Playbooks for AI agents · Reviewed 14 Oct 2026

What can I safely hand to my AI agent, and what should I approve?

Plain-English playbooks for small-business owners and solo operators who use an AI agent: Claude, ChatGPT, Gemini, Microsoft Copilot, or an open agent or automation. Every playbook gives the job a Handoff rating and names the exact moment a human should approve.

The Handoff rating, in four levels

The rating describes the job, not the agent. The higher the number, the more of a human checkpoint it needs.

  1. Handoff 1

    Read-only research

    The agent reads and reports. It changes nothing outside the chat.

  2. Handoff 2

    Drafts for review

    The agent prepares drafts or files. A human sends, files or publishes.

  3. Handoff 3

    Acts, but reversible

    The agent changes things that can be undone: labels, moves, scheduled items, record edits.

  4. Handoff 4

    Acts irreversibly, or with money or people

    Payments, messages to customers, cancellations, public posts. A human approves each one.

How the rating works

Playbooks

Showing all 30 playbooks.

"Suggested" agents are the ones with a relevant feature we found in their official documentation. Other agents may still do the job.

Want it all on paper? The Starter Pack

A printable PDF with an approval checklist for each Handoff level, an agent permission settings worksheet, a handoff policy template, a weekly review checklist, an incident log and all 30 playbooks, one page each. £19, one-off, instant download. The playbooks on this site stay free.

See what's in the Starter Pack

Get new playbooks by email

A short, occasional note when new playbooks are added or a rating changes. No spam; unsubscribe any time.

Newsletter sign-up

You'll get a confirmation email first. Sent with Buttondown. See our privacy notice.