I’ve written a piece for The AI Journal about the first time I rolled AI out across my own construction company.
It went wrong. One week the AI built a full set of subcontractor enquiries off a drawing that was two revisions old. The team caught it before anything went out, and then they stopped using it. Can’t blame them.
Looking back, four things went wrong and not one of them was the AI. It started without an induction, the one thing nobody on a building site starts without.
The article covers the four things I should have sorted first, what it took to get the team to try again, and what the AI did once it had them.