⌂ Atlas
▸ Internal · Delivery
govrn.ai
Delivery playbook. Execute on time and on quality. Build the system. Enter the access code.
Invalid code.
⚠ Internal · Delivery playbook for MCG/Kelly · confidential
▸ Execution & Delivery

Build it fast. Build it right. Deliver on promise.

Your job: execute the design, hit the timeline, and deliver zero-production-incident quality. Sales landed the deal. SA designed it. You build it, test it, deploy it, and hand it off. Here's what excellence looks like.

MCG/Kelly Services
01 · Your Charter

What you own.

You own the execution. On-time, on-budget, on-quality — zero exceptions. Your success metrics: deliver within the budget timeline, zero production incidents, client trains on day 1.

02 · Phase 1: Discovery

Map the AI estate. 3 weeks. On schedule.

SA designed how discovery runs. You execute it. Collect data from cloud APIs, telemetry, logs. Normalize it. Deliver a clean inventory to the assessment team.

Your sprint (weeks 1–3)

  • Week 1: Data access (cloud creds, API keys, log setup). First data pull. Validate quality.
  • Week 2: Normalize data into the inventory schema. Dedup. QA.
  • Week 3: Final inventory + shadow-AI report. Ready for assessment.
  • Daily: Sync with SA on data quality. Flag blockers fast.

Success criteria (you deliver)

  • ✓ 95%+ coverage (assets/models found).
  • ✓ Clean, deduplicated inventory (no duplicates, no junk).
  • ✓ Shadow-AI findings (unapproved tools + users).
  • ✓ Zero data loss or errors in normalization.
  • ✓ Documented data flow (for ops handoff).

Delivery wins: If you deliver 95%+ coverage on day 21, assessment starts on day 22 (no delay). If you miss, you cost the engagement a week.

03 · Phase 2: Assessment

Score the controls. 4 weeks. Rigorous QA.

SA designed the control library + scoring logic. You execute. Score each asset/model against each control. Gather evidence. Prepare the readout.

Your sprint (weeks 4–7)

  • Week 4: Assessment setup. Load control library. Confirm scoring logic with SA.
  • Weeks 5–6: Score controls. Gather evidence for each finding. Build the readout.
  • Week 7: QA & finalization. SA reviews findings. Prepare slides.
  • Daily: Flag high-risk findings to SA. No surprises in the readout.

Success criteria (you deliver)

  • ✓ All controls scored (no gaps, no unknowns).
  • ✓ Evidence gathered & audit trail clear.
  • ✓ High-risk findings validated independently.
  • ✓ Readout slides ready (data + narrative).
  • ✓ SA approves before week 8 readout.
04 · Phase 3: Implementation

Build the system. 8 weeks. Sprint cadence.

SA designed the architecture. You code it. Build the registry, intake engine, monitoring agents, and dashboard. Test rigorously. Deploy.

Your 2-week sprints (weeks 8–15)

  • Sprint 1 (w8–9): Registry schema + API. PostgreSQL setup. Baseline CRUD.
  • Sprint 2 (w10–11): Intake workflow state machine + approval gates.
  • Sprint 3 (w12–13): Monitoring agents + alert rules. Test extensively.
  • Sprint 4 (w14–15): Dashboard implementation + styling. Tie to APIs.
  • Each sprint: Code review, unit tests, integration tests, no tech debt.

Success criteria (you deliver)

  • ✓ All SA design points implemented (no shortcuts).
  • ✓ 80%+ code coverage (unit + integration tests).
  • ✓ Zero production bugs in week 1 of operations.
  • ✓ Performance targets met (API <100ms, dashboard <2s).
  • ✓ Deployment plan ready. Runbook written.

Delivery wins: If you ship on day 105 with zero defects, ops ramp is smooth + client confidence is high. If you ship on day 120 or with bugs, operations explodes.

05 · Phase 4: Operations Ramp

Go live. Support. Train the client. 4 weeks.

The system is live. You're on-call for month 1. You fix bugs, handle escalations, and train the client's team to run it solo.

Your ramp (weeks 16–19)

  • Week 1: Go-live support (on-call, bug fixes, escalations).
  • Weeks 2–4: Reduce coverage to 50%, then 20%. Client takes ownership.
  • Training: 4 sessions on ops, monitoring, troubleshooting.
  • Handoff: Runbook complete. Client trained. You step back.

Success criteria (you deliver)

  • ✓ Zero production incidents in month 1 (you prevented them).
  • ✓ Client team passes readiness assessment (trained).
  • ✓ Ops runbook is complete + client approved.
  • ✓ System stability confirmed (99%+ uptime).
  • ✓ Clean handoff. You exit gracefully.
06 · Your Checkpoints

What you sign off on. What Sales / SA must do.

CheckpointYou (Delivery)SalesSA
Kickoff readiness Confirms: crew ready, tools set, no blockers Warm handoff to client team Briefs you on discovery design
Discovery completion Executes discovery, delivers inventory — QA's inventory, approves handoff
Assessment completion Scores controls, gathers evidence Prepares to present readout Reviews findings, approves readout
Implementation design approval Reviews: is this buildable? Do I have blockers? — Signs off on the architecture
Sprint delivery (every 2 weeks) Code, test, integrate, deploy to staging — Reviews: does this match the design?
Go-live readiness System is production-ready. Runbook written. Client trained. — Final design verification
Production handoff On-call month 1. Trains client. Steps back week 4. — Available for ops questions

Legend: R = Responsible (does the work) · A = Accountable (signs off) · C = Consulted (input) · I = Informed (kept in loop)

07 · Quality Non-Negotiables

What "done" means. Zero exceptions.

For discovery

  • 95%+ coverage — not 85%, not 90%. 95%+.
  • Clean data — no duplicates, no junk entries, no corrupt records.
  • Audit trail — every field traceable to its source (API, log, config).
  • No data loss — if you drop a record, you know why + you've told SA.

For code (implementation)

  • No tech debt — ship production-ready code, not scaffolding.
  • 80%+ test coverage — unit, integration, and e2e tests.
  • Performance targets met — API <100ms, dashboard <2s load.
  • Zero production incidents w1 — if there are bugs, they're small + non-blocking.