Five Questions to Ask Before the Next AI Investment
Across seven investment committees I have taken part in more than a hundred investment decisions, and as founder-CEO of an AI venture platform my team and I evaluated over two thousand five hundred ventures. That volume does not make anyone prescient. What it does produce is pattern recognition — and one pattern dominates: the proposals that disappoint are rarely the ones with weak technology. They are the ones with unclear ownership of the outcome.
Here are five questions I have found separate a serious AI proposal from an expensive one. They are deliberately unsophisticated. Boards do not need to interrogate architecture; they need to interrogate assumptions.
1. Which specific decision gets better, and who owns that decision today? "Improving customer experience" is not a decision. "Reducing the time a credit officer spends on file preparation from four hours to one" is. If the paper cannot name the decision and its current owner, the value has not been located yet — and neither has the accountability.
2. What does the person on the other side of this do differently on Monday? AI creates value through changed behaviour, not deployed software. If nobody's Monday changes, the investment produces a cost and a press release. This question also surfaces the training, process and incentive work that proposals routinely underestimate.
3. How will we know it worked — and would that measure convince a sceptic? Adoption rates and usage statistics are activity, not outcome. The right measure is the business result the system was meant to move, tracked against a baseline established before launch. If the baseline does not exist yet, the honest answer is that we will not know.
4. What happens when it is wrong, and who finds out first? Every model is wrong sometimes. The governance question is whether the organisation detects the error, how quickly, and whether the person affected can get an explanation. A system whose failures are invisible is not lower-risk; it is higher-risk with better optics.
5. What are we not doing because we are doing this? The most underrated question at any investment committee. AI proposals compete for the same management attention as everything else, and management attention — not capital — is usually the binding constraint.
None of these require technical expertise. They require the same scepticism a board applies to any capital request, held steady against the unusual amount of enthusiasm that accompanies this particular category.
One closing observation. In my experience the strongest proposals are almost always narrower than the weakest ones. A team that has thought carefully arrives with one decision, one measure, one owner. A team that has not arrives with a platform strategy. Boards can learn a great deal from the width of the ask.



Comments