AI Accountability Diagnostic
A focused advisory engagement for engineering leaders building seriously with AI.
The problem
Teams are building with AI faster than ever (code, agents, AI-assisted everything), and the speed has outrun something. Most engineering organizations can no longer fully explain or defend what they've built. Nobody can quite say who is accountable when the AI is wrong, or whether a decision the system made today can still be defended in six months.
The pattern showing up across the industry: cheap construction has let organizations skip the hard conversation about what they actually need. Knowledge that used to live in senior engineers now sits in prompts and fine-tunes that nobody owns. Testing budgets shrink because the models look fluent. The governance work that would catch the difference is the first thing to get deferred. And then a quiet quarter passes, and the questions that were comfortable to defer become questions the organization can no longer answer.
What the diagnostic is
A focused 1–2 week engagement examining how your organization actually builds with AI-assisted tooling, and where the accountability, quality, and governance gaps are. The work is conducted directly with engineering leadership and senior engineers (the people closest to what's been shipped) and grounded in your real systems, not a generic checklist.
The deliverable is a clear written assessment plus a walkthrough. What's likely to break in the next 6–12 months, why, and what to do about it. Concrete enough to act on, structured enough to defend to a board.
The work draws on over two decades of testing rigor (since 2005), a multi-year track record of boring, well-governed releases at scale, and a personal habit of building the kind of AI-assisted infrastructure this work is about (see the Toolsmithing pages for what that looks like in practice).
Who it's for
Engineering leaders (CTOs, VPs of Engineering, Heads of Quality) at organizations building seriously with AI. The diagnostic is most useful when the AI work has moved past the demo stage and into production, and when "who's accountable when this is wrong" has started to feel like a real question rather than a hypothetical one.
Engagement & price
The diagnostic is delivered on-site at the client, with a written assessment and a leadership walkthrough within 1–2 weeks of kickoff. Travel inside Europe is included; engagements outside Europe are quoted separately for travel only.
The work behind the offer
- Error Budgets, Not Validators on matching AI architecture to failure severity.
- Local-First AI Systems, the personal stack: six interconnected systems, 19 services, built and tested like production.
- ettool, a tester-driven AI tool built on "AI should assist, not decide."
- Blog for ongoing writing on AI accountability and testing in the AI era.
Start a conversation
The next step is a short call to see whether the diagnostic is the right fit for what you're seeing inside your engineering organization.
Email Kristoffer about the Diagnostic