A marketing team on one tool. A bid team on another. And a handful of good people quietly expensing their own, feeding client and project data into systems nobody is governing. That isn't innovation. It's exposure — and it is already on your risk register, whether or not anyone has written it down.
Left alone, AI transformation happens by department, not by design.
Talent builds an AI approach with the tools inside its recruitment platform. Marketing builds a different one inside its content stack. Design and cost end up running on different information — the thing you spend your whole career trying to stop.
Everyone works hard, and everyone diverges. Inconsistent language, duplicated spend, contradictory outputs, no shared standard for what may and may not be put into a model. Meanwhile the curious ones — usually your best people — go and buy their own.
That is a coordination problem before it is a technology problem. And a coordination problem is exactly what a project manager solves: get the experts in a room early, surface the crossovers at design stage, build the standard in. You don't fix a clash in the product. You catch it at Stage 2.
The sector has done this before, badly. Property spent fifteen years arguing about BIM while the transition happened around it. This one is moving considerably faster.
I work as a fractional AI and change lead — in built-environment terms, a specialist you bring onto the programme two or three days a week, the way you'd bring in a planning consultant or a principal designer, rather than a permanent hire. I have no software to sell and no vendor relationships. What I have is twenty years of running programmes across experts who don't report to me.
What is actually happening in your firm right now, written down, with the exposure quantified.
The pilots stop being pilots. One owner, one standard, one set of decision rights.
Built so you're self-sufficient. If you still need me in two years, I've done it wrong.
Governance on paper is easy. Governance people follow is a different job.
Most AI policies fail the same way: they're written, circulated, and quietly ignored by the people under the most delivery pressure. You end up with a document that satisfies an auditor and changes nothing.
What closes that gap is the thing I've spent twenty years doing — building teams that will actually tell you what isn't working. If your people believe the honest answer costs them their job, you will never get accurate adoption data, and every control you write will be theatre.
So the programme names what's protected, gives people back the hours they currently spend proving they did the job, and measures what changed. The teams get better. That's the upside — it isn't the reason you'd call me.
You'll get a written picture of what's already happening inside your firm, what it exposes you to, and what to do about it in what order — in a form you can put in front of a board or an insurer.
If you do nothing, the most likely outcome isn't a dramatic breach. It's quieter than that: eighteen months from now you're paying for four overlapping tools, your people have taught three different models with your clients' project data, and a professional indemnity renewal asks a question you can't evidence an answer to.