Skip to content
Ameen Abozaid
SRB-022

AI tools across teams

  • 7 Lessons
  • 3.5 h practice (estimate)
On this page 9 topics

Overview

Topics

  • Rolling out coding assistants
  • usage policy
  • security review of AI tools
  • honest impact measurement

Learn first

3.5 h of practice (estimate)

Learning outcomes

  1. Plan a rollout of AI coding assistants across several teams with pilots, training, and support.
  2. Write a usage policy covering code confidentiality, secrets, licensing, and review requirements.
  3. Run a security review of an AI tool (data handling, retention, access, admin controls).
  4. Measure impact honestly with delivery, quality, and developer-experience metrics rather than vanity metrics.

Lessons

  1. 01AI assistants as a team capability
  2. 02Pilots and rollout
  3. 03Usage policy
  4. 04Reviewing an AI tool's security
  5. 05Measuring impact without fooling yourself
  6. 06Demo: designing a two-team pilot with a control group
  7. 07Demo: reading a pilot's results that look better than they are

Lab and exercises

Lab

SRB-022-LAB1 · Assistant rollout at FulfillCo

Outcomes: 1–4 · Time: ~3.5 h

  1. Write a pilot plan for an AI coding assistant across the four teams, with a control group, training, support, and a review gate.
  2. Write the usage policy (confidentiality, secrets, licensing, review) and run the tool security review with a role-played security lead.
  3. Write a measurement plan using DORA and quality metrics, with a baseline, a defect-rate check, and a review-load measure.

Acceptance: policy approved by a role-played security lead; measurement plan includes a baseline and a defect-rate check.

Exercises (assessment: rollout plan defense)

IDOutcomeTask
SRB-022-EX011Case: two of four teams already use assistants personally. Write the rollout plan outline with pilot, control, training, and gate.
SRB-022-EX021List three risks of rolling out widely without a pilot and how the pilot mitigates each.
SRB-022-EX032Draft a one-page usage policy and name its owner and review date.
SRB-022-EX042A developer pasted a production config with credentials into an assistant. Write the incident steps and a policy change.
SRB-022-EX053Fill a security review questionnaire for a candidate tool: data handling, retention, access, admin controls, with evidence sources.
SRB-022-EX064Case: pilot teams show 25% more pull requests and higher defect rate. Write the honest reading and next step.
SRB-022-EX074Design a measurement plan with baseline, control, three metric families, and a decision rule.
SRB-022-EX081Defend your rollout plan in five minutes against three role-played objections (cost, quality, juniors).
Show solutions
  • EX01 Pilot with two teams (including a skeptic) and two controls; training on verification and data limits; support channel; gate at week 9 requiring policy and security approval.
  • EX02 Unmeasured quality drift, data leakage, inconsistent practice; the pilot limits exposure and produces a baseline and evidence.
  • EX03 Sections: scope, approved tools, allowed and forbidden data, secrets, licensing note (legal), review rule, incident path, exception route, owner, review date.
  • EX04 Rotate the credentials immediately, report the incident, check vendor retention and deletion, add secret scanning on prompts or IDE where possible, and update training.
  • EX05 Example rows: data sent and training use; retention and deletion; SSO and roles; audit logs; repository exclusions. Answers must cite contract or documentation; mark unknowns as open risks.
  • EX06 The result is mixed: throughput up, quality and review load worse, possible novelty and selection effects. Extend the pilot, fix review capacity, and compare with control.
  • EX07 Questions with metrics: lead time and frequency (delivery), change failure and defect rate (quality), survey and toil time (experience), review time per PR (load). Rule: widen only if quality is no worse and at least one delivery or experience measure improves relative to control.
  • EX08 Evidence is the plan's baseline, policy, and gate; answer cost with measured time saved versus license, quality with defect check, juniors with training and review rules.
  • Outcome 3 is also covered by the lab security review.

Workplace evidence: an AI tool rollout or policy you led and what the data showed.

Quiz

Outcomes: O1 rollout planning · O2 usage policy · O3 tool security review · O4 honest measurement.

A self-check for you: nothing is graded, saved or sent anywhere. Choose an answer, then check it. You can try again as often as you like.

  1. 1Your teams already use assistants individually. What should the lead do first?O1
  2. 2Why include a skeptical team in a pilot?O1
  3. 3A pilot widening gate should require:O1
  4. 4In a usage policy, who is accountable for AI-generated code in a pull request?O2
  5. 5Which belongs in a usage policy?O2
  6. 6A key retention question for a tool is:O3
  7. 7Which is an integration risk to review?O3
  8. 8Which is a vanity metric?O4
  9. 9Pilot teams ship more PRs, and so does the control team. What follows?O4
  10. 10A known side effect to measure when generated code increases is:O4

Practical part

Defend the rollout plan (SRB-022 assessment) and complete SRB-022-LAB1. Pass: at least 8/10 on the quiz, a policy approved by the role-played security lead, and a measurement plan with a baseline and a defect-rate check.

Common mistakes

Measuring "lines of code generated"; rolling out without a policy; ignoring review load created by more code.

Assessment

Rollout plan defense.

Workplace evidence

An AI tool rollout or policy you led and what the data showed.

Career relevance

AI governance and adoption at organization scale (STF-021, STF-022).

Type to search levels, phases, modules and lessons.