For most of my career I studied a single uncomfortable question: why do technology projects fail? I wrote more than a thousand columns about it. The names of the technologies have changed, today the headlines are about AI, but the underlying reasons have barely moved.

Enterprise AI projects rarely fail because the model is not good enough. They fail because no one agreed on the problem, the data was never ready, or the organization could not absorb the change.

Three questions before you fund an AI initiative

First, what decision will this change? If you cannot name it, you are buying a science project. Second, who owns the outcome, not the model, the outcome? Third, what happens to the people whose work this touches? Answer those honestly and most of the failure modes disappear.

The technology is the easy part. It always was.