Capstone — Your Agency’s AI Ops Playbook
Bring everything together into one working plan you'll actually use
What this lesson is
Eleven lessons back, you audited your workflows. Since then you’ve built qualification rubrics, discovery-to-proposal pipelines, onboarding sequences, reporting templates, SOPs, an action-item system, and a knowledge base. This capstone isn’t new material — it’s the lesson where you stop treating these as separate exercises and assemble them into a single, working playbook specific to your agency.
Pull your work into one document
Gather everything you produced across this course — your workflow audit backlog (Lesson 2), your traffic-light data rule (Lesson 3), and the templates and prompts from Lessons 5 through 11. Then use Claude to assemble it into a single reference document your whole team can use.
I'm assembling an AI Ops Playbook for my agency from the workflows
I've built over the last several weeks. Help me organize the
following pieces into one coherent document with a consistent
format: workflow name, when to use it, the prompt template, who
on the team owns it, and any data-handling notes.
Here's what I have so far (pasted below, in no particular order):
[paste your qualification rubric from Lesson 5]
[paste your discovery brief and proposal prompts from Lesson 6]
[paste your onboarding sequence from Lesson 7]
[paste your reporting template from Lesson 8]
[paste 2-3 SOPs from Lesson 9]
[paste your action-item prompt from Lesson 10]
[paste your knowledge base index from Lesson 11]
Group them by client lifecycle stage first, then internal ops.
Flag any obvious gap - a stage of the client lifecycle we don't
have a workflow for yet.That last instruction matters most. The gaps Claude flags here are your real to-do list — not a hypothetical one.
Design your rollout, not just your document
A playbook nobody adopts is shelfware. Before you circulate it, decide how it actually gets used day to day.
| Question | Why it matters |
|---|---|
| Where does it live? | If it’s not where the team already works, it won’t get opened — link it from your existing wiki or project tool, don’t create a new destination |
| Who owns keeping it current? | Playbooks rot the same way SOPs do (Lesson 9) — name an owner, not “the team” |
| How do you introduce it? | One workflow at a time, starting with the highest-frequency win from your Lesson 2 audit — not a company-wide mandate on day one |
| How do you handle skeptics? | Use the side-by-side technique from Lesson 4 on real work, not a demo |
Key principle: The playbook’s job isn’t to impress anyone — it’s to make the second, third, and fiftieth time someone runs a workflow as fast as the first time you got it right. Optimize for reuse, not polish.
Set a revisit date
Your agency’s workflows will change — new tools, new client types, new team members with new ideas about what’s painful. Put a recurring date on the calendar (quarterly is reasonable for most agencies) to rerun the Lesson 2 audit process and update the playbook, rather than letting it go stale the way most documentation does.
It's been a quarter since we built our AI ops playbook. Help me
run a quick refresh: here's what's changed on our team and with
our clients since then [describe changes]. Which parts of the
existing playbook (pasted below) are still accurate, and where do
we need new workflows given what's changed?Build it: This is the real assignment. Gather your outputs from Lessons 2 through 11, run the assembly prompt above, and produce your agency’s first working AI Ops Playbook. Share it with one team member this week and get their honest reaction before rolling it out further.
Where to go from here: You don’t need to systemize everything at once. Pick the single highest-frequency, highest-pain workflow from your original audit, get it working well with your team, and only then move to the next one. A playbook with three workflows your team actually uses beats one with fifteen nobody opens.