How to use this playbook as a team lead

Status: drafted · Time: 10 min · Audience: leads Outcome: Turn the playbook into a team adoption rhythm without turning it into homework.

The job is not “make everyone read it”

If you lead a team, the playbook is not a PDF to forward with “please go through this by Friday.” That will produce exactly what most training produces: polite acknowledgement, low retention, and no shipped evidence.

Your job is to create a rhythm where people can safely try the work, see visible wins, and earn belts through artefacts the team already values: PRs, recipes, skills, better docs, cleaner handoffs, less ops drag.

The adoption unit is not the chapter. The adoption unit is the week.

Run the first two-week adoption loop

Start with a small loop. Do not announce a transformation.

Day 1: baseline. Ask everyone to take §0.10 — Self-assessment. Capture where the team actually is: not who is “good at AI,” but who has evidence at each belt.

Day 2: one shared target. Pick one target for the next two weeks. For a non-coding team, use Ops 101: one workflow saving measurable time. For a product/design team, use White or Yellow readiness: setup health, first PR, or a bug fix.

Days 3-8: visible attempts. Put attempts in the open. A stuck setup, a failed prompt, a half-working agent, a draft PR. Public attempts beat private polish because the team learns from the friction.

Days 9-10: evidence review. Review what actually shipped. Count artefacts, not optimism. A recipe that saved two hours, a PR under review, a teammate unblocked by a better context file: all count.

Then repeat with a sharper target.

Make champions useful, not heroic

Every team will have one or two people who sprint ahead. Use them carefully.

Do not turn champions into the unofficial support desk. That creates a dependency and quietly punishes the people who learned fastest. Instead, ask champions to produce reusable artefacts:

a five-minute screen recording of one workflow;
a short recipe in the team wiki;
a working prompt with the sensitive details redacted;
a small skill or checklist others can fork;
a before/after note showing time saved.

The champion’s job is not to answer every question. The champion’s job is to leave behind a path.

Protect the safety brief

The fastest way to kill momentum is a sloppy incident. Make §0.11 — The safety brief non-negotiable before anyone uses AI on real Razorpay work.

As lead, model the reflex out loud:

“Redact the specific customer details; keep the shape.”
“Do this in the approved harness.”
“Money-handling logic needs the higher review bar.”
“If you’re not sure, pause and ask in the public support channel.”

This is not fear. It is professionalism. In fintech, the guardrails are part of the craft.

What to measure

Track a few things lightly:

number of people with a completed setup health check;
number of real PRs opened with AI assistance;
number of Ops 101 workflows saving measurable time;
number of reusable recipes, skills, or KB pages contributed;
number of support questions resolved publicly rather than in DMs.

Do not overbuild the dashboard. A simple tracker is enough until the behaviour is real.

What you should carry forward

The adoption unit is the week, not the chapter.
Start with one two-week loop and one shared target.
Champions should leave artefacts, not become human queues.
Safety is part of the craft, especially here.
Measure shipped evidence, not enthusiasm.
The next chapter, §0.10 — Self-assessment, helps each reader choose where to start.

Previous: ← 0.8 How to use this playbook as a reader · Next: → 0.10 Self-assessment