What is playbooks in InstinctGTM?

Part 3 · Operate

Playbooks

A starting shape for your GTM plan, built for one type of business: which channels to run first, in what order, and what each one tests first.

01What it is

A playbook is a starting point for the AI team’s plan, one per type of business: sales-led and self-serve software, agencies, consultancies, productised services and developer tools. It names the channels to start with, in which week, what each one tests first and the starting rule for its bars, and which formats and Almanac entries it builds on. It is a prior, not a template: your own answers come first, and the plan is still written for your business.

02Why it exists

Starting a plan from a blank page asks you to know which channels fit a business like yours before any of them has run. A playbook carries that starting shape, so the first version of the plan begins somewhere sensible and the tests can do the rest.

It never outranks you. What you confirmed about the business, your hours and your budget decide what fits; a playbook only suggests the order and the first test. It also states no result: a playbook earns the label "built from tested tactics" only when every Almanac entry it links has passed a test in the system, and until then it says it is a starting playbook.

03How to use it

01Playbooks wait behind the playbooks flag, and only a playbook the admin published is ever offered.
02On the first step of starting the AI team, after your site is read, the setup says which playbook your business looks like, with the words from your site that decided it. Start from it, pick another, or choose No playbook.
03The next version of the plan starts from your choice. Each channel it kept from the playbook says so in its why line and carries a Playbook chip, and the plan page says which playbook and version it started from.
04On the plan page, Add a channel from a playbook takes one channel from any published playbook. It joins as proposed and runs only after you approve it and its rules; one that does not fit your hours or money is refused with the reason.
05Each published playbook has a public page at /playbooks. Its free plan form already has the playbook chosen, and when you start the team from that plan it is chosen for you.
06Goal playbooks, behind their own goal_playbooks flag as well, answer a second question on the Rules step: your main goal for the next ninety days (find design partners, land your first ten customers, raise from investors, or a launch week), or None of these. The goal decides which channels come first and when each starts; the business playbook beside it still decides the channels and their formats. A channel the goal put first ends its why line “First because your goal is” and the plan page says which goal and version it starts from.

04What good looks like

You give it: Ledgerline, month-end close software for finance teams, whose site says "Book a demo. An enterprise platform for finance teams, with SOC 2 security and integrations", with a zero budget and five hours a week, and no model key configured.
You get: The site’s own words suggest the B2B SaaS, sales-led playbook (book a demo, enterprise, platform, security, soc 2, integrations). The plan starts with its channels, each ending "From the B2B SaaS, sales-led playbook." and citing it, passes every gate, and runs no paid channel, because the budget is zero whatever the playbook lists.
Why: matchByWords counts each playbook’s fit signals in the business’s own words, and buildLocalPlan keeps only the playbook channels the budget band allows and the hours fit. pnpm eval:playbook-plans runs this case (0001) and eleven more.

Derived from the rules in the code

05What weak looks like, and what it costs

You give it: An agency playbook whose cold email template said its first test was "the change you saw, or the work it usually leads to".
You get: The plan written from it was refused by the claims gate: "leads" is a category word this product does not use. The seed was reworded before it shipped.
Why: A playbook is held to the same gates as every plan, so its own words cannot slip a banned word, a number or a promised result into a plan. tests/playbooks.test.ts runs every seed against the plan’s banned words and the kill list.