Company Brain Playbook · Chapter 8 · Free to read

Your first 30 days.

Start with one team. Do not announce a company-wide knowledge transformation before you have a working pilot.

Week 1

Diagnose

  • Interview your team lead and 2–3 heavy AI users. Ask: what context do you re-explain constantly? What decisions keep getting re-litigated?
  • Map where useful context already lives: Drive, Confluence, Jira, Slack, meeting notes, CRM, support, code.
  • Pick three use cases. Proven first picks: prepare me for this meeting · what did we decide about X? · draft this recurring document properly.

Week 2

Design

  • Set up the store: company, teams, products, workflows, decisions, meetings, skills, repo context.
  • Write the first documents: a one-page strategy, the pilot team's context, the top two workflows — and the last 10 meaningful decisions. Writing the decisions doc is the single most clarifying exercise your team will do this quarter.
  • Name the owner of every document and appoint 2–3 maintainers.

Weeks 3–4

Install and use

  • Connect your team's agents to the brain — first through the tools they already use, and only add a persistent runtime if the workflow genuinely needs Slack/Discord presence, CLI execution, webhooks, or scheduled jobs.
  • Run the three use cases inside real work — the actual Monday meeting, the actual PRD due this sprint. Not a demo.
  • Every gap an agent hits is a capture task: write it the same day. Start the write-back habit — every decision lands in the brain within 24 hours.

What to measure

  • Time to prepare a meeting brief (typical result: 45 minutes → 5)
  • Repeated questions answered by agents instead of people
  • Decisions captured per week — every significant one within 24 hours
  • Your team’s own answer to: “would you give this up?”
  • Runtime usefulness: did a persistent agent remove coordination work that a normal chat/tool workflow could not?

If the pilot works, you have something more useful than a tool rollout: a repeatable pattern the next team can copy.

This is 1 of 9 chapters.

The full playbook adds what to capture, what should never be ingested, the document structure with a worked example, how agents connect, and the 10-point readiness self-check — free, sent to your inbox.