Why a safety-assurance engineer is the most sceptical AI consultant you'll meet.
Before AI Business Partners, I spent years in rail systems assurance — the discipline whose entire job is standing between an impressive-sounding engineering claim and the people who have to live with it. In that world, "trust me, it works" is not an answer. Every claim needs evidence. Every change needs assessment. Every risk needs an owner. Trains run on that discipline, and people go home safe because of it.
Then I watched the AI industry sell to professional-services firms — and saw the exact opposite. Demos instead of evidence. Tools instead of outcomes. "Transformation roadmaps" that never named a dollar figure. Chatbots rolled out because the competition had one, then quietly abandoned when a partner caught the first confident hallucination in a client document.
The first number
The insight behind this business came from doing what assurance people do: measuring instead of believing. The first firm I looked at wasn't losing money to "missing AI strategy" — it was losing it to a quiet tax: senior fee-earners spending six-plus hours a week on document drafting a well-governed workflow could do in one. Costed honestly at partner rates, it was a six-figure annual leak that had never appeared on any P&L. Nobody had ever put the number on it — because the number lived between job descriptions, exactly where the unassessed model change lives in a rail project.
That became the method: find the number first. Not the tool, not the roadmap — the defensible dollar figure a partner can take to a partners' meeting. Only then decide whether AI earns its place.
What I'm against — and for
Against
- Hype-cycle consulting — selling the technology instead of the outcome
- Software-rollout thinking: buy licences, run training, hope
- "AI strategy" decks with no dollar figure and no owner
- Ungoverned tools drafting client work with nobody accountable
- Pilots that never end because nobody defined "working"
For
- Economics first — a measured, defensible number before any build
- One workflow at a time, live in 30 days, judged on results
- Partner judgement kept in the loop, always
- Client data that never needs to leave the firm
- Guarantees in writing — fee back if we miss, you keep the assets
Why professional-services firms
Because they're the mirror image of rail: high-stakes judgement work, regulated, reputation-carried, allergic to unproven claims — and right to be. A firm that bills by the hour deserves an adviser who treats an AI claim the way an assurance engineer treats a safety claim: show me the evidence, show me the governance, show me the number. That's the whole practice.
Talk to Joe directly.
Thirty minutes, partner-to-partner. Bring your biggest time-drain; leave with a number — whether or not we ever work together.
Request a conversation →