Who needs this
- A CTO/Head of Engineering whose AI investments have been running for 12+ months with no measurable result.
- The CFO/CEO expects an answer to “AI ROI in Q…,” and you need a defensible map.
- A team of 30-150 engineers.
- At least 1-2 teams already use AI with at least part of the engineers.
Who does not need this
- Teams under 10 engineers: a 1-week mini-discovery is enough.
- Companies without AI-tool experience: they need enablement, not discovery.
- Companies looking for validation of decisions already made, not an honest view.
What you get at the end
| Artifact | What is inside |
|---|---|
| 1-page Executive Summary | Bottleneck, hypothesis, recommended pilot, budget range - for CFO/CEO |
| Bottleneck Map | Where lead time is lost: scope / build / review / deploy - with senior hours per week |
| Discovery report | Full analysis: AI engineering, governance gaps, hiring risks, technical debt |
| 30/60/90 Roadmap | What changes in the operating model by week, with owner and metric |
| Pilot Scope | One boundary with exit criteria: what we start, what we do not touch |
| Measurement Plan | 3-5 metrics, baseline, target, how to measure without LoC deception |
| Risk map | AI Act, GDPR, IP risks tied to the timeline |
| Executive briefing | 1-1.5 hours with CTO + CFO/CEO, free-form questions |
The main artifact is a decision: kill / rescue / rollout. Discovery is not a report of “what exists,” but an answer to “what to do in the next 90 days.”
Structure of the two weeks
Week 1: discovery
- Day 1-2. Kickoff, access, collection of background materials.
- Day 3-5. Interviews: CTO, EM, 2 tech leads, 2 seniors, 1 review-bottleneck owner.
Week 2: analysis + roadmap
- Day 6-7. Audit of review process and CI/CD on 2 teams.
- Day 8. Governance and compliance review:
agents.md, audit log, AI Act. - Day 9-10. Assemble report and roadmap.
- Day 11. Executive briefing.
Where the evidence in the report comes from
- Lead time - from git history and issue tracker.
- Review time - from PR metadata.
- Change-failure rate - from incident tracker + production deploys.
- AI usage - from tool logs + interviews with engineers.
- Governance gaps - from repository analysis:
agents.md, security policies, audit log.
What I do not do in discovery
- I do not implement tools. This is diagnosis, not delivery.
- I do not write specifications and product decisions for you.
- I do not make nice-to-have recommendations. Only what must be done in the next 90 days.
- I do not give a “you are doing great” verdict if the facts say otherwise.
Typical outcomes I expect to see
- About 40% of companies - “fix processes first.” AI tools wait 60-90 days.
- About 35% - “ready for a pilot on 1-2 teams.”
- About 15% - “too early for broad rollout, hiring + governance first.”
- About 10% - “current AI investments are pointed in the wrong direction, pause recommended.”
Disclaimer: this is my hypothesis for 30-150 engineer teams after a year of AI rollout, not statistics. Your discovery may produce a different ratio.
Cost and format
- Fixed fee for the full format.
- No hourly billing.
- No hidden costs: all meetings, report, and briefing are included.
- If we later agree on AI-rollout implementation, the discovery amount is credited toward future work.
How to start
- 20-minute call where we check whether the format fits.
- Signing your NDA and statement of work.
- Kickoff within 1-2 weeks from the start.
FAQ
What if after discovery I decide not to implement?
Discovery is a standalone result. A recommendation to "pause" is a valid outcome. There are no direct or indirect obligations to buy further services.
Can the format be stretched to 3 weeks?
Yes, if the team is larger than 100 engineers or more than 2 teams need detailed coverage. Discussed during scoping.
How is this different from a regular audit?
An audit looks inward: what exists. Discovery looks forward: what to do in the next 90 days. The difference in the output artifact is a roadmap with owners and metrics, not a list of findings.
NDA and confidentiality?
A standard NDA is signed before any repository access. All artifacts are stored in your infrastructure or an isolated project space of your choice.