The 90-second version
Your pipeline meetings each open by rebuilding the same picture. The assembly gets done 5 times a week and the decisions get done once, badly.
Your September and October events will produce conversations nobody follows up, because the follow-up has no owner and no written first message.
Your new hires ramp on knowledge that lives in your best rep’s head, so every hire and every new market starts from zero.
The European read: none of this needs a warehouse or a new vendor. Point the agent at the CRM you already pay for, keep the data in the EU under the contract you already signed, and start Monday.
The thread: the agent is not there to make the call. It is there to hand you the page, so the call is all that is left.
Read time: ~8 min
This week’s number 1 move
Put your weekly pipeline pre-read on a schedule, and keep the meeting for the decision.
Not a forecast the agent calls. A page the agent drafts. Pacing to plan, what moved since last week, what slipped, and the 5 deals that decide the month. You read it with coffee, you correct it, and the meeting starts at the argument instead of at the assembly.
From my week: the brief that produced this edition ran without me. On Friday at 17:00 an agent scanned my calendar, wrote me a summary of my own week, flagged the 2 things it found unusual, and then interviewed me about them. I assembled nothing. I answered questions and made calls, which is the only part that needed me.
My take: most AI rollouts stall because people start with the judgment and skip the assembly. Assembly is the boring 80 percent. It eats the week, and it is the part a machine is genuinely better at. So my rule is that the agent gets the collecting, the sorting and the drafting. I keep the last 20 percent, which is the call. What makes this work is not the model, it is context arrival. A mediocre prompt on good foundation files beats a brilliant prompt with no context. You do not need an AI engineer. You need the files, 1 connection to your data, and 1 workflow on a schedule. (LinkedIn DB, “Context arrival is the lever”, and “You do not need an AI engineer, you need foundation files, one MCP, and one workflow on a schedule”.)
Do first (under 30 minutes): write the 4 questions your weekly pipeline meeting always answers. Pacing to plan, the deals at risk, what moved, and the 1 number you defend upward. That list is the spec.
The rule: if a meeting starts by rebuilding the picture, the picture belongs on a schedule and the meeting belongs to the decision.
Sources: Growth Unhinged, Dee Kapila and Fin, “How to use Claude Code as a GTM leader” (August 26); Koen Stam, “Context arrival is the lever” (LinkedIn DB)
Hi, it is Koen Stam and welcome to GTMcraft OS: The European 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 and GTM operators scaling their next 2M, 5M or 10M ARR.
Problem 1: your pipeline reviews open by rebuilding the same picture
From my week: I sat in 5 separate pipeline and forecast conversations in 1 week. Different segments, different rooms, and every one of them opened the same way, by rebuilding the same picture from the same CRM before anybody could say anything useful.
Growth Unhinged published the fix this week, and the honest part is the timeline. Dee Kapila pointed an agent at her own data and her own Slack. Her weekly pacing, risk and renewal report now writes itself as 1 page, replacing a recurring 2-hour manual build. It took 3 quarters and 67 versions, which is the number everyone skips when they quote this story.
Why this lands on revenue and not on tidiness comes from the renewal side. Daphne Costa Lopes argues a renewal has to be forecast 6 months out, not in-month. You cannot forecast 6 months out if the picture only exists in the hour before the meeting.
My take: I do not want AI calling my number, and I would not trust it if it did. Forecasting is a judgment about people and commitments. Outsource that and you have stopped leading the team. What I do want is to never again spend the first 20 minutes of a review agreeing what the data says. Give the agent a voice, not a vote. It drafts, it flags what moved, it names what slipped. A human decides which slip is a real problem and which is a rep being conservative. The gain nobody mentions: when the page is identical every week you can finally see the trend, and pipeline debt stops being invisible until the quarter it lands in.
The European read: point the agent at the CRM or BI tool you already own, and keep the data in the EU under the contract you already signed. No warehouse, no vendor review, no new budget line.
Do first: write the 4 questions your weekly review always answers. That list is your spec, and it takes 20 minutes.
Do this week: write 1 prompt naming those 4 questions, the CRM objects to read, and the date window. Ask for 1 page plus a short block you can paste into Slack.
Do this month: run it every Monday for 4 weeks, editing the prompt once a week against what you actually asked. By version 4 it stops being a template and starts being your review.
The rule: the agent drafts the page, a human calls the number. Never the other way round.
You know it worked when: your forecast accuracy holds while the review gets shorter, and nobody exports to a sheet to rebuild the real answer. Founder read: you walk into your board update with a page you did not spend Sunday building.
The play: 5 steps to give AI a voice, not a vote, in your forecast call (2026 series). Inlined in full below.
Sources: Growth Unhinged, Dee Kapila and Fin, “How to use Claude Code as a GTM leader” (August 26); trumpet Digital Sales Room, Daphne Costa Lopes on turning customer success into a revenue engine (Podcast DB)
Problem 2: your event season books conversations and sources no pipeline
From my week: I went through every event my team has lined up for September and October with them this week, and separately ran a year-end gap analysis with a founder I coach. Both lists were long. Neither had a follow-up owner or a first message written against a single line.
Jordan Crawford published the inversion this week and it is the right one for events. Do not build the agent that reads everything and then decides what to say. Write 10 messages by hand for 1 named condition, get them right, then let the agent find every account that matches. Handwriting first, scale second.
Mason Cosby points at the same discipline from the account side. Start where signal already exists, not at cold awareness. An attendee list is the purest existing signal you will get all year, and most teams treat it as a lead import.
My take: events are the most expensive pipeline source in Europe and the worst instrumented. We pay for the stand, the travel and 2 days of selling time, and then the follow-up is 400 identical emails sent 5 days late by whoever has capacity. The move I run is to write the follow-up before the event, not after it. Three conditions, 10 handwritten messages each, agreed with whoever will actually send them. The agent does the matching and drafting against those 30 messages, and a human presses send. That is 2 hours of preparation deciding whether a whole quarter of events produces anything.
The European read: the autumn event calendar in smaller European markets is dense, local and repetitive. You will meet the same 200 buyers at 4 events between September and November, so a generic follow-up is not just weak, it is the fourth generic follow-up they have had from you this season.
Do first: pull your last 2 events and count how many attendee conversations produced a second meeting. If you cannot count it, that is the answer.
Do this week: name 3 conditions you can recognise from an attendee list, and hand-write 10 follow-up messages for 1 of them. Real messages, to real named people, in your words.
Do this month: point the agent at your September and October attendee lists with those messages as the pattern. It ranks and drafts. A named human sends inside 48 hours.
The rule: write the 10 messages by hand before you point an agent at the list. The agent scales a pattern, it does not invent one.
You know it worked when: event-sourced meetings held rises against the same spend, and follow-up goes out inside 48 hours instead of the following week. Founder read: your event-sourced pipeline is a number you can say out loud.
The play: 5 steps to automate the follow-up nudge without scaling a weak loop (2026 series). Builds the follow-up sequence and the owner before the automation, so you scale a loop that already works instead of a weak one.
Sources: On the Edge by Blueprint, Jordan Crawford, “Work backwards from the message” (August 29); Exit Five, Mason Cosby on how to simplify ABM (Podcast DB)
Problem 3: your new hire ramps on tribal knowledge, so your second market starts from zero
From my week: I ran 3 case interviews across 2 open roles this week. Every one of them ended with the same private question in my head. How long before this person produces, and what exactly do they have to learn from us that is not written down anywhere.
Charlie Hills gives the cheapest answer, and it is not a training platform. Anything you have retyped twice is a reusable skill. Write it once, properly, and a person or an agent can run it. The mechanics are learnable in an afternoon, not a hire, which matters when you have 2 open roles and no enablement function.
Ross Rich supplies the urgency. AI is widening the gap between your best reps and your middle, because the tools reward taste, and an unramped hire has none yet. The undocumented motion used to cost you a slow ramp. Now it costs you a rep who never catches up.
My take: processes scale and people do not, and hiring season is where that gets tested. The trap I watch for is treating onboarding as a content problem, so somebody builds a 40-slide deck nobody opens in week 3. What actually ramps a rep is the 6 or 7 artefacts your best rep produces without thinking. The discovery question set. The pricing objection answers. The 3 things we do badly. The account research pattern. Capture those as things an agent can run, and the new hire gets your best rep’s pattern on day 2 instead of month 4. The same set makes a second market survivable, because country 2 is just a permanent new hire.
The European read: in a smaller market the senior talent pool is thin, so a slow ramp is not a delay, it is a hole you cannot backfill inside the year. Documenting the motion is how you hire the person available rather than the person perfect.
Do first: open your last 5 outbound emails and your last 3 deal reviews, and mark everything you have typed a version of before. That is your first skill, and it already exists.
Do this week: write 1 of them properly as an instruction an agent can run. The input, the standard, and 2 examples of good. Start with the discovery question set.
Do this month: get to 6 artefacts and hand them to your newest hire as their week 1 kit, with your best rep reviewing output rather than teaching live.
The rule: anything your best rep does instinctively and nobody has written down is a hire that will not ramp and a market that will start from zero.
You know it worked when: ramp to first real pipeline shortens against your last hire, measured in weeks, not vibes. Founder read: your founder-led deal share falls while the number holds.
The play: 5 steps to ramp a new rep to real pipeline (2026 series). Turns the undocumented motion into a week-by-week ramp plan with an AI layer per step, so the new hire runs your pattern instead of inventing one.
Sources: MarTech AI, Charlie Hills, “Give me 10 minutes, I’ll teach you 80% of Claude Code” (August 23); GTMnow, “What the top 1 percent of sellers do differently” with Ross Rich, Accord (August 25)
Save this. 3 AI problems from my own week, each landing on a revenue number, each wired to a play you can run today.
Send it to 1 founder or GTM operator who is about to buy an AI tool to fix a reporting habit.
Also on the radar
The version count is the story. That weekly report took 3 quarters and 67 versions to become trustworthy. Budget iterations, not a launch date. Growth Unhinged (August 26)
Keep a note this week of every prompt you type a second time. That list is your first set of reusable skills. MarTech AI (August 23)
Model names and token pricing are moving weekly again. None of it changes the move above. Use the model you already have a contract with, and spend the time on context files.
Steal this move: put the weekly pipeline pre-read on a schedule in 4 weeks
Fixes: the review that spends its first 20 minutes agreeing what the data says, then runs out of time for the decision.
Best for: 2-5M and 5-10M teams running a weekly forecast or pipeline cadence across 2 or more markets, on the CRM they already pay for.
The 5 steps (over 4 weeks):
Week 1, define: write the 4 questions your review always answers, and the 1 number you defend upward. Nothing else goes on the page.
Week 2, build: write 1 prompt naming those questions, the CRM objects and filters, the date window, and the output as 1 page plus a paste-into-Slack block. Keep the data in the EU under your existing contract.
Week 3, validate: run it against last week’s real numbers and mark every line the agent got wrong. Correct the prompt, not the output.
Week 3, gate it: write the rule out loud. The agent drafts, flags and ranks. A human calls the number and owns the commit.
Week 4, operate: schedule it for Monday morning. Read it, correct it, and start the meeting at the first disagreement.
Template: a 1-5 rubric scored at week 4. Rows: 4 questions fixed, pacing correct against source data, slipped deals caught, risk flags accurate, human commit unchanged by the agent, page under 1 screen. Under 3 on accuracy means another iteration, not another tool.
Paste this into your AI:
Read these opportunity, activity and account records for the last 90 days: [paste or connect]. Produce 1 page: pacing to plan, what moved since last week, which deals slipped and by how long, and the 5 deals that decide this month. Flag risk, do not assign a probability. End with a 5-line block for Slack.
Full play, template and workbook inside GTMcraft OS. Reply or DM me to get access.
Why this matters now
For 2 years the AI question in GTM was which model you had access to. That question is closed, and the answer stopped being an advantage.
What is left is unglamorous. Whether the page exists before the meeting. Whether a human wrote the 10 messages before the agent scaled them. Whether your best rep’s 6 artifacts are written down anywhere a new hire can reach.
All 3 are assembly problems, not intelligence problems. That is why they are worth handing over, and why buying a tool never fixes them. A tool with no context just runs your bad version faster.
Autumn is the wrong season to start a platform project and the right season to put 1 workflow on a schedule.
3 questions for the room
What does your weekly review spend its first 20 minutes on, and could a page have done it?
Who owns the follow-up for your next event, by name, and what is the first message?
Which 6 things does your best rep do instinctively that a new hire cannot find written down anywhere?
Reply or send me a DM.
Go deeper with GTMcraft OS
Scale on plays and skills, not prompts.
You do not need another newsletter, podcast or LinkedIn post.
You need the system under it.
Read the plays. Every single play is free and live in GTMcraft OS on Notion. Click through from any edition.
Build your own. GTMcraft OS members connect the full libraries, 100+ plays plus 100,000+ LinkedIn, newsletter and podcast signals, straight into their own AI or Claude context, so their team builds on the whole corpus instead of one-off prompts.
The single plays are free to read. Wiring the whole corpus into your own operating system, and keeping it current, is what the membership gives you.
Master your human GTM skills AI can’t replace inside GTMcraft,
Koen
Your (human) GTM Coach
—
PS. Co-written by Wispr + 3 GTMcraft Skills + Claude Opus 5; edited & approved by Koen
The Future GTM Operator is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.