Insights
Five questions a board should ask before investing in AI
Not whether the technology works, but whether the work changes because of it. Five questions that separate a decision from an experiment.
Most AI proposals that reach a board are well argued on the technology and vague on the work. They say what will be built, with which supplier, at what price. They rarely say what will be different in practice.
Five questions bring that difference out. They are not technical, and that is the point: they can be judged with ordinary common sense.
1. Which three processes, and why those?
A proposal naming twenty opportunities is an inventory, not a prioritisation. Ask which three it will be and what happens to the rest. The answer “we will get to those later” means in practice that nobody has chosen.
2. Who will use this, and what do they do now?
Not which department, but which people, and what their day looks like today. If that answer stays general, nobody has spoken to them. And then the application will most likely end up beside their work rather than inside it.
3. What has to change in the work?
A concrete answer sounds like: this check disappears, this meeting gets different input, this step moves to someone else. A vague answer sounds like: employees are supported and can focus on higher-value work.
If nobody can name which action is different tomorrow, nothing will be different tomorrow.
4. Where does it stand now, and how do we know?
This is the question most often missing. Ask which figures are being recorded today, before anything is built, and on which points measurement will happen again after delivery. Without that starting point, the evaluation a year from now is a conversation about impressions.
5. Who is accountable for the use?
Not for the delivery — for the use. In many engagements the building party’s involvement ends at acceptance, and after that nobody owns the question of whether it works. If this question does not produce a name, the answer is: nobody.
What these questions have in common
Not one of them is about AI. They are about work, about decisions and about ownership. That is no accident: most AI investments do not founder on the technology but on the assumption that the rest follows by itself.
A proposal that answers these five is no guarantee. But a proposal that does not answer them is not an investment decision — it is an experiment with an invoice attached.