Start with a true story — one woman's, told in an interview last week. Rachel Woods runs her business on AI agents. One Tuesday she woke up too sick to work, opened her laptop, typed a single word, and went back to bed. That word was the trigger phrase for her morning routine. While she slept, her AI triaged her inbox, rescheduled her meetings, fixed a client proposal, and sent the right payment link to a customer. Eight hours of work. Nobody driving.
The lesson isn't the software. It's the recipe she wrote before she ever got sick.
A playbook is exactly that: the way you do a piece of work, written down step by step so an AI can follow it the way your best staffer would. Four parts, one page:
Part 1
Part 1 of 4
A schedule ("every Monday morning"), an event ("after every board meeting"), or a phrase you type. If the AI doesn't know when to begin, it never begins.
Your answers are saved only in this browser. Nothing leaves your device.
Saved ✓Part 2
Part 2 of 4
The raw material for this run: names, amounts, dates, your template, the last version. An AI is only as truthful as what you hand it — vague in, fabricated out.
Part 3
Part 3 of 4
This is the whole game. Write the steps the way your best person does them, in order, with the judgment calls spelled out: verify this against that, use this tone, never do this without that. If you can explain it to a new hire, you can write it for an AI.
Part 4
Part 4 of 4
Name the finished product: a sent email, a filed report, a log entry. Done needs a definition, or the work never ends.
Now watch it work on something real: the donor thank-you.
Trigger: a new gift appears in your donation records. Inputs: the donor's name, the amount, what the gift is designated for, your last note to them, and your organization's thank-you template. Steps: verify the amount against your records — never thank someone for money you can't confirm. Draft the note in your voice, naming the specific designation. Check it like a skeptic: every fact traced to the record, no invented details. Then a human reads it before it goes out. Output: a sent thank-you, plus a log entry: who, when, how much.
Two things the hype usually skips. First: the routine work runs on its own, but a person approves before anything leaves the building. Drafting is the AI's job; sending is yours. Second: write the playbook in a document you own — not locked inside one company's tool. When the tool changes (and it will), your recipe walks with you. Own the recipe. Rent the kitchen.
And when something goes wrong — and it will — fix the recipe, not just that one note. Every miss makes the playbook smarter. That's the compounding nobody talks about: the playbook you write this month is better than the one you wrote last month, because it remembers every lesson.
Time saved: the hours your most repetitive work steals every week. Value: work done your way, even when you're not there to drive it.
Published Friday, September 25, 2026 as Issue #3 of the Sawdust Silicon newsletter — read the full issue, with this week's frontier and free fuel. This page is the playbook itself, frozen as published: playbooks never change after they go up. If a better way appears, it gets a new number, not a rewrite.
Free forever. One playbook every Friday.
Back to the shelf · ← Playbook No. 02 — Grants, faster · Playbook No. 04 — The honest energy question →
One tested AI playbook per issue — for grant writing, donor thank-yous, volunteer scheduling, board reports — you can put to work the same day you read it. Free. One email per issue. Unsubscribe anytime.