Teams and governance

Principle

A team is not “more people typing.” It is a system for making decisions when you are not in the room — including decisions made by agents on your behalf.

Governance is the minimum process that keeps trust with users and sanity with builders: who owns what, what must be reviewed, how incidents are handled, and how AI use is disclosed and bounded.

At scale, some firms treat coordination as an intelligence layer — routing context and composing capabilities instead of adding meetings. That is not the same as bolting copilots onto hierarchy. The question for your team: is AI speeding up handoffs, or replacing the handoff graph with legible owners and gates? Organization as intelligence pressure-tests the org-design bet.

The decision

  Areas (product, eng, ops) ──► owners
│
▼
Shared gates (auth, pay, deploy, schema)
│
▼
Agents inherit boundaries from docs + CI
│
▼
Incidents → postmortem → update boundaries

Lightweight team primitives

AI governance specifics

• Disclose when AI assists customer-facing output if your users care (health, legal, money).
• No customer PII in public model prompts without policy.
• Audit agent actions on production systems — logs, PRs, not chat.
• Train the team on pair and agentic chapters — shared vocabulary reduces fear and misuse.

Workflow

1. Map modules to owners — even if two people wear all hats today.
2. Add CODEOWNERS for migrations/, auth/, billing/, .github/.
3. Align on weekly iteration bet (Chapter 15) — one priority, visible.
4. Run blameless postmortem after incidents — update agent boundaries.
5. Quarterly: prune process nobody uses.

Common mistakes

• Everyone can merge everything until something breaks badly.
• Agents with prod secrets because it was faster once.
• Process theater — meetings without decisions logged.
• Hiring before workflow is documented — chaos at scale.

Artifacts

• templates/team-charter.md — owners, gates, rhythms.
• On-call or escalation doc — even if on-call is you for now.

Further reading

• Organization as intelligence — footnotes spin-out from this chapter
• Chapter 10 — Agentic development
• Chapter 15 — Iteration
• Chapter 00 — Principles (humans own outcomes)

v1.0

This chapter closes the first full pass of the handbook. Tools will change; decisions, artifacts, and revision in public should not. Bump handbookVersion when you add chapters; spin sharp principles to footnotes when they stand alone.