Use this when
- Leadership says “we are AI-native” and the evidence is browser tabs.
- Every function bought a different assistant, and nothing shares context.
- You are about to automate outbound, content, and events in the same quarter.
- A founder wants a command center before the CRM is updated.
Do not use this when
- There is no primary motion. Stay in channel strategy.
- You still need to pick which problem AI should touch. That is AI use-case selection.
- You need the architecture of one workflow. That is AI workflow.
- You need product names. Start in TOOLS.md. A community spreadsheet of logos is not this page.
A few useful terms
Keep this in mind
Score the motion you have, then climb one rung on a channel that already works. Do not automate what you have not made work by hand. Do not skip to a dashboard because a video called it a GTM engine.How to do it
Step 1: Assess each process separately
Write one line per motion you actually run: outbound, owned content, events, inbound capture, first meeting / demo, expansion. For each: which rung, what evidence, what would make the next rung true. Most teams are 1–3. “We use AI” that means Perplexity on research and ChatGPT on DMs is rung 1. It can help. It is not an OS. Over-automating a weak DM motion at rung 1 is how buyers learn to ignore you.Step 2: Choose processes that already produce useful results
AI scales what exists. If LinkedIn content already books conversations, that is a candidate. If events are a monthly hope, do not build an invite agent first. Refuse a stack that “covers the whole funnel” while no channel has a written win.Step 3: Write down the context the AI needs
Rung 2 hits a wall when every tool starts from zero. The fix is not a new SKU named “context.” It is a small, current file the workflows may read: ICP, offer, objections, banned claims, what “good” looks like. Content strategy and positioning already own those facts. If they are empty, agents will invent them.Step 4: Connect one reviewed workflow
A loop is one output that becomes another motion’s input—transcript objections into the next page, closed-won traits into the next list—with a human still allowed to stop a send. Sheets and a CRM can be the first command center. Choose a shared view someone will keep current before building a custom dashboard. Rung 4 is “enough of the stack is connected that the system can learn.” It is not “we posted that we run on autopilot.”Step 5: Treat future capabilities as uncertain
A system that senses Slack, Discord, and the roadmap and reallocates every channel without a human steer is a research story. You may watch it. You may not staff or budget as if you are already there. The human still reviews external messages—see AI workflow.Worked example (illustrative)
Sales-assist ops tool. Founder-led. Not a survey.Copy: maturity one-pager (fill)
- Motions we actually run (not the org chart):
- Rung and evidence per motion:
- Two channels already producing a sales-usable step:
- Context file the tools may read (and who updates it):
- One loop we will connect this quarter:
- What we refuse to automate:
- Rung-5 language we will not put in a board slide:
Before you start
- Each motion has a rung and a sentence of evidence.
- We named channels that already work; AI will not “create” a motion.
- Shared context exists as a file people will maintain—or we admitted it does not.
- One loop is written before a command-center SOW.
- External sends still have a human gate.
- No community AI-logo sheet substituted for this score.
- We did not buy a product because it claimed autopilot.
Metrics
Do not count AI seats, or resemblance to a six-step YouTube ladder, as maturity.
Common mistakes
- Calling browser AI a GTM OS.
- Automating a channel that has not worked manually.
- Buying a “context graph” with empty ICP.
- A dashboard of agents nobody owns.
- Skipping to rung 4 on slides while the CRM is fiction.
- Treating rung 5 as a 12-month OKR.
- Importing a Pavilion/community vendor list as the stack.
What to read next
Which problem deserves a bet is AI use-case selection. How one workflow is staged and gated is AI workflow. How a test dies or scales is experimentation. Which tool may write into the record is MarTech governance. How answers get found is SEO and AEO.Sources and evidence boundary
This is an owner-maintained operating synthesis. It is not a licensed maturity product, not a consulting diagnostic, and not a forecast that full autonomy is near. The six-rung shape (random → assisted steps → end-to-end on working channels → owned stages plus shared context → connected loops → refuse autopilot as the operating target) is distilled from a public explainer (Roman J. Georgio, how ai-native is your gtm team?! (6-step ladder)). That video is a method prompt, not a source to copy. Its founder vignette, tool shout-outs, “possible within a year,” newsletter raffle, and booking offer are not this library’s facts or a service to buy. A circulated community spreadsheet of GTM AI products (company / site / use-case tags) is a logo list, not a maturity score. It is not imported into TOOLS.md. A global survey that more companies “use AI” while attributed EBIT stays flat is also not your rung (McKinsey QuantumBlack, The State of AI, 2026 edition as reported). That series is a method prompt (adoption ≠ profit; high performers talk about redesigned work). Its percentages, “high performer” cuts, and consulting narrative are not this library’s facts.Copyright © 2026 Ivan Xu. All rights reserved. See the copyright and reuse terms. Canonical source: github.com/weilun88313/B2B-Playbook