Work with me
I measure whether AI tooling actually changed how your team delivers, using your own Git and CI data rather than survey answers.
Book a 30-minute call No agenda beyond how your team uses AI day to day, and where it bites.
The problem I get called about
Most teams have adoption numbers. Seats bought, suggestions accepted, an internal survey that says people feel faster. Almost nobody has a number on whether anything shipped sooner or broke less.
The reason is usually not the tool. The bottleneck moves rather than disappearing. Code gets written faster, review becomes the queue, cycle time barely changes, and the whole programme gets written off as hype.
What the audit is
- Read-only. I read your Git history and your CI runs. Nothing is installed, no agent goes near your repository, no production access is needed.
- Four numbers, before and after your rollout. Cycle time, rework rate, review latency, change-failure rate.
- The caveats travel with the numbers. Team size changed, a release process changed, the sample is too small to carry the claim. Where the data cannot answer the question, the report says so in the same sentence as the number.
- A written report and a 90-minute walkthrough with the people who own the pipeline, not just with the person who signed for it.
Scope and price
| Engagement | Length | Price | What it covers |
|---|---|---|---|
| Pilot | 5 days | €4,000 | One team, one repository, one before-and-after window. Discounted on the scope, not on the rate, in exchange for permission to publish the case study. |
| Standard | 8 days | €7,500 | Up to three teams, across repositories, plus the instrumentation plan for the numbers you are currently missing. |
| Re-measure | quarterly | €2,000 | The same four numbers by the same method, so the trend is comparable instead of a fresh opinion every quarter. |
Fixed scope, fixed price. Not billed by the hour.
Who it is for
Engineering organisations of roughly 20 to 200 people who have already rolled AI tooling out, and where somebody has started asking whether it paid for itself.
It is not for teams looking for help choosing a tool, and it is not a number to put in a board deck. If your data cannot support the claim you were hoping for, that is what the report will say.
What I have measured before
The same method, applied to a platform I built and ran for 20 months. Every number below is in the post it links to, with how it was counted.
- What 20 months of AI adoption measured
Per-engineer throughput doubled at flat cycle time, and change-failure rate doubled with it. Review time was the bottleneck first, not writing code.
- How to cut your change-failure rate without losing speed
Change-failure rate spiked to 37.5% in the worst month. About 35 deterministic guardrails went in, and it came back down while deploys more than doubled.
- Audit, then optimize
An OpenAI bill doubling weekly, and nobody could say which call was expensive. One table of call logs found 63% of calls were identical. Three fixes cut input tokens per result by 88.8%.
Longer engagements
I also take senior architecture and head-of-engineering contracts, remote, with EU companies. Distributed real-time systems, platform work, and engineering-culture rebuilds. Background and full stack on the about page.
How it starts
A 30-minute call. If an audit is not the right thing, I will say so on that call and it costs you nothing.
Not ready to talk? Read first.
New posts by email, about two a month. Measured results with the caveats attached, which is the same thing the audit produces.
Almost there. Check your inbox and click the link to confirm.
That did not go through. Try again, or email [email protected] and I will add you by hand.
No tracking pixels. I never pass the address on. How this is handled.