What 200 GTM operators actually do with Claude
Maja Voje and Kyle Poyar asked. The winners build GTM engines. The specific use cases, schedules, and triggers to copy this week.
Applied AI: Build The Engine - Friday July 10, 2026
For a year the pitch was to prompt Claude better. The operators pulling ahead stopped prompting and started building systems that run their GTM work while they sleep.
12 signals from this week’s Claude Signal, across 5 tracked creators. 3 problems. 1 playbook to steal. Read time ~7 min.
Routine = a Claude agent that runs on a schedule or a trigger, in the cloud, with nothing open on your laptop.
Agent = software that reads your context and takes steps on its own, instead of a prompt you fire each time.
Save this. 3 AI moves from this week’s Claude Signal, each with my operator take and a move a lean European team can run today, no ops hire required.
Send it to 1 founder or GTM operator scaling toward 10M in Europe who reads everything and ships none of it.
The 90-second version
The operators pulling ahead build GTM engines, not prompts. Maja Voje and Kyle Poyar surveyed 200 of them: 67% built workflows that were previously impossible, only 27% replaced a tool.
The smallest version is a Routine: an agent that runs your Monday competitor scan on a schedule and drops the 3 that matter into Slack, without you.
Claude will drive your logged-in stack too. Pull every stuck deal, read the notes, and tell you which need a manager this week. A RevOps morning as one instruction.
Agents drift without a shared context system. Put your ICP, positioning, and voice in files every agent reads, so a fix cascades to the whole team.
The European read: a lean team spread across markets has no ops hire to spare, so the win is an agent that covers more countries and more reading than your headcount can, in your voice.
Read time: 7 min
This week’s number 1 move
Put one recurring GTM job on a schedule instead of on your attention.
Maja Voje and Kyle Poyar surveyed 200 GTM operators for their Claude for GTM Pulse Report, and the highest-impact use case was GTM engines: systems where Claude reads your context, connects to your stack, and runs while you sleep. 67% said they had built workflows that were previously impossible, and only 27% had replaced a tool, so this adds capacity to the team rather than swapping a tool. Kieran Flanagan shows the smallest version: a Routine, an agent that runs on a schedule or a webhook (a signal one app sends another when something happens, like a form submission), in the cloud, with nothing open on your laptop. His Monday competitor digest pulls a week of competitor moves, checks them against his positioning, and drops the 3 that matter into Slack, every Monday, without him.
My take: I already run one. Every Friday an agent walks me through the same 5 questions about my operator week and hands me back a clean writeup, which is where this newsletter starts. It runs whether I am in the mood or not, and that is the point. A lean team does not need the full engine this week, it needs one job that stops depending on you remembering to do it.
Do first: pick the one GTM job you repeat every week, competitor scan, inbound triage, or content brief, and write in one line what good output looks like and where it should land, a Slack channel or a Gmail draft. That line is the spec for your first agent.
The rule: if a GTM task happens on a clock or a trigger, it is a Routine, not a thing you remember to do.
Sources: Kieran Flanagan, “The 4 New Claude Code Features for GTM Operators” (May 22) · Maja Voje and Kyle Poyar, “Claude for GTM Pulse Report” (Apr 24)
Hi, it is Koen Stam and welcome to GTMcraft OS: The Future GTM Operator. This newsletter is built from 100,000+ GTM signals collected from 100+ operators and founders, combined with 13+ years of my own lessons and failures from the trenches. I write at the intersection of go-to-market practice and AI-powered systems for founders scaling 0 to 10M+ ARR.
100,000+ GTM relevant signals from LinkedIn, Newsletters and Podcasts indexed. Translated into 100+ GTM plays for you to implement today.
13+ years over 3 GTM operator jobs across 3 GTM motions (SMB, MM, ENT). Scaling from 2-10M+ ARR multiple times. Same recipe. Different motions.
Now all part of the GTMcraft Operator Room.
Problem 1: You still run competitor intel, inbound, and content briefs by hand every week
This is the clearest GTM win in Claude right now, and almost nobody has set it up. Kieran Flanagan’s Routines turn recurring work into agents that run on a schedule, a webhook, or an API call. 3 map onto a lean team: a Monday competitor digest that pulls the last 7 days of competitor posts, product updates, and reviews, checks them against your positioning, and drops the 3 that matter into Slack at 6:30am; an inbound draft agent where a high-fit form submission fires a webhook, Claude reads the company and the role and any prior call, and leaves a first-touch email in your voice as a Gmail draft for a rep to send in 30 seconds; and a Friday content brief that reads your best posts and hands you 5 angles for next week. The survey backs it: 92% save time, 67% built workflows that were previously impossible.
My take: we are wiring exactly this into our Q3 plan, an AI layer under the mid-market pipeline motion, because the work is already repeatable and we are just still the ones running it. On a team with no ops hire, a Routine is the closest thing you get to an extra teammate, and it works across every market at once, not just the one you had time for.
The European read: the competitor scan and inbound triage that a big team gives to an ops person is the exact work a lean multi-market team never gets to. An agent runs it across all your countries on the same schedule, so coverage stops being a function of who had a spare hour.
Do first: list the recurring GTM jobs you do by hand each week and circle the one that is pure pattern, like the competitor scan or inbound triage.
Do this week: build it as a Routine on a schedule, connect Slack or Gmail, run it once, and tighten the prompt. Keep a human approval step on anything that goes to a customer.
Do this month: add a webhook-triggered Routine, like the inbound draft agent that fires off your high-fit lead form.
The rule: if a GTM task happens on a clock or a trigger, it is a Routine, not a thing you remember to do.
You know it worked when: a useful competitor digest or inbound draft shows up in Slack or Gmail without you starting it.
The play: 5 steps to put your GTM busywork on autopilot while you sleep (2026 series). Turn one recurring GTM job into an agent that runs on a schedule, without you. Get full access below ↓
Sources: Kieran Flanagan, “The 4 New Claude Code Features for GTM Operators” (May 22) · Maja Voje and Kyle Poyar, “Claude for GTM Pulse Report” (Apr 24)
Problem 2: Your reps and RevOps burn hours reading the CRM and call recordings
The other Kieran Flanagan play is Claude driving your real, logged-in browser. One instruction: pull every deal in stage 3 for more than 21 days, read the last 5 notes on each, and tell me which have a clear next step, which are waiting on us, and which need a manager this week. That is a RevOps morning as a single prompt. Same for calls: read the last 10 calls we lost on pricing, name the 3 most common objections, and show me where reps stumbled, and you have a coaching brief before the team review. Maja Voje and Kyle Poyar found the winners wire Claude into a stack, with Claude as the intelligence layer reading across the tools, and the reading and summarizing is the part to hand over.
My take: the reading is exactly the low-value time a lean team cannot spare. I would rather my reps spend their hours in the customer conversation and let Claude do the CRM read-through, especially when the calls are in 3 languages and nobody has time to review them all. Hand over the summary, keep the decision. The decision is still yours.
The European read: a multi-market team reads deals and calls across several languages, which is slow and gets skipped. Claude reads all of it and hands you the pattern, so a lean team gets the coaching brief a bigger single-market team would staff a person to produce.
Do first: pick the CRM or call read-through you do most, like the stuck-pipeline review or a lost-deal debrief, and write it out as one prompt.
Do this week: connect Claude to your logged-in CRM or call tool, run that prompt once, and check the output against the source before you trust it.
Do this month: run 2 read-through jobs on a schedule, for example a Monday stuck-deal review and a weekly lost-call debrief, and review them between meetings.
The rule: if the task is reading your own CRM or calls and summarizing, hand it to Claude and keep your time for the decision.
You know it worked when: your Monday pipeline review starts from a Claude-built table, not a manual scroll through the CRM.
The play: 5 steps to turn CRM and call reading into a 12-minute scan (2026 series). Hand the CRM and call read-through to Claude, keep your time for the decision. Get full access below ↓
Sources: Kieran Flanagan, “The 4 New Claude Code Features for GTM Operators” (May 22) · Maja Voje and Kyle Poyar, “Claude for GTM Pulse Report” (Apr 24)
Problem 3: Your GTM agents drift and cannot scale because they have no shared system
Agents are only as good as the context they run on, and this is where most GTM teams stall. Maja Voje and Kyle Poyar found that 93% of Claude power users treat detailed company context as the foundation, and the operators seeing real returns build context systems, a context file plus saved skills plus connectors, rather than better prompts. A context file is a plain document holding your ICP, positioning, and voice that the agent reads every run. A skill is a saved capability you call, so you do not rewrite the instructions each time. Ruben Hassid’s reversal this week is the how for a lean operator: stop running on loose files and folders, which leak old context and rot when nobody maintains them, and build skills you call and projects where one account’s context lives and stays put. Build the system once and every agent inherits it.
My take: this is the difference between 3 clever agents and a team that scales. We are standing up several agents at once, account search, partner intent, an AE layer, and if each runs on its own scattered prompts, they drift the moment I look away. One shared context file, in our voice, is what keeps them saying the same thing as we open the next market.
The European read: when you open a new country, a shared context system means your one positioning, in your voice, is applied on the first pass, instead of every rep re-teaching the agent what you sound like. The context is the thing that travels across markets, so a fix in one file cascades to every country at once.
Do first: write one context file for your product: your ICP, your positioning, your voice, and the 3 things Claude always gets wrong.
Do this week: turn your most-repeated GTM task into a saved skill and your top account into a project, and point your Routines at that context file so every run starts from your standard.
Do this month: give the team one shared set of skills, so a fix cascades instead of everyone rebuilding the same one.
The rule: if your agents run on prompts instead of a context system, they drift. Put the context in files they read every time.
You know it worked when: a new GTM task starts from your context system and sounds like your team on the first pass.
The play: 5 steps to turn your judgment into a context layer Claude reads (2026 series). Put your ICP, positioning, and voice in files every agent reads, so they stop drifting. Get full access below ↓
Sources: Maja Voje and Kyle Poyar, “Claude for GTM Pulse Report” (Apr 24) · Ruben Hassid, “I was wrong about Claude” (Jul 1)
Steal this move: put your first GTM job on a schedule in 4 weeks
Play: How to Become AI-First as Founder or Revenue Leader (AI-First series)
Fixes: recurring GTM work still done by hand, and AI used as one-off prompts instead of an operating system the whole team runs on.
Best for: a founder or operator at 0-10M ARR with a lean, multi-market team and no ops hire, ready to move one weekly GTM job off their attention and onto an agent.
Week 1, Define: pick one weekly GTM job. Write what good output is and where it lands, a Slack channel or a Gmail draft. Draft one context file: ICP, positioning, voice.
Week 2, Build: create the Routine on that schedule, or a webhook off your lead form. Paste the prompt and connect Slack, Gmail, or your CRM. Point it at the context file.
Week 3, Validate: let it run for a week. Check each output, tighten the prompt, and add a self-critique step until you would ship the output without edits.
Week 4, Operate: add a second Routine, share the skills with the team, and keep a human approval step on anything that goes external. Track agent output weekly, in one shared channel.
Paste this into your AI:
Here is a weekly GTM job I do by hand [describe it], and here is my ICP, positioning, and voice [paste]. Draft the agent spec: the exact steps, the input it reads, the output format, and where it should land. Then write the self-critique step it should run on its own output before handing it to me.
Also on the radar
Anthropic, “Redeploying Fable 5” (Jul 1): keep a human check on any agent whose output ships unread. Verify before you trust a Routine that runs without you.
Maja Voje and Kyle Poyar, “Claude for GTM Pulse Report” (Apr 24): the command line beats connectors on reliability, 100% versus about 72%, at a fraction of the cost, so build your most important Routine on the most reliable path.
Why this matters now
The operators seeing real returns share one move: they turned recurring GTM work into agents and gave those agents a shared context system to run on. Ownership of your instructions is the floor. The engine is what you build on it.
For a lean European team scaling toward 10M: put one recurring job on a schedule, hand the reading to Claude, and keep your hours for the customer conversation and the decision.
What is your number 1 takeaway this week? Reply or send me a DM.
Go deeper with GTMcraft OS
You do not need another newsletter, podcast or LinkedIn post. You need the system under it.
Read the plays. Every play, live and current, in the Operator Room.
Download for your archive. Take the PDF or workbook offline.
Get the markdown. Drop the raw play into your own stack or AI tool.
Build your own. Use the GTMcraft Notion databases (100,000+ signals) to spin up a play for your exact motion.
The full Plays database is over 100 plays and growing every week.
Curious to see a sneak peek of GTMcraft OS, DM me or reply on this email.
See you inside GTMcraft OS, your GTM co-operator. Not another community.
Koen
Your (human) GTM Agent
