing these low-stakes use cases first as critical to changing consumer behavior.

AI Shopping Assistants could improve on the experience of subscribe and save in a few ways. For this example, let’s say that today I have Mrs. Meyer’s all-purpose cleaner delivered monthly via Amazon. In the future, an AI Shopping Assistant may take this flow over for me and could:

• Identify better priced options for the same or similar product and swap them out (with customer approval): It might find a deal on Mrs. Meyer’s on Walmart and move my monthly purchase there.

• As the product landscape changes, identify products that better fit my needs and suggest them: The AI Assistant would know that I value non-toxic cleaning products. Perhaps if Target releases a well-reviewed private label option that is priced better, it would recommend I move my subscription to this product

• Proactively communicate about upcoming orders seamlessly: I envision a more user-friendly dashboard vs. what Amazon currently has. While Amazon has improved this functionality, an AI Assistant could make adjusting and adding to our ongoing subscriptions easier (e.g. making any changes on my behalf based on conversational cues/prompts).

• Adjust orders: This includes seamless substitutions when items are out of stock and considering factors like seasonality in subscriptions – “you may use less allergy meds in the winter – do you want to restart that subscription that lapsed now that spring is approaching?”

• Reschedule deliveries proactively based on our schedule: “I see you’re away from home these days—should I move your delivery to X date, when you’re back?”

• Suggest subscription options based on things we repurchase frequently, highlighting possible cost savings: “you’ve ordered dog food 5x this year. Let me set up a recurring subscription with Chewy and it will save you $100 per year”

What does this mean for brands?

It is already hard for brands to differentiate in the Boring/Cheap space, as many of these products are utility goods. Customers are largely getting these goods from a big marketplace site, like Amazon, Walmart, or Costco vs. transacting on a Brand.com, minimizing the opportunity for unique brand differentiation.

Subscribe & save is a boon for brands in this space, because it allows them to lock in a long-term revenue stream from consumers. In this example, the AI Shopping Assistant challenges that dynamic by introducing flexibility around brand and retailer, and handling the shopping for us. In many cases, consumers would not even see a brand listing – on either Brand.com or retailer site – prior to purchasing.

Brands would need to reach consumers elsewhere, such that I would ask my AI Assistant for Mrs. Meyer’s specifically (a monumental marketing task for small, independent brands). They would also need to optimize their product feeds and content such that they show up in AI Shopping Assistant queries.

What does this mean for marketplaces?

Amazon’s Rufus is a major contender to offer “smart” subscribe & save and take over this role, but isn’t that advanced yet. If ChatGPT, Perplexity, or others continue to advance their shopping capabilities, they could steal some share and become the go-to hub for day-to-day automated tasks. This would be a very powerful position, especially if coupled with non-shopping related tasks, like automating booking travel, appointments, etc. You can see via Operator’s interface that they are heading this way. Personally, I would test out automating some of my “boring” purchases outside of Amazon if I could get value like that described above.

What cheap/boring purchases do you want to automate? Have you experimented with AI yet to do this?

Leave a comment

Boring/Expensive: Targeted product research and comparison

While AI can simplify routine buys, its impact on expensive, considered purchases is equally transformative.

There’s a common trope that couples get into fights at IKEA. Why? Because it’s annoying to make big, complicated purchases – where you need to account for the dimensions of your new apartment, your design style, technical jargon, and the measurements of all your other furniture – in a big, crowded place. There is a lot of opportunity for technology to make this easier.

For Boring/Expensive products, we are looking to compare and understand value and tradeoffs. This couch is more expensive, but it’s real leather. Could I resell it when I move? I like this coffee table more, but the shipping is 8 weeks. This TV is bigger, but it’s also thicker and might not look good in my space. What do these vacuum specs even mean? Consumers turn to affiliate sites like Wirecutter to understand the “best” product in each category. With AI, the results will be tailored to highlight the best “for me.”

In this category, product education and comparison are critical, and AI tools are well-positioned to scrape accurate information from brand and retailer sites and summarize it in an intuitive way that can drive to purchase. Taking this to the next level, the tools should align the product specs with your needs and desires. Thus, research is the critical differentiator that customers will use AI for in this category.

What this means for marketplaces and retailers

Trust is paramount in this category. I believe that customers will buy a car with one click in the future. But for them to take this leap, they will need to fully trust the information that AI surfaces – and will need assurances that if something is inaccurate or unclear, they have recourse.

As I wrote about last week, I purchased some shampoo with Perplexity’s shopping tool, which advertised free delivery. When my item was fulfilled, it turned out I was in fact charged for shipping. Small missteps like this can erode trust in AI tools and must be mitigated before these tools can be used for complex purchases like furniture, large appliances, and even cars.

Marketplaces that are perceived to be recommending products in a way that isn’t transparent will receive backlash. I’m very curious to see how paid placements will evolve in the world AI shopping assistants, as they are currently treated as objective arbiters.

• Google remains a huge player, with Google Shopping already offering fairly mature research and comparison capabilities that make transacting across retailers sites easy.

• Large retailer sites for expensive, technical products, like Best Buy, are surely racing to implement AI tools that enable smart comparison shopping to capture share on their site vs. competitors. A comparison product grid on PDP or PLP doesn’t cut it. They want you to start your search on BestBuy and trust their tool for the best product research restricted to the items on their site only.

• As in the Boring/Cheap case, Amazon and AI discovery engines will be working to make their offerings as useful as possible, highlighting products across brands. Perplexity’s partnership with Shopify highlights how it wants to tout independent brands.

• Wirecutter and other affiliates are likely to lose out as AI tools evolve in this space. Individual influencers and tastemakers may also be squeezed.

• There is also an opportunity for niche players to emerge in this space. Just as the Sears catalog emerged as a trusted place to buy home products in the past, new AI-enabled online catalogs could emerge that offer seamless product comparison, how-to videos, repair support, resale support, and more.

Next week, I’ll focus on the Fun categories – Cheap and Expensive and how AI is disrupting how we shop for these categories.

ICYMI

My last post was about getting AI to do my shopping for me. Check it out!

What I’m Reading & Enjoying This Week

’s piece on Sol de Janeiro

’s breakdown of how running brands are growing and differentiating

- one of my fave writers on here - shared another banger on what affiliate publishers can learn from Reddit. I trust Reddit for so much information, from parenting to where to go out to eat in NYC. This is such a fascinating topic, especially as Google has prioritized forum results on the SERP.

has a great newsletter on AI and everything we need to know about it. I’m also enjoying her podcast as a quick briefing on the AI-related headlines.

Stay Curious,

Melina

What is an AI agent?

You know how in sci-fi movies, there’s always a computer who gains sentience and starts acting on its own? That’s basically agentic AI.

Many of us have interacted with primarily non-agentic AI tools, for use cases such as:

• Asking ChatGPT questions, essentially treating it as a more sophisticated search engine

• Interacting with AI-enabled assistants like Siri or Alexa

• Observing the output of AI-enabled experiences, like product recommendations online, targeted advertising, or tailored recommendations on Netflix or Spotify - a lot of advanced personalization that we enjoy as consumers is powered by AI

• Office tools like Copilot summarize meetings and help draft emails

All these use cases are static AI - they produce a predefined range of outputs every time an input is provided, and do not adapt, make decisions, or take actions. They make recommendations, and the models powering the experiences improve over time. But they can’t take your credit card and go on a shopping spree at Bloomingdale’s. At least not yet (cue ominous music).

AI agents, on the other hand, have the ability to act on our behalf to fulfill tasks. For this week, I am focusing on the emerging capability of AI agents to perform tasks, like purchasing products, on our behalf. These tools are being called “AI Shopping Assistants.” But like a lot of technology applications, a lot of the money to be made is behind-the-scenes, in unsexy parts of the business like supply chain and inventory optimization. There are also exciting developments in creative production (developing and testing assets) and customer service. Those use cases and more will be the subject of future newsletters - at this rate, the AI series is shaping up to be at least 20 parts. (Here’s Part I and Part II if you’d like to catch up).

Let’s dive in!

Can an AI agent buy my shampoo?

Because buying things online and having them delivered to my door in 2-3 days is still too taxing, I was very excited to test out the shopping capabilities of agentic AI. In the (very near) future, this capability could enable consumers to automate tasks like booking travel and making restaurant reservations. It could allow us to outsource weekly grocery shopping and buying wedding gifts from a friends’ registry entirely to an AI assistant.

As of today, Perplexity Pro is one of the only tools available to offer a functional AI Shopping Assistant who can complete transactions. I had to try it.

First, I asked Perplexity for help buying shampoo. I specified I wanted OGX. It surfaced a few options from CVS, Ulta, and other retailers. The “Buy with Pro” button triggers the agent-enabled purchase flow.

Then it prompts me for some details to complete the transaction.

I was surprised this step wasn’t more seamless. I had just provided my credit card information to sign up for the free trial of Perplexity Pro. So far, it would have been faster to buy this on Amazon, but I’m trusting vision!

There is also a section within cart to specify size and color for SKUs that have multiple color/size combinations. This hints at some limitations with Perplexity’s functionality.

As Maxwell Zeff at TechCrunch points out when he tested this in December, it appears that Perplexity is scraping the web for available product details, vs. listing items for sale themselves in their own marketplace. In other words, I just gave $10 to Perplexity, and Perplexity is going to turn around and buy the shampoo from CVS with that $10. This raises a lot of challenges, for instance - if CVS happened to turn on a promotion between when I paid Perplexity and when they executed the transaction, would I get a refund? Conversely, how much above what I paid is Perplexity authorized to spend if the price changes in the other direction?

Ultimately, I received an email from CVS confirming my purchase 4 hours after I completed the transaction on Perplexity. I immediately noticed an issue - I was charged a $5 shipping fee by CVS that I was not made aware of when I bought via Perplexity. In fact, both screenshots above highlight “Free Delivery” as a benefit of purchasing via CVS. This essentially increased my purchase total by 50% - from $10 to $15 - and shipping cost is a key criteria in purchasing a low-cost item like shampoo.

Reflections from my first AI purchase

Perplexity’s agentic shopping capabilities are in the early stages, and I really respect the team’s hustle to bring this to market and be a first mover, even though it’s pretty bare bones. Perplexity is clearly making shopping a differentiator, rolling out features over the holiday geared to recommend and purchase gifts.

Right now, Perplexity Pro acts as a bridge between shoppers and retailers rather than operating its own storefront. That means it’s scraping product data and executing transactions on my behalf—but the process isn’t always seamless.

OpenAI Operator: Ordering a new toy for my son

ChatGPT introduced AI Operator, their agentic assistant, 3 weeks ago.

I will not attempt a detailed technical explanation, but it’s interesting to note that Operator is essentially trained to act like a human and click through to complete actions on a website like we might do, vs. using API integrations. This enables more use cases to ramp up faster, as Operator can essentially do anything a human can do online, if it has access to the right context.

I had to test it.

Under shopping, there are prompts to help get started. I want to buy my baby a new toy.

I used this prompt: I have a baby who is 8 months old. I'm looking for a developmentally-appropriate toy that is educational, costing between $10-15 and available with free shipping. Can you share some options? I'd also like you to buy it for me.

In our Chat, Operator opened up their own browser and started searching for toys…crazyyyy.

It selected a toy that fit my criteria and opened a window in Target.com for me to buy it. Operator didn’t actually buy the toy for me—it surfaced the best option and then walked me through the purchase flow. Unlike Perplexity, it doesn’t complete transactions autonomously (yet).

What agentic AI might mean for the future of shopping

It took Google 4 years after launching search to launch Google Shopping (in 2002), which then ramped up fairly slowly to the several billion monthly searches it commands today. Meanwhile, these AI companies have waited mere months to move into shopping. I think it’s likely that by this summer, many of us will be using AI shopping assistants to complete tasks.

Practical examples of how this could look in our day-to-day life

• You ask your preferred AI Shopping Assistant to reorder all of your consumables, like pet food and toiletries using either voice or typed commands. Because you are sharing your order history with the AI, it knows your preferences and can reorder the same items at the best price.

• Your AI assistant has access to your Google calendar and knows you have a wedding coming up in 6 weeks. It prompts you to order a wedding gift and asks if you want to order something from the wedding registry link provided via email. You can share a desired price range and mail a wedding gift without ever opening the registry yourself.

• You notice your socks are worn out so you ask your AI to order another pack similar to the last ones you ordered a few months ago. Because you’re sharing order history, it knows your preferences and can surface the relevant item and some better-priced, similar alternatives.

• You rented an Airbnb with friends for the weekend. In the group chat, you send some links for recipes you want to cook together. You send these links to your AI assistant, who assembles an Instacart order with all the ingredients you need and schedules delivery for a few hours after your check-in.

Who will own the consumer-AI assistant relationship?

If I am interfacing with an AI shopping assistant for many shopping tasks, I won’t be visiting Amazon, Google, or brand sites as much as before. I’m sure this scares legacy retailers and marketplaces, who want to maintain their position in the value chain.

This is prompting a scramble to develop agentic AI shopping assistants.

An AI shopping assistant from OpenAI, Perplexity, or another AI-first company

These companies are at the cutting edge of building AI capabilities and thus are well-positioned to build consumer-facing products like AI shopping assistants. OpenAI Operator and Perplexity’s recent launches highlight their advances in the space.

I believe a challenge will be how best to add commerce to the equation. As I observed when testing Perplexity’s shopping capabilities, products aren’t yet listed for sale on a true marketplace. With Operator, the AI is searching external sites for me, but I am still owning the transaction (though that may change).

These platforms also lack some of the context on who I am as a customer - what are my interests, order preferences, etc.? Theoretically, I could grant them access to my purchase history in some way - bank statements? emails? - so they could best tailor recommendations to me. Like any personal assistant, there will be an onboarding period.

A Google AI shopping assistant

Google is well-positioned here because so much of my day-to-day life is managed in Google tools - Gmail, Calendar, and Search. Google has a rich portrait of my purchase history (via order emails), my schedule, and interests. There is a lot of context for an AI shopping assistant to tap into to make recommendations for products or to anticipate needs. Being truly useful is critical for these applications to go from “fun to test” to truly sticky behaviors that change the way we shop long-term. I wrote recently about Google Shopping’s AI focused makeover. This is setting the stage for agentic AI shopping assistant capabilities.

An Amazon AI shopping assistant

Amazon is investing heavily in AI across their business. In the shopping assistant space, they have rolled out Rufus, a chatbot that helps with shopping queries inside Amazon’s app. It’s not an autonomous shopping assistant (yet), but it could evolve into one.

Amazon has a data advantage. I buy the majority of my everyday purchases - from baby supplies to kitchen stuff - on Amazon. They own my order history data and know a lot about my preferences. If I were Amazon, I wouldn’t want to share that data for free with an unaffiliated AI agent who might use it to recommend products from other retailers - in fact, Amazon was one of the earliest companies to block OpenAI from crawling their site. Any outside AI shopping assistant may be hampered by a lack of access to a lot of my purchase history.

I think Amazon will build their own assistant to work within Amazon - for instance, allowing me to buy groceries via Whole Foods and schedule delivery using conversational text and voice prompts. I’m sure Walmart will work to offer a competing tool. It’s an oft-quoted stat in ecommerce that 50% of product searches start on Amazon. Using an Amazon AI assistant ensures those dollars stay on Amazon.

Will brands push back on AI agents “shopping” on their sites?

When an AI agent completes a transaction on a customer’s behalf, brands lose the ability to “sell” in the traditional sense, meaning they can’t tempt an AI to grow their basket in the same way they can prompt a human to buy a pair of shoes to match the jacket in their cart. These bots could represent low-value, unprofitable customers to some brands.

Some retailers already restrict AI-powered bots from scraping their sites (Amazon being a notable example). If AI shopping assistants gain traction, we could see more brands push back or require stricter API-based integrations.

Putting on my brand hat, I can imagine a nightmare scenario where AI assistants place lots of orders on my site, which affects my inventory levels and operations, only to cancel or modify them when they don’t meet the consumer’s requirements. Especially in the early days, this could create a lot of noise and turn me off to hosting this type of traffic.

On the other hand, restricting AI agent traffic entirely can be shortsighted, since Brands want to drive sales, and these agents may make up an increasing share of traffic in the coming years. Perhaps AI agent traffic becomes like any other affiliate channel - taking a small slice of revenue for recommending your brand in return for the sale.

The death (re-birth?) of SEO

The age of keyword stuffing and SEO hacking is over. As more people start their searches outside of traditional search engines, organic traffic won’t just be determined by rank on Google. It will also be determined by how relevant AI shopping assistants find your products and content.

Just like how brands need to maintain product feed data to surface on Google Shopping and other platforms, they will need to format and maintain product data in a way that AI shopping assistants can parse it. This can drive advantages in discoverability by these tools, increasingly inclusion in AI assistant recommendations, and hence, conversions.

It’s still unknown exactly how brands should “optimize” their sites and content for these tools, as they are so nascent. But brands that move early to make themselves discoverable on these platforms will be big winners.

How do you sell to an AI? With another AI, of course!

So consumers will all have AI shopping assistants. What about brands? Brands will have AI ambassadors.

For example, on a fashion site, the brands’s AI might introduce itself in machine terms: “Here are the latest arrivals and current promotions,” and your AI might respond, “My user is looking for a size M black dress, ideally A-line, under $100, and prefers cotton. What do you have?” The two AIs can rapidly exchange info – availability, prices, detailed specs – far faster than you could clicking filters.

This sounds very futuristic, but you can imagine the implications for ecommerce if a large share of traffic starts coming from AI assistants. The impetus to lay out websites in the traditional sense - to be easy for humans to understand - would become less important vs. other objectives.

There’s so much more to come here, but for now I’ll work on using Operator and Perplexity to automate my life so I have more time to play around with AI tools and write this newsletter. Stay curious! — Melina

ICYMI

The most recent installment in this series was all about visual search. Check it out!

Useful references

Techcrunch December 2024 - The race is on to make AI agents do your online shopping for you

Crunchbase February 2025 - OpenAI’s Operator Signals The Agentic Era Of Commerce Is Here

New York Times February 2025 - How Helpful Is Operator, OpenAI’s New A.I. Agent?

Podcast - Sept. 2024 - Bret Taylor, Invest like the best

Podcast - Dec. 2024 - The Race for AI—Search, National Infrastructure, & On-Device AI, a16z Podcast

To that end, I’d love your help deciding what to publish next week:

Often Overlooked: Strong Post Purchase Communications

For Christmas I received a Dyson Airwrap. I am loving it! But the learning curve is pretty steep. The Airwrap comes with several attachments that all have a specific use case - curling, taming flyaways, straightening, etc. Upon opening the package, I had no idea how to use this thing. Luckily, Dyson stepped in with some very helpful resources.

Registration and video resources

In the box, there was a QR code to scan to register the product. While I did have to download an app (a personal pet peeve), this flow was very simple and took no more than 2 minutes. Once in the app, I had access to video tutorials and resources.

Email flow to reinforce and remind

A week after I registered my Airwrap, I received a series of emails introducing the tool and re-surfacing some of the tutorials and resources. Having this content pulsed across multiple channels (app, email, and website) is a good strategy because customers are busy - this makes it more likely that we will find and use the resources.

The timing is also well thought-out. A week from registering the device gives the customer time to experiment and try things out. Perhaps at this stage, they are struggling a bit and tempted to give up on the product. Pulsing email resources now makes sense.

Opportunity for Dyson: Organic social integration with post purchase communications

I also watched a lot of tutorial videos on TikTok when researching this product and figuring out how to use it. Tying TikTok and user-generated content into the post-purchase flow (e.g., featuring relevant trending videos in post-purchase tutorial emails) is another way to drive engagement. The tutorials provided by Dyson, while useful to understand product functionality, are a bit dry. I found it more engaging to hear real customers use the product and share “hacks” from their experience. Dyson’s US TikTok channel has a relatively low follower count, while creator videos reviewing the tool have 500k+ views.

Post-purchase insights for Brands

For many brands, post-purchase communication looks something like this:

• Order confirmation

• Shipping updates (often from the carrier)

• Delivery notification (often from the carrier)

• Leave us a review (often timed too soon after the product arrives - this is another personal pet peeve)

• Silence

This leaves a lot of opportunity to build a solid customer relationship and drive loyalty and repeat purchases.

Of course a brand like Dyson, which makes complex, expensive, and technical products will send communications explaining how to use the product. But I believe that many consumer brands would benefit.

• In the beauty space, brands should be sending educational content on how a new serum fits within an existing skincare routine. One problem I’ve encountered when I buy a new skincare product is that I’m unsure how to order the product application with other things I already own. As a result, I may not use it as much. The Brand wants me to use this new product regularly so I see results and want to repurchase. Post-purchase communications like this can build trust and authority, encourage utilization of the product (and hence repurchase), and speak to upsell/cross-sell (did you know the new serum you just bought pairs well with this moisturizer).

• In the apparel space, brands should take a similar approach, focused on how to style new items the customer has just purchased. For Christmas, I bought my husband some wool pants from J.Crew, which he really likes. But he hasn’t worn them very much because he’s not sure how to style them with other things in his wardrobe. J.Crew should be sending editorial content on how to style those pants. This again presents upsell/cross-sell opportunities to build out the full outfit. It is also an exercise in brand-building to share UGC or other content showcasing how the products are worn by real people.

• Tech is a key enabler here. Taking this strategy to the next level would involve personalizing the communications, e.g. in the post-purchase summary email, allow customers to request styling guidance on specific items in their order. With adoption of AI tools to automate creation of email marketing assets, creating “how-to styling resources” for specific SKUs could be more scalable across a wide product catalog.

Implementing effective post-purchase communications has been shown to drive customer satisfaction and loyalty. Post-purchase emails see a 217% higher open rate than traditional emails, and have a 6.4% click-through rate (source). Also, lifecycle communications, including post-purchase email campaigns, generate 90% higher revenue per recipient compared to pre-purchase emails (source). Brands should be investing more here!

My AI avatar is cooler than me

I was invited to test out Doji’s private beta. Doji enables users to create an AI likeness and use it to try on real products and shop their favorite looks.

This was so fun!

Doji seems to be targeting us - individual fashion lovers - as consumers, to encourage creativity and discovery. I had tons of fun with the experience and I could see it as an effective discovery mechanism for new brands or styles. I could also see this tool being a powerful add-on for brand and retailer sites - for example, enabling an experience on a Nordstrom jeans PDP where you can see your avatar in a bunch of other shoppable looks - essentially replacing the “Shop the Look” module with something more personalized.

While not the stated purpose of Doji, I see this tech as an interesting evolution of the “smart fit” solutions that have been bouncing around the retail tech world for many years. Some of these use your preferred sizing at other brands to estimate your size at a new brand. Others use your self-reported measurements or body shape. Others use AR interfaces. Personally, I haven’t seen anything revolutionary from these tools as a shopper. I still often order multiple sizes of an item if I’m not sure about the size. Perhaps leveraging AI could be a path to encourage adoption and trust in sizing recommendations.

Big purchases on the big screen: Is there truth behind the meme?

Gen Z has found another way to make fun of us millennials. They have pointed out (accurately imo) that we tend to browse on mobile, but make our “big purchases” on the desktop screen. I admit I fall into this stereotype - I like having multiple tabs to compare before I buy! So I was curious to see what the data has to say.

What does the data say?

Yes, Gen Z shoppers are more likely to be predominantly mobile purchasers vs. millennials. But the gap isn’t as big as what the memes might suggest. A 2019 survey revealed that 55% of Gen Z consumers in the United States made most of their online purchases via smartphone, compared to 42% of Millennials (source). These numbers are a bit dated, but the generational gap is not that large.

In the industry as a whole, mobile commerce is growing at a ~2x rate vs. desktop. This holiday season saw nearly 55% of total sales on mobile across all ages - highlighting how fast things are shifting since the 2019 survey was conducted (source). This is likely due to better optimization of mobile sites for commerce, which drives comfort and trust with making increasingly larger purchases via mobile.

I look forward to seeing the memes in 20 years when the brain-computer interface has rendered mobile completely obsolete.

What I read and loved this week

published an awesome deep (haha) dive on DeepSeek, the Chinese AI firm quickly and cheaply spinning up models competitive with OpenAI and others. Read this if you are enjoying my AI series and want more details on the technical landscape behind the user-facing tools.

SatPost by Trung Phan

DeepSeek: Links and Memes (So Many Memes)

Thanks for subscribing to SatPost…

Read more

2 years ago · 5 likes · Trungphan2

wrote about Flip, a video shopping app that has content similar to TikTok Shop. It’s interesting to see how startups are bringing live shopping and social commerce, already huge in other parts of the world, to the US, and capitalizing on TikTok’s success. Personally, I go to TikTok for the organic content and for creators I enjoy, not for shopping. If I am inspired by something I see and want to buy it, I’m happy that Tiktok makes it easy. I am dubious that people will go to an app just to shop, unless there is a key draw, like early access to products, special discounts, etc. Really enjoyed her analysis!

Delivered

The app that pays you to watch shopping videos

If you’re new here, welcome to Delivered, a newsletter covering the intersection of content & commerce. If you haven’t already, subscribe to get all posts straight to your inbox…

Read more

2 years ago · 4 likes · Elaheh Nozari

wrote back in October about Merit’s influencer strategy for their recent perfume launch, which included a Substack strategy. This just crossed my feed, and it’s a great evergreen breakdown of influencer marketing and how a brand can use it effectively.

Following Up by Clare Moore

Retrospect: Merit's Influencer Strategy from Tease to Launch

Merit is one of my favorite brands, and when they teased their new launch earlier this month, I would have not guessed fragrance. On October 6th, Merit cleared their Instagram feed, teasing a new launch. Two days later, they revealed their new fragrance, Retrospect, with a launch date of October 22nd. Over the next two weeks, Merit slowly released more …

Read more

2 years ago · 52 likes · 4 comments · Clare Moore

is a great publication on consumer research and marketing. This is a useful primer on some updated rules of the game.

Pulse Check

10 Research Rules Every Brand Should Follow in 2025

Consumer behavior is rapidly shifting, so here’s my modern, practical guide for brands and researchers conducting effective and culturally nuanced market research this year. Inspired by The New Rules of Media, I’ve compiled my thoughts, with hopes that following these steps will ensure more people will reach for their wallets to support your brand and t…

Read more

2 years ago · 18 likes · Kate Whalen

ICYMI

I wrote about visual search!

My once-peaceful thrift trips are no more. My local shops are packed with people, phones in hand, searching for the best resale opportunities to upload to Depop, eBay and other platforms. On the flip side (pun intended), thrift and consignment stores are using visual search to better price their wares, so unicorn finds like my designer coat are becoming more rare. While thrifters have always researched the provenance of their finds, visual search is a one-click, dummy-proof way to understand the market value of pretty much any consumable item. You don’t need to know about fabric composition. You don’t need to know about designer labels. You just need a phone with a camera.

Beyond the thrift store, this technology has implications for how we shop.

Visual search is getting better

I’ll leave the detailed explanation to the experts, but here’s a quick summary of how visual search works:

AI visual search technology relies on advanced image recognition and machine learning techniques to understand the content of the images and match them with relevant items, products, or information from a database.

Machine learning ensures visual search becomes faster, more accurate, and increasingly personalized over time, aligning better with user needs and providing more intuitive experiences. Without getting too deep into the technical details, a few factors drive the continuous improvement of these tools:

• More data - many of these tools (like Google and Amazon) are getting billions of monthly searches. This expands their dataset and improves accuracy.

• Signals from users - The quality of results is measured in part via user feedback - when you engage with the search results, e.g. if you end of purchasing something, you’re signaling that those results were relevant. The algorithm can then adapt based on that learning.

• Technical advances in the underlying models powering these tools

• Advances in hardware and processing ability

Taken all together, consumers have observed a very fast improvement in the power and accuracy of visual search tools in the past few years.

Visual search usage: Limited but growing fast

Unsurprisingly, with improvements in visual search capability comes increased adoption. While overall adoption is still relatively low, it’s growing fast. Like many new technologies, adoption is stronger among younger users. A few key stats from eMarketer:

• While only 10% of US adults regularly use visual search, 42% are at least somewhat interested in using the tool, as of August 2024.

• Younger consumers are most likely to use visual search—22% of consumers ages 16 to 34 have seen or bought an item via visual search compared with 17% of 35- to 54-year-olds and 5% of consumers ages 55 and older.

• Visual searches worldwide increased 70% YoY, per Amazon.

• Google gets 20 billion Lens searches a month, 4 billion of which are related to shopping. This compares to ~1.5 billion monthly searches that start on Google Shopping.

Pretty impressive!

More players are releasing visual search products

In the summer and fall of 2024, I saw a lot of announcements about new visual search products hitting the market. The retailer in me immediately thought that platforms and brands wanted to get these out ahead of holiday - when many consumers are breaking from their typical consumption patterns to search for gifts in new ways. These products span the markets of tech, marketing, and ecommerce. They range from Apple’s offering, which will embed visual search technology in the native phone camera, to startups like Phia, which leverage visual search for the specific use case of shopping and price comparison.

Because I’m focused on ecommerce, I’ll take that lens when considering the impacts of visual search. But this technology goes far beyond shopping.

Consumer-facing visual search: a market map

I labeled this chart “consumer-facing” because there is a whole other world of tools that businesses use that leverage visual search technologies. For example, resale platforms use visual search to authenticate goods. There are also a lot of emerging players building visual search tools that plug into brand websites. But that will need to be the subject of a future post!

• All the major big tech players are experimenting with visual search - no surprise

• In retail, only big players have developed this feature so far. This makes sense, because searching via image is most useful if the retailer/marketplace has a vast/frequently changing catalog. It must be costly to license/build the core technology to enable high-quality visual search. So many retailers are probably taking a “wait and see” approach.

• I’m particularly excited about visual search’s role in resale. The transformational nature of the tech for this industry in particular is why we’re already seeing several players investing in this.

• I’m also so excited to follow emerging players, like Phia, and Perplexity’s shopping tool.

Visual Search: Changing how we shop

I initially wrote back in October about how visual search could impact brands. I’m fleshing out those ideas more now.

Reshaping product search - the move away from filters

Ecommerce traditionally has had a very linear search architecture. For example:

Women’s > Dresses > Mini > Black

It’s easy to see the limitations of this approach. Perhaps there’s a perfect navy blue mini dress that I would love, but I’ll never see it with this filter structure. That gap represents money left on the table for brands. Selecting from a long list of filters is also annoying as a shopper, especially on mobile, where the UX is often subpar. Often, shoppers opt for the “endless scroll”, browsing all dresses, to be sure they don’t miss a good option. While brands can design their navigations and collection pages to prioritize best sellers, the level of personalization is pretty limited. Shopping online can feel like a chore.

I discussed in Part I how Google’s AI search tools are enabling a shift to more conversational search - for example, you might type “I want a dark colored dress for a semi-formal event” and see curated options. Visual search is another tool for brands to strengthen their customer funnel, making it easier for customers to find relevant things and keeping them engaged on-site.

Reduced pricing power for sellers

While in-season items often have an MSRP that is consistent across retailers, any time an item is marked down at one retailer, the others are at a major disadvantage unless they match. Consumers were always savvy about comparison shopping across major retailers (especially with increased popularity of Google Shopping), but now they can look across eBay and other resale sites too. This will further erode the position of department stores and advantage any digitally-savvy player who can effectively monitor their online listings.

An opportunity for smaller brands

Many marketing technology trends aggregate gains to the biggest companies that can spend the most money on paid formats and run strategies at scale. But I believe visual search is positioned to advantage smaller, independent brands through enhancing discoverability.

• More brands in consideration set: If I see that an unfamiliar brand makes something nearly identical to what I’m searching for, I’m likely open to try it - especially if there are reviews and other forms of social proof that I can reference. Before Google Lens, I might check around on my favorite brand sites to see what was out there, but the consideration set would be more limited.

• Reduced brand loyalty: Driven by the above factors, consumers’ loyalty to a given brand is challenged by increased visibility of relevant options.

More robust resale/used good market

I think a lot about the resale market because the majority of my closet is secondhand and vintage. I believe that resale, already a vastly growing area of ecommerce, will be a beneficiary of this technology.

• Increased share of wallet from resale/used items: With visual search, it’s easier to find an item without knowing the brand and style specifications. This will bring more traffic to resellers. It will also force them to make their pricing across platforms more competitive.

• Imperative for brick-and-mortar thrift and consignment to move online: Visual search will unlock more traffic to secondhand listings, which creates an opportunity for brick-and-mortar resale businesses to move into ecommerce, or at least list items online to drive foot traffic to their physical stores.

• Increased interest and reduced barriers to selling used items: With an increased ability to see the value of your items on the resale market, there will be an increased interest in selling. For the past few years, growth of resale has been constrained on the supply side - it’s annoying to create listings as a single seller and manage them through to fulfillment, all for a relatively modest payday. New tools are simplifying listings across multiple platforms, and these tools themselves leverage AI and image recognition to reduce the manual data input required to list items.

I am gobsmacked that The Real Real hasn’t yet released any visual search features. This is a huge miss when compared to eBay’s and Poshmark’s visual search offerings. I’m a big TRR fan and I don’t want to see them miss the boat on this!

Personalized marketing opportunities

Visual searches contain a lot more context than a search term. In the past, I may have searched for “blue Prada tote bag.” Now, I can search this image:

This image contains a lot of valuable context - it shows that I like the style blogger Blair Eadie. It indicates that I might like 70s style and bold prints. It gives clues as to other brands and styles I might gravitate to. Capturing these insights could allow tech companies and brands to better tailor recommendations to me beyond a single query.

The future is here - Multi-Modal search

A Google Lens update from October 2024 sheds light on how we’ll all be shopping in a few months:

Lens search is now multimodal, a hot word in AI these days, which means people can now search with a combination of video, images, and voice inputs. Instead of pointing their smartphone camera at an object, tapping the focus point on the screen, and waiting for the Lens app to drum up results, users can point the lens and use voice commands at the same time, for example, “What kind of clouds are those?” or “What brand of sneakers are those and where can I buy them?”

Of course, I had to try it.

Using the same image above, I prompted Lens to focus on the purse and added the text “this purse but green.” The search did turn up the exact item in green. Unfortunately, it was sold out, so I will not be blowing $3k on a cute bag today.

*Side note - to improve this product, Lens could prioritize retailers where the item isn’t sold out.

Conclusions

In Part I of this series, I wasn’t 100% sold on the utility of AI Overviews, at least for commerce. As a shopper, I’m sold on visual search - it’s very intuitive and returns relevant results a lot of the time. Multi-modal search opens up the potential for even more relevant results. I’m also excited about how this technology may increase discoverability for independent brands and secondhand products and sellers. Will paid placements in these visual search tools start watering down the experience? More than likely, yes. But for now, I’m a fan.

Sources I found useful while researching for this piece

1. ’s “For Dummies” explainer on LLMs (thank you for the super clear explanation!)

2. Leeway Hertz primer on visual search

3. Retail Brew: Google Lens gets a precise tweak before the holidays to push shopping - October 2024

4. TechCrunch: AI-powered visual search comes to the iPhone - September 2024

5. AdWeek: Poshmark Takes on Google and TikTok With Gen AI Visual Search Update - October 2024

6. Wired: Google’s Visual Search can now answer even more complex questions - October 2024

7. ThredUp Resale Report 2024

ICYMI

Thanks for reading, and stay curious!

Melina

Leave a comment