Five Questions Every Board Should Ask Before Approving Another AI Investment
I've watched an eight-figure AI investment get approved on a deck that couldn't answer question two. Nobody in the room asked.
Save this. Bring it into your next AI investment review, and don't let the paper leave the room until it can answer all five.
First: what decision or process genuinely works differently once this ships? Not "improves." Differently.
Second: who, by name, owns "is our data good enough for this decision"? Not a team. A name.
Third: what happens, operationally, the first time the model is confidently wrong? If the answer is a shrug, the programme isn't ready.
Fourth: which existing process gets retired because of this? If the answer is none, you're adding cost, not removing it.
Fifth: who is accountable if it fails in production? A name, not a department. Departments don't answer to regulators. People do.
I've watched an eight-figure AI investment get approved on a deck that couldn't answer question two. Nobody in the room asked. That is not a technology failure. It's a governance failure wearing an AI costume.
If your last approved business case can't survive all five of these, it isn't ready for sign-off. It's ready for one more round of questions, and that round is cheaper now than the incident review will be later.
Nigel Guy
The Disruptive AI Enabler™
One useful disruption every week.
Get essays like this one before they publish anywhere else.
More writing
Human After AI
What Twenty-Five Years in Transformation Taught Me About AI
In 2003 I sat in a room where a sign-off took four signatures and eleven days. It took another six years to change it.
The Enablement
A Framework for Turning AI Ambition Into Execution
Seven letters. I built this because every failed AI programme I've seen fails in a predictable place.