The Playbooking Method (Rachel Woods)
A reusable framework for defining a problem with a client (or yourself) so AI can actually run the work. The core shift: stop asking “how do I prompt AI?” and start asking “how would I teach a human to do this?” Write the employee handbook, then hand it to the AI. The result is a playbook that runs in any tool — Claude Skills, ChatGPT custom GPTs/Agents, Microsoft Copilot, Gemini Gems, or one-step-per-Zapier-action — so you own the playbook, rent the tech.
The Four-Component Playbook
Every playbook is just these four things. When defining work with a client, this is the discovery scaffold.
| Component | What it is | Example (newsletter playbook) |
|---|---|---|
| Trigger | What kicks it off — schedule or event | Monday 9:00 AM |
| Inputs | What varies each run (the human-shaped raw material) | Brain dump of topics top of mind |
| Steps | The actual process the AI follows | Pick topic → outline → draft → edit for voice → write subject line |
| Outputs | The end result | Drafted newsletter |
The newsletter playbook collapsed 2–3 hours/week of writing into ~30 minutes — ~130 hours/year saved on a single playbook.
The Mental Model Shift
When AI feels like a fight (you correct it, it adds emojis, you correct again, you give up and write it yourself), the missing piece is the playbook. AI without a playbook is a contractor with no brief.
To set up a human assistant well you’d give them: company context → what to do → how to do it → an employee handbook to follow. A playbook is that handbook — written for AI instead of for a person. Once it’s in writing, you can trust the output the same way you’d trust the assistant who follows the handbook to the letter.
Writing the Steps (The Specificity Layer)
A playbook is only as good as how each step is written. For each step:
- Be clear and non-contradictory — the most common failure mode is instructions that quietly conflict.
- Inject background context — what does the AI need to know about the company, audience, or constraint to do this step well?
- Show, don’t tell — give examples for anything easier to demonstrate than describe (voice, formatting, tone).
- Define success criteria — a checklist of what a good output looks like. The AI optimizes against it.
Where to Find Playbook Candidates
Four hunting grounds — useful both for your own work and for client discovery sessions:
- Existing SOPs — any documented process is already 80% of a playbook.
- Recurring calendar meetings — look at the last 1–2 months. For each recurring meeting ask: what’s the prep, what’s the follow-up, and could this meeting just be a playbook?
- Recurring to-do items — anything that takes >4–6 hours/month is over the threshold; playbooking it unlocks meaningful time.
- The wish list — things you’d do if time were unlimited. Some of the highest-leverage playbooks live here.
Running a Playbook
- Save the playbook in the AI tool of choice (Claude Skill, custom GPT, Copilot agent, Gemini Gem, Zapier flow).
- Invoke with the trigger phrase: “Run my newsletter playbook.”
- The AI loads the saved instructions, asks for the inputs it knows it needs, and walks the steps in order.
- You stop prompting per-task and start operating with AI.
Using This With Clients (Discovery)
When defining a problem with a client, drive the conversation toward filling out the four boxes:
- Trigger: When does this problem fire? (calendar event, inbound message, end of a campaign, end of a call)
- Inputs: What do you have in hand each time it fires? What’s variable vs. constant?
- Steps: Walk me through how you (or your best person) does it today. Then sharpen each step with context, examples, and success criteria.
- Outputs: What does “done” look like, and who consumes it next?
If a client cannot describe one of the four, that gap is the consulting work. Operational debt — the difference between how things should run and how they actually run — almost always shows up as a missing or fuzzy component.
Operational Debt (The Frame)
The gap between how you know things should run and how they actually run. Symptoms:
- Work on your plate that’s below your pay grade.
- Busy all day, not productive.
- Re-explaining the same thing to the same person more than once.
- Trying to delegate, going back-and-forth, giving up and doing it yourself.
Playbooking is the fastest way out — every playbook closes a slice of the gap.
Related
- business-streamlining-playbooks — concrete playbook candidates for an AI consulting practice, modeled on Rachel’s stack.
- extract-playbook-from-chat — drop-in prompt for mining an existing AI chat for the steps component.
- architect — the simplicity / cost-of-change / blast-radius lens applies when deciding which playbooks to build first.