Are you celebrating Halloween tomorrow?

I am happy it’s on a Saturday because it’s made this year way more exciting than previous years when I was worried about work and getting the kids to school the next day. It’s also the last time Halloween will be on Saturday until 2037! So you gotta live it up! 🍬 🍫 🎃 🧛‍♀️

Anywho.

I do have a treat for you today….

I want to share some of me and ’s latest thinking on where we think AI in GTM stands right now.

TLDR: the models are incredible. The harnesses (e.g. new AI-native tools, coding agents, MCPs/CLIs, etc) are getting better every week. But companies still haven’t properly implemented AI into their GTM.

Read on to find out why we think this is…

As AI has developed and has been adopted in GTM, certain layers have been barriers to full adoption and value. First, it was the model layer, where models were smart but hallucinated and struggled with complex GTM use cases.

Then, over time, the models kept on improving, but it was a struggle to figure out how to “harness” the intelligence. Most existing GTM tools hadn’t figured out how to add AI into their features yet, the new AI native tools were non-existent or emerging, and no one in GTM had used Claude Code. Genuinely useful MCPs/CLIs were rare, and no one was talking about context management. We had excellent AI but minimal tools to deploy it.

Then in late 2025 and into 2026, the harnesses started to become very useful. Claude Code took over, lots of new GTM tools emerged, existing tools added agentic features and MCPs/CLIs, operators started figuring out context systems, guardrails, and observability. We can now build agents that can do real work.

But, still so many companies still are stuck with a few pilots and localized AI usage. The bottleneck now is firmly in the implementation layer.

Why?

Implementation is f***ing hard! 🫢

As operators, we know implementing ANY new technology is hard. Broken data, legacy systems, tech debt, processes that have to be redesigned, and change management all get in the way.

Now try implementing AI into your GTM where you have to fundamentally change how you think about and architect important (previously human) jobs/processes on behalf of the GTM teams and implement all sorts of new technical frameworks (token/credit usage, evals, agent guardrails, etc). While all this is still evolving, literally daily.

So today, at the end of 2026, most companies are stuck in what we call the “AI operator” phase instead of truly running an “agentic GTM operating system.”

AI operators are individuals at your company running AI in a silo:

• One task at a time, triggered and guided by that person. Could be Claude Code, Grokbot, etc.

• It makes the person faster and better, but the value caps out at their individual output.

• It’s still reliant on human-driven prompting, so it’s inconsistent and hard to scale.

• Often working locally so if they go on PTO or are sick the work stops. And very few of the optimizations get fed back across the team. One salesperson is killing it while the rest are stuck in 2020.

• It’s useful, but not transformational.
  

An agentic GTM operating system is different, it’s where you’ve deployed production ready centralized agents that reliably execute core business processes and jobs to be done. The operating system has:

• Centralized, automated agents, not reliant on individual human prompting to trigger them. It reads shared and self improving context, and writes to the system of record.

• Reliable, consistent, scalable, secure, and built with guardrails.

• Able to operate within the fabric of a company’s GTM 24/7. No PTO outages.

But as I said, it can be hard to build it all out. But if you get there, that’s where the potential of AI is realized.

So the main focus for GTM teams now should be closing the gap between what models are capable of (ie all those ideas your C-Suite is hearing about) versus what they actually have deployed in their systems.

Here is an example to show the “gap” by looking at the “job to be done” of account research with an AI operator approach versus one within an agentic GTM operating system.

AI Operators Doing Account Research:
> You give your sales team access to Claude (plus connectors to SFDC, Hubspot, etc), and each rep is able to chat away to try to help them create a list of accounts to go after.
> They can use Clay’s MCP or take that list developed by Claude and build a list from another tool with all the account and contact data.
> They can then use Claude to do deep research on the accounts to then draft emails to send to the prospects.

This is all fine. It helps the seller individually. But everyone is operating in silos. There is no way to measure or manage it, one rep finds a great method and the others don’t benefit, some use it and others don’t, and it still takes a load of time.

If this is the promise of AI we should short the Anthropic stock once they IPO.

Account Research Agent (in your operating system):

> An account agent is created (you could use Clay, a tool like Rox/Actively, build custom, etc) that ingests the full context on your ICP, your key personas within your full buying group, your current customer base, call recordings, your target accounts by seller, your top signals for outbound, and key messaging by persona for each signal, your inbound intent signals, and where to read that data on contacts under accounts in your CRM.
> It is doing research on all existing accounts and contact behavior as well as net new accounts showing signals on a regular cadence.
> It is then creating net new accounts or updating existing accounts with a priority score and adding signal data to a custom object on the account in the CRM where sellers want to see the data.
> It is then doing research on the key contacts at that account across the whole buying group and adding them into CRM.
>It is then surfacing these contacts to a seller in a daily report in Slack with suggestions on email cadences to add to by signal or allowing the seller to reach out with their own personalized message with context from that signal or activity on the account.
> If any of the agent-prioritized accounts have opportunities generated, then it will log that positive result and continue to see if there are iterations or learnings it can then suggest for review or automatically make to itself.

So, how do companies close that gap?

Companies will have to allocate significant time and resources to true agentic implementation. It takes work to rebuild your GTM for agentic work.

So instead of having an internal AI mandate focused on the team using specific models or tools and working in silos, the GTM team should focus on becoming more centralized in their AI in GTM strategy and approach, which I think should be led by GTM operators at the helm since they are the most clued into technology, data, and processes already.

And this re-architecture will take time. But time that will be most valuable in the long run.

Because when the team actually takes the time to truly focus on creating agents that handle their most valuable GTM team processes or “jobs to be done,” they will start seeing the true promise of AI.

And now you have a whole team focused on an agentic operating system that runs predictably / all the time, and is governable, scalable, and can be tested and iterated on.

Where is your team today? Still in the AI operator phase, or starting to build toward an agentic GTM operating system?

I’d love to hear where you’re at and also if this messaging is resonating with you.

Latest from LinkedIn

Join the spicy convo

What We Are Cookin' Up (for real)

This past weekend I threw a little pumpkin carving party for my daughter and her friends (plus little sister tagging on).

I made the following spread:

• Simple Mills apple muffins (I added diced fresh apple)

• Chocolate and butterscotch chip oatmeal cookies (which I subbed coconut sugar for the brown sugar)

• Homemade cereal snack mix (I used Cheerios instead of Chex and left out nuts. I also used less syrup to make it less sugary and used part algae oil, part butter)

• A charcuterie board

• And some cute little mandarin “pumpkins” and “banana ghosts.”

I thought the fruit wouldn’t be a hit with the kiddos, but they actually were! So much so I had to make more ghosts on the spot.

The pumpkins were so easy. Just peel a mandarin or clementine and add a pretzel (or piece of celery) stick in the middle.

The ghosts were just a banana cut in half, then into fourths, and I added mini chocolate chips for the eyes and mouth.

You can find more family friendly Halloween treats (that aren’t candy) here!

We’re Hiring at CS2! 🎉

CS2 is looking for our next Salesforce architect. If you love all things CRM, data, AI, and automation and want to work with an awesome group of other ops experts to learn from and service some of the top companies in B2B, then our new role might be a good fit for you! Note: CS2 is a fully remote company. We are open to hiring from senior manager to senior director level for each role.

Find the job description here to learn more.

If you have questions, feel free to reach out to me directly!

Links We’re Liking 👍🏻 on LinkedIn

“Who killed the positioning?”

This is a sad example of what you shouldn’t outsource and how, without a lot of context your results from AI are just 💩

Loving the direction of the scoring model in Phave. (ps. Conversion’s is great too!)

Speaking of which, new AI marketing automation solutions are coming for Adobe Marketo.

Loveeeee these AI-themed work group costume ideas!

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

One challenge for some GTM teams is developing the right pipeline goals, then building the reporting structure to help them identify how to create that pipeline or uncover actions they can take, so they don’t miss those targets.

Oftentimes, it’s just finance sending the GTM an overall revenue goal. One that is aspirational or one that just takes their past performance and adds on a % growth they all want, and then marketing and sales are left to take those targets and create goals for their team.

Unfortunately, this is not precise enough, and a big reason is that the finance team isn’t working with the GTM Ops/ RevOps teams who are closest to the pipeline and funnel data that should inform those goals.

A second challenge for some GTM teams is that they have pipeline goals that were built off accurate data, but they do not actually use their data (including funnel, campaign, and pipeline data) as a team on a regular cadence so they can uncover risks to hitting their targets before it happens, or create action amongst the marketing and sales team to hit them.

This is often because actually doing the slicing and dicing on the data that is meaningful to provide action takes A LOT of work and it’s just picked up by looking at a simple dashboard in SFDC.

This summer, we worked with a client to tackle both of these challenges and with the help of AI are able to do it at a more regular cadence.

To set the stage, CS2 helped the larger enterprise tech company deploy a consistent sales process for their opportunity management and deployed our custom funnel tracking object (also known as Project Insight) to track all their inbound, outbound, and channel-driven customer journeys all the way to closed-won customers.

One of the driving factors for getting this funnel data was that the client has a very long deal cycle, but marketing needed a better way to understand how they can achieve those broad targets from the channels / campaigns they produce.

Now they needed more precise pipeline goals.

So Alison leveraged all of the historical data for the client and came up with a model that wasn’t simple. It took into account history on their deals, and various other inputs like slippage, seasonality, and outliers to develop opportunity conversion rates and pipeline goals for their fiscal year with accuracy. This model was also smart enough to know if, looking at their current pipeline in stages 2+, they have enough to make their revenue goal. This spreadsheet was on steroids for the amount of formulas it needed, and Alison’s brain + Claude Code made it possible.

Now they could work backwards to create a “MQL” goal.

Again, our client wanted a clearer idea of how many sales-ready leads they needed to achieve their target.

So instead of picking one average and calling it a goal, we looked at the range of how opportunities had historically been created (the slow quarters, the normal ones, and the great ones) and worked backwards.

That 1,700 number got everyone’s attention. Which was exactly the point! For the first time, the goal was tied to something real, and everyone could see the math.

The weekly funnel dashboard and pipeline review call is where the magic happens

After the goal-setting activity, Alison built an ongoing dashboard to review quarterly, monthly, and weekly funnel metrics against plan (with the aid of Claude Code). This was built in Excel and pulls from data derived from only 2 SOQL queries. Alison used Excel so we could better shape the data, and because they don't yet have a BI tool we can use to build this out. But the goal is to eventually have it built out in BI.

Building the workbook alone (without AI) would have taken hours and hours previously, but was done in an afternoon.

Every Friday, our team refreshed the data in the Excel dashboard, generated a word doc with a set of analysis /insights to review and we then led a call with the team to review the insights.

Because we also pull in data on target accounts and campaign influence etc we have been able to draw a lot of different types of analysis and insights for the team beyond just the the obvious funnel, win rate, pacing metrics

Some of which include:

• Marketing-sourced vs marketing-influenced pipeline: pipeline marketing created, plus deals that marketing campaigns touched, through to Closed/Won.

• Channel / UTM attribution: which campaigns and channels brought in leads and opportunities.

• Contact Us vs other inbound: the Contact Us form tracked as its own channel, plus non-Contact-Us MQL volume and conversion by account tier.

• Speed-to-lead: how fast each rep follows up on Contact Us leads, and how response time affects conversion (the 4-hour finding below).

• Target account tiers: funnel volume and conversion for Tier A-D accounts vs untiered accounts.

• Monthly business review (MBR) snapshot for the leadership meeting.

• Pacing scorecard: where H2 stands against goal on each metric, the change since last week, and marketing’s share.

• Automatic “why are we behind” analysis: runs for every metric under pace and checks three things:

  • Has the marketing vs sales mix shifted?

  • Are we ahead on kW only because of one or two large deals (deal concentration)?

  • Which reps are carrying the number or falling behind?

• New Stage 3 deals since last week: each deal linked to Salesforce, with source, status and expected close date.

• New Closed/Won deals since last week: each linked, with how it was sourced and the specific campaigns that touched it, plus blank bullets for live discussion on the call

Example of how insights drive change

By September, we were looking at speed-to-lead on their “Contact Us” form, the channel behind about 70% of their marketing-sourced Stage 2s and Stage 3s.

And we found responding within 4 hours doubled conversion!

So the question the GTM team was trying to answer now shifted from “how do we get more leads?” to “how do we get the most out of the leads we already have?” A very different (and much cheaper) problem to solve. And created actions for the ops and sales team to help improve this.

Where they are now

By late September, marketing-sourced pipeline was right where it needed to be.

But the real success story is how we had built them a process that helps guide the whole GTM team and fill them with confidence in achieving goals.

It also created a set of data the whole team could rally around and create action.

This is nirvana for many GTM teams, and having confidence in the data the models are built upon is the first challenge to tackle.

Liked what you saw and have a question? Post in the comments or reply directly to this post.

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

Last week, I shared one of ’s recipes she has been making and I am doing it again!

You all know how much I love soup. I have actually been reading more about Chinese medicine and they believe in the health benefits of eating broth-based soups, especially for your gut and overall digestion.

I have been eating soup for lunch pretty much every day recently when I can and I do find it much more soothing and helpful in aiding my digestion over powering down a cold, raw vegetable-based salad.

Laurie shared a recipe for a gluten-free chicken noodle soup and almond flour biscuits that I am dying to replicate. How good do those look?

Here is the recipe for the soup.
Here is the recipe for the biscuits.

Laurie added frozen peas to her soup, which is a power move I do with mine as well!

We’re Hiring at CS2! 🎉

CS2 is looking for our next Salesforce architect. If you love all things CRM, data, AI, and automation and want to work with an awesome group of other ops experts to learn from and service some of the top companies in B2B, then our new role might be a good fit for you! Note: CS2 is a fully remote company. We are open to hiring from senior manager to senior director level for each role.

Find the job description here to learn more.

If you have questions, feel free to reach out to me directly!

Links We’re Liking 👍🏻 on LinkedIn

“Second wave” for operationalizing an AI SDR.

Has there been a “blue washing” of marketing?

Where to upskill when moving into GTM Engineering.

This is a funny dig at Salesforce…

A bake-off of classic Salesforce vs ClaudeForce (by Anthropic) vs Replit vs Lovable

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

One other thing that isn’t really a project and is only alluded to briefly above but more of a practice is that our team is tapping into the APIs of our commonly used tools (HubSpot, Marketo) and the CLIs (SFDC and Clay) way more and it’s big a hugeee time saver for projects. If you are able to do this at your org please do as you will be surprised at how tedious tasks, troubleshooting, auditing, and even building can be done so much more easily.

Over the next few months, we will also be working toward developing more standardized agentic processes that support entire “jobs to be done” for the GTM team. Stay tuned!

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

Last week, shared a few of her go-to recipes and this one Shred Happens for “unbelievable slow cooker beef” caught my eye!

It’s slow cooker season. With busy workdays, school events, and sports, I know many of you are leaning on your slow cooker.

And this recipe, which is super simple, healthy, and adaptable, is something I can completely get behind.

You can also make it in your pressure cooker!

Served up with roasted sweet potatoes and veggies? YUM!

Links We’re Liking 👍🏻 on LinkedIn

Agree - this is genius.

Uhm - hello! I think I was missing from this Marketo Mafia ; )

A great example of why you can’t trust all the content on LinkedIn

A very true but harsh reality for female leaders.

I need to plan a gals trip stat.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

For most of the day after the main sessions, partners split out by solutions partners versus data and integration partners.

The big focus for solutions partners was around how to partner with their sellers and how to sell into the enterprise. I spoke on a panel about selling to the enterprise given the very quick success we had with selling and implementing Clay for some of our enterprise clients.

Lastly, the day ended with awards!

And guess what?!

We won the Studio Partner of the Year Award. After only 9 months of being a partner. This was a genuine surprise and we felt so proud and stoked!

Since I saw Clay in 2023, Ireally wanted to be a partner because I knew the power of the tool and what it can mean for clients even when it was so primitive. It’s expanded as I projected, and our whole team has been doing such a great job selling, implementing, and using Clay for our client base. Especially Xander, Christie and Sarah, who were all early adopters.

I have to say the best part of partner day was all the listening the Clay team did with partners, including us, about how things can be improved. All the way to Varun their cofounder, we were heard. But we also let them know how great of an experience its been so far.

Sculpt Conference Day

Back up to the city we went.

The morning started with the public seeing the same roadmap and announcements as we got the day before. Here are the few things you should know if you missed it:

1. Workflows will be coming out of Beta soon and should be adopted ASAP.

2. Workflows + Audiences are the “future” of Clay instead of just tables.

3. Sculptor will be retiring soon because…

4. Clay announced their new agent builder which you can chat to and architect workflows for you.

Part of this also includes a knowledge hub which will act like a context layer in your org and continually learn what is working. For example if you craft a campaign targeting your ICP accounts and Clay picks up on you booking demos with telehealth companies for example, the agent will suggest adding more telehealth companies to the campaign.

The only kinda bummer to this is using the agent builder will require credits. So additional credit budgeting for using it will be necessary.

5. Finally - Clay announced their upcoming AI SDR Agent.

This agent which you can add to your website will help chat, qualify, give live demo and scheduled a meeting for your inbound prospects. The agent will use voice chat but no (creepy?!) visual of a person doing. IDK why but this is my preference honestly. With this feature, I see a big point solution consolidation play for clients if they decide to use this.

For more details on the announcements, see this blog.

The rest of the day was full of sessions and roundtables. But I didn’t get to attend too many since I was busy meeting folks IRL, which was my main goal.

I did attend Jordan Crawford’s session, though on personalization and segmentation for outbound (done right) and I liked it a lot!

A few highlights:

1. Finally meeting Brendan Short from The Signal IRL, and we filmed a video together!

2. Finally meeting Emily Kramer from The MKT 1 Newsletter IRL, and she is such a joy!

3. Seeing some of our awesome past and present clients for fun chats and strategic discussions.

4. Chatting with other agency owners/operators like Troy from GoNimbly, Patrick and Mathias from The Kiln (now 2x), my buddy Jill Melchionda from Marketbridge, Jordan Crawford from Blueprint, and Nico from Scalantec.

5. Last, but not least, I had a long time listener to our podcast and newsletter subscriber, Carlton Parsons stop me and say hello and also tell me about how he loves our content and how he has been using it in his roles. This made me so happy I had to take a picture with him!

Overall, the day was a buzzy, excitement (and sunshine) filled day in the city and I loved it!

Hope to see you all next year at Sculpt!

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

It’s October. And I just had my first purchased Pumpkin Spice Latte this week from the only place I will get one near me since most are bleh to me.

And I am making pumpkin breads and pumpkin pancakes for my kiddos with some help from some boxed heroes.

I love this pumpkin bread/muffin mix from Simple Mills. I made it into a brownie baking pan this week to cut into squares for Ava’s lunchbox. One top tip is to add more canned pumpkin. I added 1/2 of a can, and it still came out so well.

I also like this pumpkin spice oatcake mix from Seven Sundays. It’s nice to stir up some pancakes quickly on a fall weekend, and this one is so good. All you have to do is add water, but I add eggs to mine as they suggest.

Links We’re Liking 👍🏻 on LinkedIn

This marketing campaign. I’m crying.

87% of men in tech have mustaches?

This post made me LOL. But really though - pick up a book/kindle!

When your head of finance actually understands marketing.

What the Loveable revenue team is building and shipping.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

So here's how it went down.

Sunday

Welcome dinner at La Posta. An Italian-ish, coursed, family-style dinner filled with fresh dishes made with ingredients from the farmers’ market. Swoon

Was a great way to settle into the offsite together!

Monday

Welcome presentation from Charlie and Me
Monday morning we went straight into sessions in the meeting room at La Bahia. Charlie and I covered a deck on the latest thinking on GTM AI, an updated version of our GTM Ops framework with agents at the center, and some of our top learnings from the year so far.

One of the biggest learnings we discussed was that we feel like we were “wrong” for talking about AI maturity through levels. It was too confusing for clients, as they may think Level 4 (autonomous agents) should always be the goal, and that is not right. If anything, we see very little use for truly autonomous agents in GTM right now, and also some processes should be using a Level 4 of AI. We knew this. But the talk track was not clear enough.

Team Activity - AI Use Case Mapping to GTM “Jobs to be done”

We decided to go as analog as we could during this session. When Charlie and I showed the “new” GTM Ops framework, a big change to it was that we identified the key “jobs to be done” that a GTM team has because those jobs are where we can identify specific use cases for agents to support those jobs.

The “jobs to be done” of the GTM team we nailed down were:

1. Find, prioritize, and sell to the best accounts.
2. Convert inbound demand fast.
3. Win more of the deals already in the pipeline.
4. Retain and expand customers.
5. Run programs across every motion that produce pipeline.
6. Analyze GTM data and plan revenue.
7. Keep the operating system running.

For this session, we had everyone write down their agent ideas and didn’t tell them the jobs to be done. Then we had big sticky notes with the jobs to be done as the category and had everyone put their use case/agent ideas on the job that matched it the best.

Afterward, we had one person on the team try to group them all to see if we could find overlaps in the ideas for the agent - signaling to us how important and useful it would be for all clients.

This was such an incredible exercise, and we came up with a ton of great ideas!

We selected three to build. We discussed the top ones as a group, and each of the three teams, split by roles at CS2, chose one.

The final 3 agents to build were:

1. Account Targeting Agent
2. Opportunity Forecasting + Sales Next Step Agent
3. Pre-opportunity Analysis Agent

Each team had time to brainstorm and build before we left the room and head to our next activities.

Walk at Wilder Ranch

Took the team to one of me and Charlie’s favorite places to hike together. But we chose the flat and shorter coastal walk since we have learned in past that not everyone wants an intense hike. lol The views though are epic!

Dinner & Pub Quiz (aka Trivia) led by Rouse the Riddler (Alison)

I honestly do not like too many sit-down dinners as people don’t mingle enough. So we booked a back space as a pour house with good food and this also allowed for us to do our our pub quiz led by Alison who organized it all.

It was so much fun. Thank you to Ted Lasso for inspiring me to do this with the team.

Tuesday

Team Activity continued - Build + Present Agent POCs
We wasted no time and starting at 8:30 am, each team went directly into building mode again to complete a POC of their agent back to the team.

They had to present 3 things:

1. The value / selling props of the agent
2. The prerequisites from a data and tool and foundation standpoint
3. A sample “output” or version of the agent

*Note - we had a very loose definition of agent and didn’t just think these would be autonomous agents. They could also be a process that triggers an agentic workflow or could be a series of agents.

For brevity of this newsletter - I won’t go into the detail but I will say all the agent ideas and POCs were great and ones we could build for any client and they will receive a ton of value.

CS2 Team Standards Workshop

After presenting the agents, Christie led a workshop on setting team standards and practices that we will stick to as a team. Every team member filled out a survey on what teams are doing well or could improve to help inform the discussion and creation of those standards.

We have values at CS2, but it was great to get each team to nail down their “principles and rituals” that Christie suggested. This means defining what standards we want to uphold as teams and what rituals we will do to support those. We left with immediate actions for the teams and are still working through those now since we ran out of time.

Beach Boardwalk

Tapped into our inner child and played arcade games, laser tag, and an awesome pirate-themed mini golf.

Sunset dinner and beers at Humble Sea on the wharf

It was a lovely afternoon of sunshine, so we closed out the night on the Santa cruz wharf just a short walk from our hotel. Some of us didn’t want to end the night too early, so we also had a nice chat around the fireplace back at the hotel before me and Charlie said goodbye. And the next day everyone departed back home.

It’s always a great time getting together as a group, and this year I loved the mix of work, fun, and lots of strategizing as a team (even the trivia!).

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

We're in a moment of produce in California that I love.

I make it a point to try to get to a farmers market because not only do we still have amazing summer fruits and veg still here (strawberries, peaches, melons, eggplant) we also now have fall’s finest apples, grapes, leeks and even some squashes!

The past two Fridays, I made an incredibly easy meal that just involves some cutting, a quick sauce, and then letting the oven do the work.

I made a tahini maple chicken and veg skillet meal and changed up the ingredients for both weeks.

The first week, I included leeks, sweet potato, and grapes. And the second week, I used fennel, leeks, eggplant, honeynut squash and some sad grapes going bad in the fridge.

I take all these chopped veg and cover with a “sauce” of 2 tablespoons tahini, and 2 tablespoons maple syrup, some dried sage, and some vinegar to thin it out. These are rough measurements and you can just eyeball it all.

I then take my deep cast iron pan and make a bed of the chopped veg, add the chicken, and then pour the sauce on top and cover it all.

I cook the skillet at 375 degrees F for about 40 minutes (or longer if you want it be crispier).

If you prefer to have an exact recipe, check out this one which is what inspired my creation!

The result is a dish filled with caramelized veggies, carbs, and meat. Delish and so comforting. I especially love the pops of sweetness from the roasted grapes and leeks. Yum!

Links We’re Liking 👍🏻 on LinkedIn

I too, rather see human art than AI art. ; )

Interesting tech company gossip here…

Sharing experience of using Claude Tag in GTM.

Interesting (and more positive view) of AI “taking” our jobs.

I love Mrs. Doubtfire and I LOLed at this one! Cringe but funny.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

But some might still be thinking.

Well, what is “the foundation?”
And what “data” do we really need to fix?

And the answer is “it depends.”

Because an agent that is useful typically helps support a specific job to be done. And that job has its own set of “prerequisites” for the following areas in order you to set up a proper agent for it.

You need:

Context

Unified Data

Tech and Infrastructure

Process

And you need a human (or team of humans) with good judgment to help define the strategy and determine what the agent is doing is “right.”

For example, let’s talk about one of the most valuable use cases we have right now:

Opportunity Risk Identification and Forecasting

You need to give it context on what risks and close-lost reasons you often see. You also need to tell it what good things to listen for and even examples of what a good conversation with a prospect looks like and what is the expected key phrases or even deal milestones that contribute to an opportunity closing.

You need to give it unified data. This means giving it the ability to read call transcripts and emails from prospects who are part of an open opp.

You need the technology that allows you to have those transcripts and email history and be able to access that data from where you are building your agent.

And you need to define a process for how you will deliver risk details to AEs and management and potentially even update open opportunity data and stages automatically in your CRM.

All of this needs to be done and figured out before you can even think about touching Claude Code.

But this honestly should sound familiar.

Most important ops processes HAVE ALWAYS required this.
And we only harp on about this constantly because we know how important the foundation and data are for making any process or technology work in GTM Ops.

But people in GTM sometimes think AI can outsmart it all and not need a lot of that to operate well.

WRONG.

I am sorry.

At least for now.

And on top of that, once you start building your agent, you need the experience and judgment to determine whether that agent is working well or how it can be better.

Technology does not equal strategy.

AI does not equal less structure or foundation.

AI also does not mean simpler.

AI means smarter to me. And we still need to define the strategy and build a foundation to make that agent work.

Because really, that agent is just trying to replace a human’s work.
And we all know how hard it is to work, when the process is broken or when the data is a mess and not clearly accessible or defined.

It’s the oldest problem. And it remains.

But now we can also tap into AI to speed up creating that infrastructure, fixing that data, or getting access to data we never had.
And that is the coolest part.

Anyway - hope that was a good little thought experiment for ya.

This week I have been super busy leading up to our team offsite in (hopefully) sunny Santa Cruz, CA. One of my favorite places, and I am so looking forward to showing members of the team.

But also because of that, next week I will be taking a break from the newsletter. But will be back after and hopefully sharing more about what we did during our offsite.

Stay tuned!

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

I am not one to fall for viral recipes, but I did cook one this past weekend.

I had seen someone mention in a Substack how she cooks Emily English’s cottage cheese meatballs in tomato sauce almost weekly recently, so I had to give it a try.

I often take long breaks from Instagram for reasons I don’t think I even need to explain anymore (lol), but in the past I would follow Emily, and her recipes are SO GOOD. I love that she is actually English (not just her last name!) and her recipes and shares on food remind me of all the good parts about English or European cooking but just a bit more healthy.

The more interesting thing is this recipe is a riff on Hailey Bieber’s recipe. So this recipe is viral x2.

As usual, I also riffed on these, and instead of adding in breadcrumbs, I used gf oat flour I made in my blender to make them gluten-free. And instead of making the sauce I used a jar of Rao’s Caramelized Onion Sauce. I served ours with Goodles gluten-free penne, and it was delish!

Recipe ratings from the family:
Parents: 10/10 delish!
Kids: 6/10 (yummy meatballs but we still don’t like tomato sauce unless it is on pizza)

Links We’re Liking 👍🏻 on LinkedIn

I love all the advice in this post. Especially the one on not tearing other people down!

Why your AI output needs judgment.

Still need to dive into this “self-updating” CRM.

The case for tracking buying groups.

Are people just calling their outbound tools GTM platforms?

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

And the most recent tool we created, I feel like, could benefit any ops team - even a team of one!

That tool is our roadmapping and hours planning tool.

At CS2, we use a retainer model with clients where they sign up for a specific range of hours that we can use per month. This range is typically dependent on the budget of the team and translates pretty well to how many projects and everyday support we can take on.

But oftentimes the client will request bigger projects or have more needs than what the monthly retainer can allow for, and we need a way to show them why potentially moving up in hours or range would help them get those projects done quicker or on their expected timeline.

And a conversation or a Slack message doesn’t always get the point across, so we would sometimes need to plan it out in a spreadsheet just to show the project with the number of hours and how it would span across multiple months to get them done, or else some other items would need to be deprioritized.

And this scenario isn’t just relevant to us.

GTM Ops teams have multiple stakeholders and teams they are supporting internally.

All of them have different projects and requests being asked of the same team of resources, and GTM Ops teams often find themselves struggling to prioritize or even get the work done because they aren’t effectively communicating what they can even get done at a given point in time.

They also often underestimate how much of just their ongoing support and maintenance work is taking from that same pool daily.

Projects get stalled. Deadlines get missed. And the stakeholders that get the best experience are the loudest ones.

Sound familiar?

So I thought it would be useful to chat through not only the roadmapping tool itself, but also what processes you can put in place to help create an effective roadmapping tool.

Suggested pre-reqs for the roadmapping tool

1. Get the whole team using a project management tool.
   
   Even if that team is just you. Do I even need to talk about the benefits of doing this in 2026? Honestly, it can even be a vibe-coded tool, but the two features that are necessary are the ability to report out on past projects and pull them via API and then also the ability to track time to them. Which leads me into the next point.

   

2. Start tracking time.

   
   One benefit of time tracking we have found at CS2 is that it not only gives us insight into how we are meeting our retained hour commitments, but also gives us insight into how much time across the whole team specific projects have taken us and how much time is spent on things like maintenance and project management depending on the client. Those Slack messages or “quick requests?”

   That is valuable time, and operators often underestimate or don’t include this time when they are planning out their roadmap or committing to deadlines. Mainly because they don’t have any historical information on it. ClickUp includes a timer setting on all tasks and subtasks (which time gets rolled up under the main project). And it’s very likely whatever project management tool you are using does too.

   

3. Track ongoing master tasks for ad hoc requests.

   
   As operators, we spend lots of time in meetings, answering Slacks, or troubleshooting. You can have trackers for each of these buckets and even break them out by “types” (i.e., MOps, RevOps, or CS Ops, for example). Having these not only give you rough averages you can include in your roadmapping, but also give you insight into maybe which teams need more enablement or deflection tools. ; )

   

The roadmapping tool and how it works

1. First, we used Claude Code to create a database of project types (aka workstreams) that had the total # of hours for the project plus the length of time it took to complete the project and what tier the client was on. This used 4 years of ClickUp data, and because we also started to use ClickUp templates, it made the standardization of the data much cleaner. If you don’t have a history of projects and timeline to pull from, just create your own estimates. These are the projects that you can build into the tool that you can add as a line item on the roadmap.
   

2. We created a roadmapping tool using Claude Code that we added as a section under our “client hub/dashboard” that Charlie described in this newsletter. This is password-protected and hosted on Vercel.
   
   The roadmapping tool is made up of a few sections you will see in the image below:
   
   - The workstream library, which pulls from the database created above.
   - The workstream area, where you add in workstreams (either from the library or manually).
   -The roadmap - where it displays the workstreams and hours across the allotted month.
   -The inputs area (up at top) where we add in the starting month, # retained hours, total months, monthly calls average, and buffer %, which allows you to add on buffer, especially if you have a very complex client/systems/etc.

   

   

The nice part is a lot of it is editable. You can adjust hours, timing of the workstreams and move things with a simple drag and drop without doing any “excelpymics.”

Charlie rolled this new tool out to the team, and folks started using it instantly, including myself, so I could show why a client needed to increase their monthly hours since the delta for what they wanted done quickly was around 25 hours.

And all I had to do was send them a picture of the roadmap and the reasoning for why, and both VPs were able to go and get more budget from finance.

You can see I was a happy girl.

In the context of working in-house, this can be a great tool to reprioritize, ask for agency resources, or even try to get more headcount on your team. The visual nature also just paints such a great story. And you can create different versions by adjusting the hours based on your current resources/headcount versus your requested resources/headcount.

Obviously, there is a lot of work and tinkering I just glazed over here, but it’s really about the tool and the value it provides that truly matters.

But if you do have any technical questions or want help creating something like this, just let me know! You can reply directly to this newsletter.

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

I had a revelation post-lunch yesterday.

After a bowl of soup, I wanted a little sweet bite, so I took a piece of fresh medjool date and added a piece of apple and some almond butter.

The result?

It tasted JUST like a caramel or sticky toffee apple.

Just much healthier and easier on the teeth.

I told the foodies channel immediately.

Now my weekend project will be to create a little treat that my date-loving 3.5-year-old Lily (and me) can have.

And I found this date recipe for caramel apple bites which I think will be perfect!

Don’t those look delish?!

Links We’re Liking 👍🏻 on LinkedIn

I have been seeing all of these WINN.AI posts everywhere on my feed!

Corp Bro dancing into Friday will forever get me smiling.

Using Zapier does not equal native integration.

Love this “map the buyer, pick the the person” social posting approach by Clay.

Who will be the winner in the context layer?

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

After months of warning people about vibe-coded CRMs, I accidentally built one for CS2.

I say accidentally because my goal was to build an internal tool to track CS2 client health. Things like: how each client engagement is going, where we can improve, where we are crushing it, utilization, sentiment, what stakeholders are involved, etc. And then to surface actions for our team to keep improving our work for clients.

I talked about it a bit in this LinkedIn post where I explained how much of building agents is really just data work, and I showed this image of the data pipelines being built that feed into it.

I am pulling in Slack data, call recordings, internal client calls, call attendance data, ClickUp project management data, and delivery data. All running through a data pipeline where it’s sorted, analyzed, and stored in a centralized database for agents to reason over and to surface important information to Crissy, me, and the team.

Some of the highlights are:

• Stakeholder management: We can now track call attendance across clients, enrich the contacts with Clay so we can see their titles, and monitor to see if we are losing stakeholder interest or if important stakeholders leave the company and we need to engage another point of contact.

• Sentiment analysis: All the positive and any negative statements on calls or in Slack are captured and analyzed to help us understand how our clients are feeling about the engagement.

• Client Health at a Glance: We have a 4x4 grid and score to map clients to see which ones have the most opportunity to do more great work vs. which clients need extra focus

• Quick TLDR summaries: Each with quotes and data on every client so Crissy and I can easily get a sense for how everything is going across all clients without having to get our team to write out long updates each week.

It’s magical, has been a lot of fun to build, and has instantly become our daily hub to monitor and improve CS2 each week.

What does this have to do with “vibe-coding” a CRM?

Well, as I built this out, I realized I basically built the plumbing to gather a lot of the typical data we would need for a CRM. The only thing that seemed to be missing was the sales/pipeline management.

So I worked with Claude Code to build it, and in a couple of hours, I had a place where I could manage all new business deals and any customer deals.

What was awesome about this is that because I already had all the data pipelines, analysis, and tables built, it all could flow into the deals.

For example, if a client needs to expand their relationship with us, it is usually captured in a call or Slack message. I was already bringing that data into the Client Health system, so then I just made sure it was also tied to my deal database.

This means that on every deal, I have all the activity tracked, the sentiment of the client, who we are speaking with (and stakeholders we are not but should be), what is going on at that client, auto-stage progression, etc.

And with AI reasoning over all of that data every day, it can help give the leadership team at CS2 a quick summary of what we need to know to progress things.

I didn’t set out to do this, but I built an agentic CRM for pipeline management that gathers all the context I need, moves stages, and helps me manage my client contacts and sales pipeline end to end.

So that is the long story of how I use agentic coding (my preferred term over vibe coding) to build the “new CRM” at CS2.

Should B2B tech companies create their own CRM?

If you are reading this and are anything like our clients (scale up to Enterprise B2B tech and AI companies), I do not recommend you do all of this. But I recommend you do part of this.

It’s quite nuanced, so keep reading.

What you should NOT do:

TLDR: You should not agentically code a CRM from the ground up.

If you have SFDC or HubSpot as your CRM, keep it.

The underlying architecture of what I built is great for a 15-person company like us that has one person managing the sales process (me), relatively low-complexity internal business processes, and is ok with this running locally on my computer (for now) and using up ALL my Claude Fable usage every day.

My build would not be ok for a company that has tons of salespeople, lots of complex business processes, requirements around permissions and access, 100% uptime/reliability needs, requirements for a support team, a UI for sales/analysts/etc processes, and a ton more that true enterprise CRMs offer.

The cost/benefit just does not work out once you factor in the time, effort, resources, and lost productivity you would experience if you tried to rebuild Salesforce or Hubspot from scratch and then release it into a bigger org with salespeople and many other internal stakeholders.

There is a reason why Anthropic uses Salesforce. And there is a reason they have partnered to create Claudeforce.

What you should learn from this (and can probably apply to your org):

TLDR: build your “intelligence layer” custom yourself (for now). Or get CS2 to build it 😀.

The intelligence layer is all the stuff I talked about at the beginning of this post. The data pipelines to gather unstructured data from various sources, AI reasoning/analyzing that data and structuring it, the dashboard/views/automation driven by this data. You can and probably should build this.

The core CRM architecture is the foundation. It’s where all this stuff lives: pipeline, forecasting, users, business processes, integrations to other tools (like finance, marketing, etc). And it’s secure, reliable, and has support. You shouldn’t build this from scratch.

How to build your intelligence layer:

1. Map out all your important unstructured contextual customer and prospect data

Things like: call transcripts, emails, internal Slacks, etc. You have likely been piping some of this into SFDC which is fine. But to truly extract the value out of the data you need an agentic workflow that can gather the data > run agents to add structure by categorizing or analyzing the unstructured data > add that data into its own table.

You need to find all the sources of this data, where it lives, if the tool has a CLI/MCP you can use, how the data is structured and stored, API call limits, etc

Figure out what ingredients you have to play with.

2. Then extract signals and data from the unstructured data and structure it in tables

For example, in my pipeline, I have a script that pulls call recordings and Slack messages each day into Claude, which spins up parallel agents to reason over the transcripts and messages and is trained to extract postive/negative sentiment and expansion/churn signals and logs them in a “signal” table with different columns tracked.

Here is my signal table:

It’s this unstructured > structured step that is so important.

And right now, IMO best to build custom yourself vs. buy a tool that may not meet your needs.

You want to own how this operates. Even something simple like what a positive sentiment signal is for your business may be different than my business. A tool may lock you into their way of it working which you’ll get frustrated with.

An example of what you could build now would be:

• Pull in all your call recording, email, and slack data related to all your prospects with active pipeline

• An agentic workflow extracts the important deal risks, opportunities, extra context, stakeholders who should be there but are not, any issues around the deal stalling, etc

• Logs each piece of data in a database you can reference

• This is surfaced in Slack alerts, dashboards, etc (this brings me to the last point)

3. Build dashboards and processes based on this new dataset

For example, we surface our sentiment data in the dashboard in the image below. We also use it in a client summary we review each week, and it powers a client scoring model that builds out our 4x4.

So to sum this up:

Step 1: Figure out all the rich data you have that isn’t really structured in a way that you can use yet.

Step 2: Build data pipelines to gather it, and then use agents to analyze and review it and structure it.

Step 3: Take the now-structured data and surface it, summarize it, display it, and action it however you want.

You now have built an intelligence layer that you can use and sits on top of your underlying core CRM architecture. The best of both worlds.

It’s a fun time to be a builder. But don’t build the wrong thing. 😀

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

Although your thermostat might tell you otherwise, according to Starbucks and grocery store aisles, it is pumpkin spice season.

Although I will be skipping out on the Starbucks sugar, this week I started to put some chai masala on top of my homemade flat whites, and this past weekend I made this pumpkin cake (without the frosting) as a homemade snack for my kids’ lunchboxes or after school.

Side note: I realized I was sending my daughter to school with just some chocolate masked in the guise of a “healthy granola bar” so am going to attempt to make her more homemade snacks or replace with things like a protein drink or yogurt pouch for more protein.

For the cake, I subbed the oat flour with gluten-free flour, I used only cinnamon instead of pumpkin spice, and instead of adding the frosting, I topped the cake with hemp and pumpkin seeds.

The result was a super moist and slightly sweet cake that has stayed good for 4+ days now, and it didn’t need the frosting at all!

Links We’re Liking 👍🏻 on LinkedIn

Life flashing before a sales rep’s eyes. So real!

Your enterprise AI strategy without clean data.

Your next valuable AI project will come from chatting with a BDR.

Guess who worked on the operations of those 3 nurture tracks? 🙋‍♀️

If corporate America was run by women.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

1. Understand when to be fast and when to slow down.

2. Confidence is just as important as competence.

3. Pretty much everything we do is about data.

4. Never think a big project is done.

5. Just because you CAN do something doesn’t mean you should.

6. Strike a balance of simplicity while still being cutting edge.

7. Wear the shoes of your stakeholders constantly. Build with them in mind.

8. Update your execs/stakeholders more on what you need to be doing, rather than just what you are.

9. Know what you shouldn’t be spending time on just as much as what you should be.

10. Never take an operational request from someone without challenging it or understanding the reason behind it

11. Never overestimate how much stakeholders know about operations.

12. Buy your infrastructure, but build or own your intelligence. * A new one I talked about here.

13. Your neediest stakeholder isn’t always the highest priority.

14. The best form of documentation for a project is one that you have time to create and that people will actually use.

15. Your “ways of doing something” should always be challenged.

16. Your career built off expertise in specific areas/“tools” will never grow as much as if it were built off “how” you acquired your expertise.

Does your neck hurt from nodding so much while reading this list? 😆

Would love to hear in the comments on this post what your most important lesson has been over the years!

Latest from LinkedIn

See the Full Post Here

What We Are Cookin' Up (for real)

Ok, so this week I am not talking about a recipe but rather an interesting new product I tried.

Lately, I have been always doing all my grocery shopping online unless it is for some veggies and fruit at our local farmers market on Saturdays if I have time to go.

But one thing I miss about pre-kid life was just roaming slowly down the aisles of Whole Foods looking for new products to try.

So lately I have been trying to figure out ways on their app how I can browse new products. This is sometimes using the trending or seasonal product areas, but this last week I just did a search for “protein” to see what comes up in terms of protein sources outside of my normal chicken, salmon, ground beef etc.

And I was pleasantly surprised when I found this new soy-free tofu by Big Mountain Foods.

It is soy-free because it is made out of fava beans. And it packs a big protein punch without any carbohydrate content unless other sources of non-meat protein like beans or even dairy.

I have to admit, it is not great tasting while it is in its raw state. But I covered it in sesame oil, miso, and rice vinegar and cooked it in the oven until crispy, and it was a yummy replacement for firm tofu and a great quick protein I used this week on top of my salad.

Definitely check it out if you are into tofu and looking for something new!

Links We’re Liking 👍🏻 on LinkedIn

This behind-the-scenes look at the things this gal has done for her job is interesting!

SFDC and Anthropic announce Claudeforce. So interested in learning more about these “37 prebuilt skills for sales.”

Knak’s upcoming webinar on automating campaign execution if you are interested!

CRO’s are being forced to become technical.

Love this share from Ashley Ward at LeanData on how to build an automated process for customer proof requests with UserEvidence and Claude.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

I will start with a little bit of insight into how I got to the lesson I want to share.

All of you probably know that my husband (Charlie) and I both run our agency CS2 together.

And we are also parents of a newly 7-year-old and a 3.5-year-old.

Our 11 years of running CS2 have been a very “equal” split of the responsibilities of the business.

And when you run a company with your spouse, the work and mental load follows you basically at all times.

As our kids have gotten older, and their needs have gotten higher and their schedules have gotten fuller, we have continued to try to pull in resources to help us spend 40+ hours on running the business.

But, as many working parents know, you will have things that will still fall onto your plate physically, mentally, logistically, etc.

This summer has been a great example of a “perfect storm” of how much it can take a toll.

From traveling to Europe for Charlie’s Mum’s wedding…

To camps starting later and ending earlier than our typical “school and after care” days…

To 2x per week tutoring for our daughter who needs the extra help…

To having all camps end at the same time our preschooler is out for school break for 2.5 weeks and having to try to piece-meal babysitters and my Mom together as best we can…

To organizing our daughter’s 7th birthday after coming off a week in Tahoe to fill the gap of both girls off school...

It’s been a lot.

And when I have been operating in it. I just go into operating-only mode.

Until I looked back at the last 8 weeks this week with Charlie, and I just feel upset.

I wasn’t upset because I wasn’t working hard and putting in the same effort despite the distractions.

It was because I realized that with all the mental load of life and work, I was only using hard work as my barometer for good.

And as a leader, this is not the right measurement scale.

Working hard is only one small piece of what is important.

And if you are any level above Director, you will realize this.

What I realized in myself is that when I am in those phases of high mental load, I tend to stay in the path of least resistance.

I do the things that are easy to execute. (Client work for the one account I am now, answering team questions, calls, 1:1s)

I do the things that I have a system already for. (This newsletter, company updates, quarterly calls with clients and strategizing on them)

But during these phases I really struggle to take something from blank page to execution. Especially if there is no client or anyone else telling me to get it done.

I also avoid focusing far out into the future.

And I also get easily distracted by all the aforementioned “easy to execute” work.

Does this mean I was coasting? No!

I was still working on important things and having tough conversations, and delivering great guidance.

But again, my barometer for a good week was how much effort I was putting in and how much I was doing.

So I decided I needed to change how I was measuring my success.

Here are the questions I will use to help me measure it:

- Am I providing value to the business and the team?

- Am I putting my energy into things that deliver the highest amount of value to the business first?

- Am I working more ON the business rather than just in the business?

- Am I taking brand new projects from idea to completion as efficiently and quickly as possible?

- Am I carving out enough time to teach and guide my direct reports while also learning myself?

- Am I spending part of my week focusing on things that benefit the long-term success of CS2?

But I can’t get there without some changes.

And here are a few things I am going to do to ensure my success:

- I am slowly handing off the 1 client I have left on my plate to an awesome new hire who is starting.

- I am going to measure my weeks based on how much I helped my team, how many important initiatives I moved along, and how many new ideas for the future I brought to the team.

- I am going to leverage AI even more in areas to help me speed up work that is easy (ie summarizing slacks, getting quick answers to things etc) and also use it to help brainstorm or plan with me as a partner to eliminate some of that blank page syndrome I get.

- I am going to protect my calendar from things that don’t provide value, but give more time to my direct reports and team with more personalized coaching and guidance so that every client is getting an even better experience and not just one that I was advising as coverage.

- I am going to take dedicated time out of my week to solve or improve areas of the business or things that will help with the long-term success of CS2.

Now, I know this is all a bit personalized to me.

But if you are reading this and you too are feeling a bit crispy (aka burnt out) at the end of the week but using only that effort level as your barometer of success, then you may want to rethink some of your own habits and goals too.

Especially now, when there is so much pressure to outperform AI, defend your job, but also make sense of a lot of changing playbooks and targets, it can feel like a lot to do and manage. Both workload and mentally.

But I caution you to really avoid just tackling this with JUST working harder.

It needs to be working smarter and delivering the most value.

Does that require hard work? Yes.

But all hard work isn’t your best work.

As a Type A perfectionist, it’s hard to put these vulnerable feelings and reflections out to the world. But at the end of the day, we all have areas for improvement, even when we are doing good work.

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

I asked the team this week what they have been cooking, and I got a great answer from Claire:

I love it because I am a huge fan of sheet pan dinners! I also loved a one-pot or skillet dinner, which is what I do often too.

But sheet pan dinners are so great because they require very little work and clean up which is a win for pretty much all of us right now who are busy with summer ending, kids back at school, traveling for work, or evening activities starting.

Plus a lot of sheet pan dinners can be thrown in tupperware if you're so busy you need to eat dinner on the go.

Here is a great list of 20 sheet pan dinners from a food blogger I like to get you inspired!

Links We’re Liking 👍🏻 on LinkedIn

Mastering Claude Code from the creator of Claude Code.

I totally agree with these 6 silent growth killer of GTM B2B SaaS

This is all a bit too “simplified” to me but I like the image!

The bottleneck in GTM Ops that AI cannot solve.

As a Mom, I totally get how this can happen.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

Oh, and think it’s just us who are against AI content? Take a look at some of the other companies cracking down…

Clay recently just posted about their internal AI Writing Policy (which pretty much aligns to how we think too!)

Also, Anthropic just announced they have started embedding machine-readable watermarks in text from new Claude models on August 2, 2026.

These aren’t really a traditional watermark but actually a use of words, so it’s very hard for a human to detect this. You can read more about that here.

And lastly, Substack (yes, where you are reading this) now uses Pangram to detect AI written content on posts and even gives you the ability to display this with a message on all your content.

Of course mine scored 100% real. : )

There are so many great applications for AI, but when it’s used in places where a human seems more appropriate its just going to fail.

Last week I was on vacation with family and friends in Lake Tahoe and it was such a great reminder about how at the end of the day, people still crave “humanness.”

There is nothing more satisfying than seeing nature.

Or making food with friends and enjoying a laugh in the fresh air.

Or feeling the warm sun and grainy sand between your toes.

Or hearing the laughs and giggles of your kids and family waft through the air.

AI can’t replicate those moments. And it’s why some of the leading brands, even AI companies, are not using perfectly curated or robotic marketing.

They are actually displaying real humans front and center in their marketing.

Clay and Anthropic are great examples of that.

So again, in brand, content, communication with people, and writing. AI loses.

But leveraging AI to do tasks, to find data, to make sense of data, to automate manual work, to segment your database, to prioritize accounts, etc is awesome.

So yes, humans still have a purpose in GTM. And over time it will become clearer and clearer where AI isn’t the best solution in GTM like this one.

Am I wrong? Let me know. Would love to hear it!

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

My favorite dessert is hands down ice cream. There is no sweeter thing than chocolate itself when it comes to dessert.

However, I do find that most ice creams, even if I have a small amount, leave me feeling not my best due to all that sugar.

This is ok, and I am not demonizing ice cream.

But, if I can have a sweet that also doesn’t send my blood sugar through the roof, why not, right?

So the past few weeks I have tested out making my own “ice creams” that are no-churn and just set in the fridge.

So far I have made one with coconut milk and one with cottage cheese.

The verdict from my family is that the coconut milk one is better due to its creamy texture. And I found the recipe on Instagram here.

So if you are looking a hormone-healthy and nourishing ice cream recipe that is also easy - here it is:

Chocolate Tahini Date Fudge Ice Cream

1 400ml tin of coconut milk (mix of milk + cream)
70g (1 cup) cocoa powder
110g medjool dates (about 8/10 dates)
1 large tbs tahini
pinch of flaky sea salt

Blend everything in a blender, pour into a container and allow it to freeze for about 4 hours before serving. If you leave overnight, allow it to thaw in the fridge for 30 mins before scooping.

To add a bit more pizazz to it, I added crumbled up Simple Mills cookies and added some chopped dark chocolate to it, and it was so delish.

And I also remade it again this week for my daughter’s birthday dessert with crumbled up cookies, Hu chocolate chips, and replaced tahini with peanut butter! I also find you can use fewer dates. For this last batch, I used only 6!

Links We’re Liking 👍🏻 on LinkedIn

GTM Engineers be like…

The reality of taking PTO as a solopreneur.

AI Slop in presentations. HALP.

Workflows are game-changing for Clay.

RevOps / GTMOps. “What, like it’s hard?”

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

Every 38 seconds, someone posts on LinkedIn that they have automated the entire sales process with AI. I built an AI agent to track this and send me a report every…. Actually, I’ll stop here. I just made that up, but you believed me for a second, right?

That’s because the sales process is probably the #1 use case companies are trying to solve right now (and therefore what tons of people talk about online).

Specifically: Find the best Accounts (w/ Signals) > Pick the right Contacts > Write the best messaging > Send emails > Create pipeline.

This is the dream we’re all being sold in this age of AI. And it’s not snake oil. There are real reasons we’re all trying to crack this code.

• Say you have 50 salespeople getting paid $250k

• That is $12.5m (plus another ~$4m full of loaded costs: benefits, laptops, travel, office space)

• Now imagine that 25% of their job is clicking around Salesforce trying to figure out who to sell to and 25% of their job is writing emails to those people.

• We’re talking ~$8m spent on “busywork” that should be easily automated by AI.

It’s definitely a goal worth chasing. But what the “I built this convoluted Claude Code process/skill/agent that runs the whole sales process and we don’t need humans anymore” content misses is:

1. It’s hard to get this right when it comes to the technical build and data. Not just kinda hard. Hard, Hard.

2. Humans still possess certain skills, discernment, nuance, and experience that Agents struggle with.

3. Very few people have the measurement infrastructure to know if the agent is outperforming the humans, or at least on par.

I see lots of people talking about #1 and #2. But very little discussion and data on #3.

Luckily, we’ve spent the last 15 years building the measurement systems and fixing client data, so we have some data on agent performance.

Let’s look at a real client example:

One of our clients has the following running:

1. Clay finding buying signals and pushing signals into a SFDC custom object tied to the Account. It will also create new Accounts and Contacts.

   1. *This was not mandatory. We just use this as a way to feed the agent signals to use in its prioritization and contacts to attach those signals to.

2. Actively.ai powering the Account Agent layer that

   1. Looks over all the Account data and scores Accounts

   2. Chooses several Accounts each week per salesperson (based on the best Accounts to target now, according to Actively)

   3. Picks the best Contacts for the salesperson

   4. Pushes the Contacts to a Gong Flow (and customizes the emails based on the context)

   5. Sends Sales a Slack message with their Contacts and the Gong Flow info a few times a week.

3. GTM team using our Unified GTM Pipeline Model (which tracks the funnel/pipeline in a Custom object in SFDC) to track the outcomes for each Contact/Account

   1. We are also tracking non-Actively Sales Outbound and all other channels

4. CS2 using Claude Code to run an analysis of Actively outbound vs. non-Actively to see how it’s performing.

So, is the agent any good?

TLDR: it’s nuanced.

The first thing we looked at was the conversion rate of Accounts from Sales Ready (this is when sales initiates the outbound) to Qual (a meeting is booked), non-Actively performs best. We’re seeing almost 3x the conversion rate.

So that’s not what we want to see. But if you look across all the stages (image #2), the picture gets a little better. Once there is a meeting, the conversion rates are essentially the same. It’s just getting that meeting that is tricky for Actively.

Important caveat: I do think there are some nuances to this data. For example, we tried to remove scenarios where sales may only track their work on Accounts they book meetings with and not track on the non-responsive Accounts. So we removed any where meeting booked date = outbound started date. But there is probably some of this still in the data.

That said, a 3x diff still isn’t great. It makes me assume that Actively isn’t as good as humans at picking the right Accounts and Contacts. Maybe it’s in the context provided to Actively, maybe it’s their model, maybe it needs more time to learn.

The one thing to still consider though, is that Actively did source 18% of the meetings. You could argue these meetings and the resulting pipeline would not have happened otherwise. You could also argue that the CAC on any customers acquired by Actively is much lower, as sales had to spend very little time to generate this pipeline vs. their own outbound.

Always a fun debate.

The next thing we looked at was how signals played into it. For this client, we are tracking a few signals:

• New important hires

• Account has an open position for their target buyer roles (and the job description calls out a buying motion for their product)

• M&A & funding

The results of this were very interesting. When Actively picks an Account with a buying signal, they convert almost 2x higher. So signals work (awesome!). But not so fast, for non-Actively (sales selected accounts with a signal) they convert a bit worse. So, in short, overall sales selected accounts, signal or not, still outperformed any accounts selected by Actively.

The other thing we looked at is overall “intelligence events.” Actively can help be a “co-pilot” on deals to help a rep research, choose, write, etc. When sales used Actively as a co-pilot, conversions did go up.

Now that could be that Actively helped OR this could be because they only used Actively on Accounts they were really trying to work. Is it correlation or causation? Hard to tell.

What is my conclusion on all this?

Well, first, it’s that data is hard. Analyzing data to find insights is hard. Sales is hard. Outbound is hard. It’s all hard. So a magic bullet to automate it all and get amazing results = HARD.

Maybe AI is going to get there, but for now IMO humans have a real place in the sales process and I think the data proves that.

But, that doesn’t mean to kill all your AI agents either. Agents can be incredibly useful throughout the sales process but given how most companies have very nuanced needs throughout the sales cycle, the agents need to be deeply customized. Otherwise, they will underperform your humans.

So my take (right now, I reserve the right to change my mind) is that I am less bullish on the vendors trying to own the whole process and enforce their methodology (sorry Actively) because you are tied to their opinion on how it should work and you don’t own enough of the intelligence layer.

I may get hate for saying this, but it reminds me of 6sense and how their 6QA’s and Intent are a complete black box with very little customization for your unique business. Actively is a step up and there are some areas you can change (they are also innovating on their platform fast), but it’s still outsourcing your intelligence to a vendor trying to solve needs for hundreds of different customers, which IMO gives you more generic/average (non-human beating) results.

For now, if I could make the final decision for my clients, I’d recommend we review each stage of the sales process and build custom agents for each process/goal:

• Account Summarization Agent: Brings together all the signals, history, call analysis, etc) in one place for Humans and Agents to learn and action.

• Account Priority Agent: Tracks behavior, and signals to highlight the highest propensity Accounts for humans or agents to action. It can provide a score / rating with reasoning for the selection / prioritization.

• Sales Outbound Agent: Reviews the best Accounts that sales isn’t actioning and proactively surfaces them to the sales owner with a summary, messaging, and email sequences.

• Buying team agent: Agent that reasons over Accounts being worked, finds Accounts where the Contact data is limited, finds/enriches the right Contacts, creates them in CRM to builds out the buying team.

• MCP and Skills leveraging your data: Gives sales a place (could just be in Claude/ChatGPT) to ask questions on Accounts and Contacts that have all your data available and the right context to answer questions correctly.

• Data analysis agent: Builds on top of our Unified GTM data to look at what is converting best to build a self-improvement system for all the agents above

• Call analysis agent: Reviews sales calls to find true closed lost/DQ reasons and gathers feedback to help improve the agents above.

Build each of these custom so you can fine-tune each one to your business.

To do this, you need an agent building tool like Clay, Agentforce, Dust, custom app, Claude/Open AI agents, etc. I’d recommend choosing the tool(s) that:

1. Have access to your GTM data already, ideally baked into the tool where you build the agent. Dealing with another integration, timing, API calls, etc is annoying.

2. A tool that your GTM ops team has permissions to build in quickly and to iterate (without crazy red tape. For example, we’re finding Agentforce is surrounded by red tape vs. many clients IT/Ops teams are more comfortable with Clay build-outs)

3. Ensures you own the Intelligence layer and can deeply customize (the layer where you can customize the agents, the tools they access, the data they access, how they reason, how they action process steps for sales, etc)

So there you go. That’s what were seeing and the direction we’re heading in with our clients.

It’s a fun journey and will share more as we learn.

Keep building :)

Charlie

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

We are basically in August, and many of us are dealing with some hot hot weather.

IDK about you, but when it’s super hot, I tend to crave some nice light and easy food. Especially for lunch.

And one protein I like to keep on hand is tuna.

I can be a bit lazy and just mix it with some pickle juice or lemon and add it to salad, but Laurie’s latest share with the CS2 foodies is making me think I need to be a bit more creative.

She has been making this canned tuna ceviche, which still looks easy and yummy! It can easily be swapped with canned salmon as well. Which I have to say, I am NOT into, but whatever floats your boat!

Links We’re Liking 👍🏻 on LinkedIn

Great perspective on wanting to measure AI ROI.

Hear from Clay’s own Head of GTM Ops on this recorded event.

This is how I feel about my LinkedIn inbox. 😭

Check out Everett Berry’s Days of Clay API posts. This is Day 6.

This made me LOL. Do not give me Outlook. ☠️

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

I do these periodically with clients to gather feedback on our engagement, but if there is no lengthy feedback, I can’t help but put my consultant hat on and quickly try to uncover opportunities with them.

And right now, my obvious focus is uncovering opportunities to do more with AI.

Instead of asking things like: “What AI tools are you using?” And, “Where do you see an opportunity to use AI?”

I like to ask these two questions: “What are some things that the GTM team is struggling with or has pain points?” And, “What are the goals or top priorities for the GTM team right now?”

Hearing pains or top priorities (which usually align to problems) quickly gives me insight into what operational work could align to those problems.

This comes naturally to me since I have been advising for so long. And it’s what all of us at CS2 aim to do as we work with clients. Solve the true business problems/needs instead of just receiving requests.

And with just 20 minutes left in the call, I used these questions that resulted in a conversation uncovering 5 AI projects we can (and likely will) support them on.

Here is a recap of the pain point/priority they uncovered and my initial thoughts on the project that will help solve this:

Pain Point: Manual and tedious work for campaign execution processes

Right now our team has been helping with some of the overflow work for their live event processes, which involve some manual updating of a Google Sheet to then update the members in the campaign just to get approved for an event attendance. Also, the main POC complained about list imports being “the bane of her existence” and that they have more webinars coming up for the team to operationalize too.

Project Identified: AI Execution for the Campaigns

Leverage Claude Code to set up a campaign execution process that starts with a campaign brief request and builds out the assets and workflows in HubSpot automatically. Also, create a skill that helps with cleansing and importing list imports directly into HubSpot. And lastly, potentially create a way for event managers to approve or decline registrants automatically through Slack that updates their event tool and HubSpot.

Pain Point: Account Prioritization for Sales

Right now, the sales team is taking a role in trying to select their accounts, but they don’t do a great job of it. A lot of that selection is more aspirational and based on the client logo than data-backed. They are also spending hours manually finding the key contacts at these accounts. Meanwhile, their data analysts are doing some analysis on which accounts just aren’t a product-market fit or which product a certain account can fit into.

Project Identified: Leverage Clay Research Agents and Clay for Reps

Considering the data analysts are generating great data on which accounts are a good fit and what product matches them, creating an Account Research Agent in Clay that will marry that data along with signals and engagement data to prioritize Accounts for sales, as well as identify what product and messaging they should use for that account would be so useful. Then, we can set up a workflow that can find the key buyers and create them under the account in SFDC. If we feel like the rep wants to do the contact acquisition on their own, we can look into giving them the ability through Clay for Sales Reps feature where they can run those on their own via Claude.

Pain Point: Net New Customer Logo Acquisition is Down

Right now, the client is struggling more in the last year with acquiring net new customers instead of just expanding their existing customer base. Right now, the whole executive team is creating their own hypotheses as to why, and the data that the team sees in Salesforce doesn’t tell a great story, or they uncover why it’s at risk only after it’s closed.

2 Projects Identified: Use AI to analyze past conversation info and emails for analysis AND Set Up Clay Research Agent.

Conversation data is a gold mine for uncovering what is going on with an account and the opportunity itself. We can first do an initial analysis with a Claygent to set an engagement score on that account/opportunity by looking at conversation data from Gong, SFDC, and email activities. Then we can use a Claygent to pick up and analyze any key risks or known reasons as to why the opportunity closed lost or on current open opps to see if they may close lost. We can also do analysis on opportunities to see what is going well or showing signs it will close won, and on past closed won to see if there are any trends or key factors noted by the prospect that signal it to close.

With all of this new data, we then can create a Clay Account Research Agent for churn risk or close-won likelihood in agent-managed fields on the opportunity.

Pain Point: Attempting to Contact > Connected Conversion Rate is Low

The executive team is also very concerned the conversion rate from “Working” to “Connected” is so low. And their team noted it would make them all very happy if they had clearer insights into why.

Project Identified: Use AI to analyze the funnel data (we built for client) for both outbound and inbound.

This one isn’t really groundbreaking, but it actually could be hard for a company that didn’t have their funnel/lifecycle data set up in a way as we do for clients. Because we have this clean data set, we can use AI to uncover trends, split by inbound and outbound funnels, to see if there are any insights on the overall fit of the prospects, the type of follow-up or outreach they are getting, including volume, the different channels or campaigns they are part of, and even who is doing the follow-up. All of this data lives on the custom object we build for clients for their funnel/lifecycle reporting and makes it SO much easier for AI to analyze.

There you go, friends. 20 minutes and 5 projects with A TON of value on our roadmap and on the agenda for their next client call with the GTM consultant.

So, do you have a problem or pain point? Yo, I’ll solve it… (Sorry! had to) But really, shoot me a note, and I would love to advise on the how!

Latest from LinkedIn

Join the Conversation

What We Are Cookin' Up (for real)

I have been totally in a dinner rut lately. Most of my meals, I am just trying to think about what my kids will eat and not fuss over, and I can get done in a short amount of time.

And honestly, dinner time has been the WORST lately. IDK what it is, but my kids go mental, have tantrums, push food away, or just won’t sit in their seats. It’s usually some of those things, but now it’s all of it, and my 3-year-old likes to just let it rip during that time. 🫠 I was even commiserating about this with a client who has young kids as well this week. I know I am definitely not alone in this fight.

Anyway - because of this food rut, this past weekend I made a point to try to make a “grown-up” feeling meal for my sanity. And ever since I had seen NYT Food talk about cold noodle salads, like this one and this one, I had to make my own version to go alongside some sesame honey roasted sockeye salmon I made.

Here is my own “sesame noodle salad” recipe:

Step 1: I cooked 1 packet of gluten-free soba noodles and rinsed them, and let them cool.

Step 2: I roasted some broccoli and then chopped veggies like sugar snap peas, cucumber and fennel, and then shaved carrot strips and put them into a bowl.

Step 3: I made the dressing (with my daughter Lily). I grated a small knob of ginger and some jalapeño, then added it to 1 tbsp tahini, 1 tsp peanut butter, 1 tsp toasted sesame oil, some rice wine vinegar, 2 tsp maple syrup, and some water, and stirred in 1 ice cube (which helps the tahini emulsify).

Step 4: Mixed the cold noodles with the salad dressing and veggies, added a pinch of salt and any extra vinegar if needed. Sprinkled with toasted sesame seeds. You can add fresh herbs or even chopped peanuts!

This keeps in the fridge really nicely and gets better with some time in the fridge!

**For the kids, I saved some undressed noodles and roasted broccoli and served those plain on the side. If you also want to appease some picky palates ; )

Links We’re Liking 👍🏻 on LinkedIn

Charlie has joined the X bandwagon, driving through B2B town.

Clay launched Account Agents this week (stoked for this!)

All of us loved having VLOOKUPs in our list of tricks, right?

I loved this advice from John Cena on what traits can lead to failure.

Can’t validate if the skills are good, but love these AI use cases.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

But also over 12 weeks during Summer, they are rolling out 12 new features during their “Summer of Shipping” campaign.

And so far, the ones they have rolled out have been pretty useful and I have been loving the marketing behind them too.

Here is a taste of some so far. Click on the image to go to the LinkedIn post to read more.

1. Clay for Reps. How many times did you wish you had something like this in place for your sales team?

2. Open-weight models for Claygents. Not all operations need the same PHD level brain (and cost) applied to it!

3. Clay CLI & API. An efficient and more cost-effective way to leverage Clay data and actions within your terminal. It can only build workflows now, but it is looking like workflows will maybe be replacing workbooks anyway?

And if you are interested in getting the full rundown, Clay's CEO, Kareem Amin, will be demoing all of these features at this live event next Tuesday!

Also, if you missed it, last week we announced that CS2 is now a Studio Level partner of Clay!

We talked about this in our last podcast episode, but it’s been a really cool evolution from how we started using Clay for our clients to now working closely with the Clay team to deliver expert-level support and strategy to our clients.

And it truly has been a partnership. Just look at my Linkedin post below and you will see why!

Latest from LinkedIn

Join the Conversation

What We are Cookin' Up (for real)

Some of you may be avoiding fresh food right now due to cyclospora. I get it.
I am still eating summer’s bounty from my garden and the farmers market because I love tomatoes, cucumbers, peppers etc.

And this past weekend I also picked up some yummy fresh hummus, baba, and fresh beet salad from a local vendor that has made creating a cute greek-ish lunch or dinner so easy.

I also finally put some halloumi in my fridge to use and forgot how good pan-fried it is.

And I actually got the idea to get halloumi cheese after seeing this Taverna Salad recipe that Claire shared in the foodie channel a few weeks ago.

Looks like an incredible vegetarian main or side dish for summer, am I right?

Links We’re Liking 👍🏻 on LinkedIn

This is an epic car work setup but I do love having my big screen.

Wise words about showing humility in the workplace by ex-CEO of WD-40.

5 work-from-home lunch ideas. (They actually seem more like dinner effort to me though lol)

Don’t let AI make your marketing job less joyful.

Zapier is hosting a practical webinar about using AI in GTM is next week!

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

Running an ops agency alone has taught me that standardization in best practices and advice is key, but the ability to build custom solutions based on the people, processes, tech, and strategy involved is what people will really pay for. And want!

Looking back, you’ll see that some companies’ success was tied purely to this idea of customization.

Take Salesforce: a SaaS tool with an open API that encourages customization on everything and integrates with just about anything. Companies and users were ecstatic.

Even marketing automation platforms like Marketo were able to beat out competitors purely based on their flexibility for customization (albeit also a downside for marketers who found it overwhelming!).

Where we saw buyers become let down is when they realized the tool, and what it provided, was so constrained that they started questioning its actual benefits. Tools like 6sense, which built their models in a black box, did not explain why an account was in a certain stage or became a 6QA, and simply said “trust us,” are now losing clients.

That’s because companies can build their own solutions, leveraging first-party data as well as the intent data that matters most to them, from sources that go beyond the mystery third-party intent “snake oil” they were sold. On top of that, they can actually create a reasoning field to show WHY an account was prioritized.

Looking at this AI boom and what’s getting traction, you’ll see a lot of it comes down to customization. Here are some examples:

1. Clay initially gained popularity for its ability to enrich data well beyond a single provider like ZoomInfo, giving GTM teams the ability to go beyond generic data and leverage Claygents to research data they never knew they could access but needed.

2. n8n garnered attention because it became the most flexible automation platform, able to integrate with basically anything and leverage AI within workflows with ease, all within one shareable platform. This is a great example of buying for infrastructure, but building processes with your own intelligence within it.

3. Claude Code became a sensation overnight by giving non-technical operators the ability to code anything their hearts desired, with no coding experience required. While it first started as a terminal-based tool, now with the addition of shared team instances and Cowork, teams are building solutions and automating work customized to their GTM or individual needs.

4. Just yesterday, Lovable announced it’s in talks to raise another $300 million, which would put its valuation at $13.2 billion! A tool that literally takes verbal direction to vibe-code software is now being used by large enterprises such as Workday and Nvidia, proving that vibe coding isn’t just for small companies looking to build a storefront or a custom website.

In my opinion, any company that wants to survive should always keep this in the back of their mind: “Is what we provide so much a part of their infrastructure that they wouldn’t want to build it themselves? And even if what we deliver isn’t something they’d want to build themselves, can we add in as much possibility for customization as we can?”

Companies that straddle the line of being that infrastructure — allowing for centralized use and governance, but also a lot of customization — will continue to get adoption.

And I see companies trying to do this now: building their MCPs and CLIs, and opening up their APIs. ZoomInfo announced GTM.AI, HubSpot announced its CLI and MCP, Freckle announced its CLI and just today, Clay announced its API and CLI following its MCP announcement about a month ago.

While social media algorithms and even B2C brands have already capitalized on this inherent want for customization, I think more and more technologies will make it part of the fabric of their solution, since buyers will simply expect it.

What do you think?

This Week on the Cooking Up GTM Podcast

We’re back from our short vacation hiatus with a fun announcement! We’re now Studio Level Clay Partners! Listen to us discuss our Clay partnership and how it validates our internal focus on the platform and the great work we have been doing for clients.

Then we dive into two topics we’ve touched on before to see how our opinions have changed. First, when should you build v. buy in the AI era? Second, how has AI changed the job market?

ps. We’re planning on bringing on some guests soon, so if you or someone you know might be interested, we’d love to hear from you!

Watch Now

What We are Cookin' Up (for real)

This is going to sound sacrilegious, being 75% Italian (my Mom immigrated here from Sicily when she was 16!), but one of my favorite ways to eat pasta is pasta salad.

Yup. Cold pasta might just be my favorite. Unless it’s pastina or tortellini in brodo. Both are elite ways of eating pasta.

And recently my go-to side dish for BBQs has been a pasta salad. And it may seem like I am a Goodles ambassador at this point, but I have been using their pasta to boost it with protein.

I am kicking myself because I didn’t take a picture of any of them. But my go-to has been a “Caprese-style Pasta Salad” consisting of:

- 1 box cooked Goodles
- chopped cherry tomatoes
- chopped nectarine or peach
- cubed fresh mozzarella
- sliced Castelvetrano olives
- fresh arugula
- fresh basil
- And dressed with aged balsamic vinegar, good olive oil, red wine vinegar, and salt and pepper

But pasta salads can take SO many different forms, and this week I was drooling when NYT Food sent out a list of great pasta salad recipes.

Links We’re Liking 👍🏻 on LinkedIn

Great simple Clay use case for when your champion leaves the company.

A B2B marketer “getting robbed” lol

Enjoyment is also a form of self-care!

I do this on Zoom calls and at my kid’s school events. ugh.

Love this argument for attitude versus experience in hiring. But I also think attitude can drive experience!

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

We also experienced yet another heat wave we brought to the UK. Third year in a row! But this year it was humid and hot without A/C or any ice in sight.

One of the most interesting things that came up as we had various chats to our friends, people at hotels, and family and friends Charlie’s Mum and her partner.

We realized how AI was still such a hot topic on everyone’s mind.

And the funniest part is that in many of these places, especially Cornwall, we barely could get reliable WIFI or cell reception, and there was no Uber in sight most of the time.

The opinions and familiarity with AI were quite different across the board. Sometimes due to the industry they work in. Someone even noted maybe it was generational and that the younger generations, being more technologically savvy, are embracing it much more.

But whether it was a restaurant owner, a nurse, a therapist, or even a B2B tech agency employee, they all had AI on the mind and their own applications they had been testing.

One gal, who worked for a UK-based PR and Comms agency, even noted that it was a mandated requirement that they have to use AI and even noted that sometimes they were putting our content for a client that they wished they hadn’t used AI for or saw the client use AI for but had to deploy it anyway.

After being away, I did also have some reflection on just how we are inundated with information. Not just AI-related. But everything!! LinkedIn, Instagram, or YouTube influencers are constantly telling us what we SHOULD be doing or not doing or what we should buy or what we should not. It’s relentless.

And it’s no wonder many folks feel a bit paralyzed by it all.

I also fear that with people relying on OTHERS to tell them what to do, they will lose the ability to discern and make their own informed decisions without trying to get an outside opinion or approval first.

But this is dangerous when thinking about where to apply AI. Especially in GTM. Because so many of us are just figuring it out.

And every week there is a new model, a new tool, a new security or compliance risk, a new pricing structure etc. etc.

So it would be very normal to feel a bit lost on where to start.

This is why we have been leading AI workshops with clients to really focus on their pain points and then how we can apply AI whether in a workflow, an agent, a new tool, or even providing a skill.

And I suggest doing the same. Take a part of the funnel or even a specific pain point and see how you can maybe work on just that specific process. Then keep adding on.

Because I know from experience, while in the attempt to do too many things. You very rarely get them done fast or in the most effective manner.

And if you are looking for some use cases to start with, I used AI (hehe) to analyze all our client AI roadmaps and completed projects to come up with this list of our most popular use cases (right now) at CS2 by category:

*Note some of these reference specific tools for example purposes but could be handled in different ways.

Enrichment & data quality

• Clay enrichment / waterfall (email, phone, firmographics, tech stack) written back to CRM

• AI title/function/level normalization to expand MQL eligibility (via Clay or native HubSpot AI Data Agents)

• AI junk/spam scoring on inbound forms to block bad MQLs

• Account research & classification (Claygents replacing manual research)

Scoring & prioritization

• AI lead scoring / lead rating (LLM infers ICP fit vs. rigid rules)

• Buying-signal & intent research scored and mapped to sales plays

• Email-send prioritization and suppression recommendations

Outbound & campaigns

• Personalized outbound pilots (accounts → buying group → AI-drafted email → sequencer)

• Buying-committee/persona rollups on accounts

• Marketo tokenized templates & AI-assisted email builds

Deal & revenue intelligence

• Deal-risk analysis on open opps from Gong transcripts (weekly digests)

• Closed-lost analysis comparing rep-stated vs. AI-inferred loss reasons

• AI-assisted revenue models and funnel-vs-plan dashboards

CRM / Martech ops acceleration

• Salesforce/Marketo/HubSpot build acceleration (flows, Apex, field mapping, data adjustments, backfills)

• Lead-source automation via single custom-code block (replacing dozens of if/then branches)

• Field-level security and picklist automation

• SFDC architecture audits and sync-error root-cause analysis

• Flow/process diagramming via Whimsical MCP

Packaged skills & tooling

• List-import skill (clean, dedupe, enrich, QA, auto-build workflows)

• Address-parsing / list-QA skills

• Technical-documentation generators (from transcripts/Loom/help docs)

• MCP servers for marketing platforms (HubSpot/Marketo)

One-off analysis

• Ad-hoc data analysis (bounce-keyword mining, contact reparenting, UTM effectiveness)

Emerging / not yet deployed*

• Inbox-monitoring agent (unsubs, job changes, positive replies)

• Predictive audiences (likelihood to register/attend/unsubscribe)

• AI as CRM data validator

Hopefully that inspires you a bit for your own roadmap. And if you have any questions, as those were all very abbreviated descriptions, just let me know!

ps. If you are in the USA, Happy 4th of July weekend!

Latest from Team CS2 on LinkedIn

Join the Conversation

What We are Cookin' Up (for real)

I have talked about this before, but I absolutely hate food waste. It’s one of my biggest pet peeves.

I am known to always take the doggy bag, to save the bones for stock, to freeze anything I can to ensure it doesn’t go bad etc. etc.

And I also am a fan of using things that normally get trashed as an ingredient. And one of my most used ingredients like this is pickle juice!

I use pickle juice as the thing it best resembles: a vinegar.

I use it to marinate meat. I use it to add acid to soups or stews, and my most used application is using it to dress salads.

Oftentimes, I will just use the pickle juice itself (preferably Bread and Butter pickles these days) to dress a “fridge salad” without having to add any oil at all. This lightly pickles the sturdy veggies/beans in a nice way.

A few weeks back, Christie was actually making a salad and noticed she'd forgotten the dressing, so I told her to look for pickle juice, and she did!

This week, I spruced up the pickle juice dressing with sesame oil, a little bit of rice vinegar, and a dollop of peanut butter to make a yummy peanutty dressing for my napa cabbage and chicken salad. It was so good!

So next time you are about to throw out that pickle juice, think about me and save it and use it!

Links We’re Liking 👍🏻 on LinkedIn

Every time I accept the LinkedIn request because I thought it was a GTM Operator…

This is painfully accurate about Americans who just went to Europe. lol!

HubSpot’s quiet announcement that was maybe quiet on purpose?

Love this framework on creating high agency in employees.

Great thought on what it takes to align a leadership team.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

This week’s newsletter is going to be pretty short and sweet. It’s been a very busy week personally for more than a few reasons, but the main one is that next Wednesday, both Charlie and I fly to England for his Mom’s wedding! So it’s been all about prepping for that and ensuring we have done everything we needed to before signing off.

And on that note, there will be no newsletter for two weeks. I know you will miss it but don’t worry, I will be back before you know it.

Anyway, this week I wanted to talk about a few things that have caught my eye or been on my mind in GTM AI, including an exciting announcement from us at CS2.

1. Using Slack Salesforce Channels (with Agentforce) for Deal Rooms

Last week, I was chatting with a new parent friend from my daughter’s school, and he is a sales manager at a well-known private company valued at $1bn. He mentioned his Slack briefly and showed me his app on his phone, and I nearly got sick.

I can barely handle it when my Slack messages go above 10 unread. It was lit up more than Rockefeller Center at Christmas. I asked him, “Do you all create a Slack group/channel for every deal your team has?” And not shockingly, he said yes, and I am on every single one.

Prior to this conversation, I had always thought that this practice was somewhat harmless, and maybe even effective since it creates a “deal room” where the sellers and their manager “live.” But after seeing (and hearing) how he can barely keep up with all the messages, I started to think that this whole idea of Slack being the AE’s central command center is maybe just not the way.

Salesforce says it’s the way (selfishly). And some people even commented on my post this week about the topic and said it can be the way. But like another person said, it seems like it will make managers become notification managers instead of sales managers, and they will probably barely keep up with those notifications at that.

I do think there can be a lot of value in using Slack Salesforce Channels, which connect directly to SFDC and even let you invite agents via Agentforce into the channels. So I haven’t completely written this off and definitely want us to explore the best ways to do this with clients.

2. Using Claude Code to Assist with Editing SFDC Flows

This is something that Alison had talked about with the team a few weeks back and just posted about on LinkedIn this week. She is finding a lot of value in using Claude Code to edit SFDC flows by combining it with SFDC’s flow version comparison tool to help validate the changes it makes. Alison noted that she still finds some mistakes, but they are often ones that could have been easily avoided with a refined prompt or clearer instructions.

Check out Alison’s post if you want to try it yourself!

3. The Case for Still Having Sales Reps Capture Data (Like Updating MEDDPICC Fields)

Shoutout to newsletter reader Greg Travers, who shared with me this episode of The RevOps Review podcast with Nikko Georgantonis, Head of Revenue Systems at Hightouch.

In it, they discuss how some sales leaders actually still want their reps to input data like MEDDPICC so they don’t get lazy, stop asking the right questions, or stop thinking critically about the prospect’s responses and their deal strategy. At Hightouch, they use Glean to create a MEDDPICC summary alongside the reps, so instead of doing all the work for them, it helps validate what the reps documented and suggests changes for accuracy.

I am always thinking about how to automate tasks like this for sales, so it was an interesting perspective to hear.

4. We Are Hiring!

Yes, super excited to announce we are hiring for another Director/Sr. Director of GTM Ops here at CS2! This is our “client lead” role, where you consult and work directly with clients to roadmap, strategize, execute work, and coordinate projects across our supporting practices on the account. What makes this GTM AI related is that we are also adding a new skill requirement we haven’t included before: someone who has already been using AI in work they have deployed for their company or clients.

If you are or know someone who fits the bill and wants to join an awesome company working alongside some great minds in GTM Ops fully remote, check out the job posting here. Please note this role is only open to US based residents.

Alright folks, that is it. I will see you in your inboxes in a few weeks. Don’t miss me too much. ✌🏻

This Week on the Cooking Up GTM Pod

This week, we discuss something that’s been on everyone’s mind lately: AI token and credit spend. How do you choose the right tool and model for the job? Do you *really* need AI for that deterministic workflow?

Then we test HubSpot’s AI SDR Avatar (powered by 1Mind) and see how things have changed since the last time we reviewed an AI Avatar. Lastly, we plug a great job opportunity… here at CS2!

Listen or Watch Here

What We are Cookin' Up (for real)

There are a few ingredients that I feel like have so much versatility that you are crazy not to have them on hand.

Eggs. Cottage cheese. Yogurt. Beans. Oats. Rice. Onions and carrots.

And another one is bananas.

Even a banana gone bad is so valuable.

You can freeze them for smoothies. You can blend them into “nice cream.” Or you can make banana bread!

But this weekend I found another great use. Banana bread granola!

One of my good friends/neighbors just had their third baby, and I didn’t have enough time to prep a whole meal for them, so I made her a batch of granola to snack on. Oats and omega-3 fatty acids are great for milk production, so it was a no-brainer.

I saw a recipe on Instagram, and I didn’t want to comment since I don’t use the Instagram app, so I went and found it for free on this post here.

The only change was that I cooked mine at a lower temperature (around 275 degrees) once I realized the mixture was still a bit wet and cooked it for about an hour. I actually prefer slow-cooked granola to make it ultra-crunchy. Just be sure not to break it apart until after it’s cooled, so you can get those crunchy chunky pieces!

I have to hide this bag from myself after scooping out some on my side of morning yogurt because it’s so yummy and easy to eat the whole bag.

Links We’re Liking 👍🏻 on LinkedIn

For as long as I can remember, I have always been a morning person.

If you are a start-up looking for a fractional senior marketer - Mark Huber can help..

Marketo MCP capabilities + Marketo API capabilities in 1 custom MCP.

Companies have gone from AI-pilled to AI poor…

There are 50 grams of protein in…

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

This is something we have been thinking about and talking about as a team when it comes to getting more AI projects on our own roadmaps, and as we’ve been having those conversations, it made me realize a few ways operators can combat this. But the most important way to tackle this is to identify pains and needs, and not AI use cases.

You see, many folks on your ops team or the broader GTM team are likely still thinking about doing things the way they always have. Or they’ll come in with a request and not the reason why. But uncovering the true pain and needs of the GTM team, and then challenging yourself to tackle that pain with something beyond the status quo, is what will hopefully drive you to solve something with AI.

As much as we like to always be understanding the pains and needs of the business, and I know this is something we focus on with our own roadmapping with clients, sometimes it can fall to the wayside when the GTM is moving fast and just piling on requests and projects.

So here is some advice on how to incorporate more AI projects that benefit the GTM team and not just personal efficiency into your (and/or your team’s) workflow:

1. If you are not overloaded with requests but unsure what AI projects to work on, go find problems.

A few of the areas most ripe for AI in GTM are the sales process and prioritization, outbound and demand gen campaign execution and engineering, and analytics. Go meet with your SDR leader, the folks working on campaigns or outbound, and your marketing team who are likely looking for more insights in their data. And once you figure out what those pain points are and start to brainstorm solutions, think outside the norm and really consider how they can be solved with an AI solution.

2. If you are being sent tasks from the GTM team, instead of quickly accepting the project or planning to execute it the way you “normally would,” go into exploration mode.

Ask what the goals are. Ask what pains and needs the stakeholder or team has, and see if you uncover a bigger project and one that can be best executed now with AI, whether that means including AI in the workflow, building an agent, or leveraging an AI-native tool or technology.

Here is an example of this in practice:

You have a meeting with your SDR leader who says they’re struggling to get their SDRs to input the qualification criteria they’re receiving from meetings with prospects. AEs are showing up to demos uninformed, making the prospect repeat themselves and leaving them frustrated with a bad experience.

The SDR leader’s proposed solution: put in a validation rule requiring all qualification criteria fields to be completed on the Contact before it can be transferred to the AE once the meeting is set and logged in the CRM.

Now, this could “work” but it just creates more friction for the SDR, who then has to try to remember what was said in order to fill it out.

A better way to think about this: how can you take the work off the SDR and make it more effective for both the SDR and the AE?

Most orgs already have a call and/or meeting recording tool that delivers summaries which are “helpful” when sent over email, but that doesn’t solve the actual problem of extracting the data and putting it where you need it.

So instead, you can identify a project that leverages AI in various ways to help. For example, you could use an automation tool like n8n or Clay to extract call recordings for SDRs, listen for key mentions or insights, and then update individual fields or one consolidated field with that qualification criteria. That data then gets mapped to the meeting or opportunity record for the sales rep to reference, and it automatically sends the AE a Slack alert with a link to the opportunity, a summary of the qualification criteria, and anything the prospect mentioned they’d want to focus on during their tailored demo. Just make sure you train your SDRs to actually ask for this on the call!

You can also leverage tools within your CRM like Einstein Conversation Insights for Salesforce or HubSpot Conversation Intelligence to take this even further and apply call insights not just to the SDR to AE handoff but throughout the entire buyer journey with your team.

The result? SDRs happy they have less admin to do. AEs happy they can show up informed and ready to delight the prospect on the demo call. SDR leadership reassured that AEs will convert the meetings they’re generating into opportunities.

All without using a “gate” like validation rules or requiring manual input. Win. Win. Win.

Hopefully this inspires you to put this into practice and let me know if it helps you get more GTM AI projects onto your roadmap!

This Week on the Cooking Up GTM Pod

We have another guest this week! You probably already know him - and love him. Our GTM Engineering Lead, Xander Broeffle, joins us for a conversation where we tackle a new tool by Zoominfo, what is working for us in Clay, and our thoughts on Marketo catching up on the AI front.

Listen or Watch Now

ps. We loved having Alison and Xander on so much that they will likely have regular appearances along with some other folks here at CS2!

What We are Cookin' Up (for real)

A while back, I shared how I was powering up my kids’ Goodles Mac & Cheese with Cottage Cheese and many of you clicked into that Goodles link so I know it resonated.

Well, today I will tell you about the hyperfixation dinner I am currently making weekly.

Protein-Packed Shrimp Pasta Primavera

Goodles also makes some great dried pasta, and I love both the loopdy-loops and the pipe rigate (aka curveballs).

So to make both me and my kids happy, I have been making the plain pasta which I power up with some blended cottage cheese and milk sauce separately. I serve them their version “plain” with shredded cheese on top, and then serve some of the cooked shrimp and veggies separately. Kids hate things “mixed together.” IYKYK.

Then, in another pan, I make a sauce of sauteed onion, veggies like peas, asparagus, zucchini, broccoli or carrots and unthawed frozen deveined shrimp cooked in fennel seed, lemon juice, and garlic powder. I cook this down in the pasta water until soft, and then add in the rest of the pasta that has the cottage cheese sauce.

The water from the pasta, the cottage cheese, and milk solution make this sauce glossy with some time under low heat.

I finish things off lately with a sprinkle of fresh basil from my garden.

It is so good and packed with so much protein. From the sauce to the shrimp to the pasta! So you can feel full and not have a huge spike in your blood sugar.

Links We’re Liking 👍🏻 on LinkedIn

I love all these learnings from founders / AI experts Mark Roberge shared.

How Canva aligned its GTM team with Winning By Design’s SPICED framework.

4 Claygent ideas from Sara McNamara + see the last slide for new features for Claygents!

Dreamdata announced its “Analytics Hub!”

Such a brave post by Kathleen Booth and another reminder to be your authentic self!

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

So I decided to map out the skills (especially the not-obvious ones) that I believe anyone working in GTM AI, especially operators, needs to master to be successful, and share them with folks and our team as a guide for where to focus.

It is a lot. And some of these things have “best practices” that aren’t even fully established yet, or will evolve as AI evolves, but it’s still worth investing the time to learn what should be done now and how to implement at the orgs we support.

1. Data Architecture and Hygiene Expertise

Most good operators will have a strong skillset in data architecture, but you’d be surprised how often I meet folks who don’t. Data architecture and hygiene are proving to be the most important prerequisite for AI effectiveness. However, it’s an area many GTM teams have struggled to invest time and money into unless it was for contact acquisition/enrichment. Yes, data across leads, contacts, accounts, and opportunities will remain the most important and keeping those objects clean, duplicate-free, and enriched with segmentation data is high priority. But other data also becomes important, like structured funnel data and clean campaign structure and tracking, so you can leverage AI to draw the right insights or make the right decisions.

2. Product Marketer, Enabler, Manager, and Engineer Skills

Now that we can build sophisticated solutions with AI, many teams have decided to replace tech with homegrown tools or solutions. One skill here is actually knowing where you SHOULD build versus buy. But if you do build, you need to know you’re essentially building a product for your org and you need to treat it like one. This means you need to effectively market and enable the GTM team on it, manage and innovate it, and maintain the bandwidth to build those innovations. You also need to be on the hook for any patches, fixes, or adaptations if new models cause issues.

3. Selecting the Right Tool/Model for the Task

Before AI, the biggest question for operators around the right “tool” was what platform or technology they were going to use to build something. But now, with AI’s ability to function either as part of a workflow or as an autonomous agent, operators are being pushed to decide when each is most appropriate. I saw this post last week from the team at Anthropic and shared it with our team and I thought it gave great advice.

On top of that, operators will also need to understand what model should be used for the task (unless Claude/ChatGPT itself is trained to determine this). This post by Francis Brero was interesting on this topic. TL;DR - you shouldn’t be paying for a premium model when the task doesn’t call for it.

4. Budgeting and Controlling Spend

Woowee, is this a hot topic in GTM AI. I went to two events in the last two weeks and it came up a lot as a concern from GTM teams who have found that in this era of experimentation and decentralization of AI, they’ve burned through a lot of credit usage and spent more than originally budgeted. Now CMOs are pulling back the reins and even moving toward centralization until they can get a process around budgeting and budget adherence in these tools. I think this is best suited for a direct owner or set of owners, depending on the teams using AI and where they get their budget from.

5. Governance and Risk Mitigation

Another very hot topic. I have seen some SFDC instances already taken down by one rogue person who got connected to production. I have also seen clients hit their API limits. And I know for a fact that companies are allowing read and write privileges to their most sensitive production data through Claude Code or even connected platforms that are building agents for them. I do support the enthusiasm of teams wanting to leverage AI, but it sure does need governance and risk management. Structure (and potentially a set of owners focused on this) is crucial. And it’s an area we’re still firming up our best practices on here at CS2, because as you can imagine, it is very nuanced.

6. API Integration Management

Before AI, most operators were solely concerned with monitoring the integrations between tools like their CRM, marketing automation platform, sales engagement tool, BI tool, ABM platform, etc. Operators figured out the best way to handle this was with dedicated integration users and permission sets. But now, API integration also means being accessed by local terminals through Claude Code, for example, and from various systems you may never have thought about giving API access to. Operators will need to be mindful of putting in API limits, cracking down on who has access to API endpoints, what data is being accessed or updated, and what tools are actually executing that.

7. AI Conductor Skills

The future of GTM AI will be leveraging autonomous agents, and this doesn’t mean humans won’t be involved. For the immediate future, our foundation layer (i.e., CRM, marketing automation) will likely stay the same, but AI agents will be used for intelligence: deciding what workflows to run, what data to analyze, and executing across multiple systems or with other agents. In the future, I see more and more of the human work being about devising what agents should be built, building and training them, and monitoring them. Autonomous agents should be improving over time, but I think they will still very much need oversight (like all workers! lol).

8. Self-Learning, Training, and Enablement

I cannot stress enough how this is a time where EVERYONE needs to be open to learning at every chance they get. Being at an agency, we have the added pressure to become experts, but whether you’re in-house, agency, or contractor, this will mean getting your hands dirty, learning by doing, and proactively seeking out knowledge, best practices, and experimentation at any free moment you get. Maybe this means creating a whole upskilling system or goal for you and your team. Maybe it just means having a regular block on your calendar if you’re going it alone. And guess what? Sharing is caring. I believe a great system across a GTM team is one where everyone can learn from each other and I highly encourage teams to “own” specific domains or initiatives and share knowledge with the wider team. You can even look at creating a whole “center of excellence” around this, similar to what some orgs did around the rise of marketing automation and MarTech in the 2000s. (I was part of one at Marketo.)

So I could go on and on, but my week has been crunched, so this was just a brief overview. I’ll likely keep sharing our own learnings about the skills that fall into these categories in future issues. But I’m curious, what “new” skills have you been honing with the rise of AI?

This Week on the Cooking Up GTM Pod

We have a special guest today! CS2’s Chief Solutions Architect and Analyst Alison Rouse joins today for a fun discussion, including a chat about how she leveraged AI for business planning for a client and how you can too.

We also dive into a few trending social media posts with some hot takes that we dissect.

Watch it Now

What We are Cookin' Up (for real)

My daughter would hate me repeating this story, but she is 6 and can’t read this, so I am going to anyway.

A few weeks ago, we were with her “Uncle,” and she was finishing eating a lemon lavender scone we picked up from Voyager Coffee.

When she got near the end, he asked her for a bite, and she stopped and instead of giving him the last bite. She popped it into her mouth instead.

She realized how cruel/rude that is, so we decided that the next time we see him, we will give him lemon scones that we make ourselves.

So this past Saturday, I found this awesome lemon scone recipe while Ava picked a few huge Meyer lemons from our tree and we got to work.

The thing I liked most about the scone recipe was how easy it was. If you have a mixer, you are literally looking at just 4 steps of work. And they came out sooo tender and moist. (Apologies for using the m word.)

Instead of the icing the recipe used, we made our own by blending up blueberries with lemon juice and powdered sugar. It was a bit thinner, but it also had a lot less powdered sugar.

I highly recommend bringing these warm out of the oven to a friend. Like we did ; 0

Links We’re Liking 👍🏻 on LinkedIn

A recap of my “Maycember”

This post by Charlie, “RIP Salesforce API Limits,” really struck a nerve with folks.

19-year-old founder video cracked me upppppp.

My mom also just learned what G-drive was a few weeks ago...

ZoomInfo announces Nooks integration.

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM

Partly driven by Track Star* on YouTube, a music show where the host plays a track, and the guest has to name the artist and talk about them a little bit. Start with Cillian Murphy’s ep.

But also, YouTube allows you to stumble across random videos like this one on the BBC archive. 2000: Are Internet Mobiles Really the Future?

Before you start laughing at how the title was even questioning the mobile internet and the use of the term “internet mobiles,” give it a watch if you have time.

What struck me was how most people, even the smartest people/analysts, had no bloody clue about what would happen in 2, 5, 10, 20 years.

There was also a lot of skepticism and 🤷‍♂️IDK-ing

Some of the major points behind the skepticism and lack of clarity were:

1. Technical limitations that will hold things back. The screens would never be big enough.

2. No revenue/profitable business model. Internet data was 1% of revenue back then. So people wrote it off as insignificant.

3. Too technical. Some people can’t even use VCRs, how could they then use the internet on a phone?

4. The public doesn’t even want it. Why watch a movie on a small screen when you have a big TV?

As I was watching this, I couldn’t help thinking about some of the AI skeptic talking points over the last 12 months.

1. Technical limitations that will hold things back. AI is just autocomplete on steroids. AI hallucinates and you can’t fix it. There isn’t enough power.

2. No revenue/profitable business model. All the AI companies are bleeding money. We’re in an AI bubble. The only reason AI revenues are up is that people are tokenmaxxing.

3. Too technical. It’s too hard to integrate AI into real production systems. All the agent harnesses are way too risky for non-technical people. Claude Code is for developers.

4. The public doesn’t even want it. Everyone hates AI right now. It’s going to take all the jobs and crash the economy. Students are revolting at graduations.

Now, I expect that most of you reading this are in tech and GTM, so I doubt you think AI is a fad. You’re probably anticipating that AI is going to change a lot of things.

But everyone should think about whether they see themselves in the people in the video. Maybe being a bit too skeptical and unsure.

Even if you’re bought into AI, have you genuinely modified your daily habits, routines, and priorities based on this yet?

Now, before I go further, I have to say that I am very much aware that AI is going to have some downsides. And people have mixed feelings on AI. My feeling/hope right now (that I reserve the right to change at any time) is that humans will always be needed (especially us in GTM and Ops) and things will work out.

I feel it is quite likely that the best GTM operators that harness AI will be even more valuable than ever. Especially as GTM Ops takes over even more of GTM, see this post and image:

But if the video from 2000 convinced me of anything, it’s that we’re going to look back at AI in the mid-2020s like we look back at the early 2000s. Things we can’t imagine now are going to feel so obvious and quite normal in the 2030s. And it’s going to be a world we can’t picture right now.

Even the most bullish people on mobile internet in 2000 didn’t anticipate: Uber, Waymo, Apple Pay, WhatsApp, TikTok, Instagram, podcasts, food delivery, mobile video calling, etc. The world changed more than anyone thought.

The video reinforced my belief that our job right now is to avoid being like anyone in the 2000s who read the skepticism and waited too long to adapt.

I still see some skepticism out there, but more so, people who are not sure where to start, or are waiting for someone else to show them the way.

So to make this practical, here are 10 ways me, Crissy and our team are challenging ourselves to stay ahead:

1. Crissy and I went back to building. As the owners of CS2, we could easily step out of the client work and delegate everything. But there is no replacement for actually having to build AI systems with your own hands. If you’re a leader, I’d recommend you carve out some of your time to be an IC again and build stuff.

2. We found a few AI-centric podcasts and newsletters to immerse us in what other people are doing and the art of the possible

   1. A selection that I enjoy: 20VC, AI Daily Brief, In the Weeds, Cannonball, AI Builders, The Signal, Also this newsletter :)

3. We set up weekly round robin AI calls between people on the team and me to learn what they are doing with AI.

4. We have a weekly learning session with the whole team to walk through AI use cases.

5. We’re consuming content on LinkedIn more than we were before. Love it or hate it. There are some really great people posting some interesting stuff.

6. We’re going to more events, both in person and virtual.

7. Crissy and I are building skills and agents for our team internally. Nothing beats having to actually ship a production-level agent to the team. Again, if you’re in leadership, don’t feel like this is below you.

8. We’ve aligned on KPIs around AI adoption. One goal we have is to always be working on an AI project for 100% of our clients as part of their overall roadmaps.

9. We launched our GTM engineering team to go deep on AI. We also have our other teams all discovering and sharing use cases to find ways to build quicker with AI and deploy AI for clients. It’s a top-down and bottom-up approach.

10. We have a cross-functional AI tiger team meeting weekly to optimize all of the above.

TLDR:

1. Spend time consuming AI content to see what is possible. Even 2 hours a week can make a massive difference.

2. Take the initiative and build production-level AI tools, AI workflows, and AI skills with your own hands. Note: production level, not a demo. Don’t wait to be asked.

3. Document at least one AI-related KPI for your team that is measurable and creates an incentive.

For anyone practicing all this consistently, I am hopeful the next era we’re entering has a lot more opportunities than risks. And it might be quite fun.

This Week on the Cooking Up GTM Pod

If there is anything that going to two events in under two weeks has taught me is that the content that people really need is some REAL examples of what they should be doing to prep for AI and with AI in GTM.

This week’s podcast, we are telling you the data you need to focus on for AI. Then, we cover 6! examples of how CS2 is using AI for client work and also to help us as an agency (but still apply to GTM teams.)

Check it Out!

What We are Cookin' Up (for real)

One thing that I will never do is throw out a banana.

Well, unless someone accidentally let it go rotten in a hot car or bottom of a backpack.

Because once a banana has gotten too ripe for raw consumption, they either make it into a baked good or they go into my freezer for future smoothies.

This week, I wanted to make a little snack for us all to share throughout the week. Whether it’s my daughter’s lunch or after lunch for Charlie and I.

I found this healthy banana oat cookie recipe and was sold. For the chocolate chips, I used my favorite Hu Kitchen chocolate chips, which I shared in two newsletters ago.

For those of you who are more visual learners, you can watch the video of the cookies being made here!

They make a nice little lunch dessert or afternoon snack if you are into that.

Links We’re Liking 👍🏻 on LinkedIn

This post had me at “aux-cord privileges.”

I don't fully agree, but this GTM Eng then vs. now is kinda funny.

Intelligent people do these things. Always be willing to change your mind!

I have to admit it. I do not like their ads/billboards…

Not GTM or work-related - but this was so heart-warming!

🙋‍♀️ P.S. Is there a topic in GTM Ops or AI that is top of mind for you that you would want me to cover in a future issue? Reply to this and let me know!

Share Cooking Up GTM