One common reason AI pilots stall is simple: the money is split into too many small amounts. Many companies rank their AI ideas and then give a small budget to as many as they can. Each pilot gets a little money and a little attention, which is rarely enough to make the new way of working part of a team's daily routine. In our view the missing step is choosing a few and funding them properly.
Why a ranked list is not a decision
Ranking has real benefits. It makes a long list easier to manage, a score for value and effort looks objective, and nobody's idea is turned down, which keeps everyone happy. Trying several small things at the start is also sensible, because nobody knows yet which ideas will work. The trouble is that a ranking rejects nothing. It puts ideas in order, and every idea on it still expects some of the budget.
What too little money does to a pilot
A small pilot is easy to start and hard to finish. It rarely gets the people, data or tools it needs, and nobody is held accountable for whether it works. In a 2025 report, McKinsey estimated that about nine in ten generative AI projects built for a single business function were still stuck at the pilot stage. Many things cause that. In our view one cause is that no one chose these projects over other uses of the money, so no one is responsible for making them succeed.
What choosing costs
Funding one AI project properly usually means another gets less, a team drops something it had planned, or money moves from one department to another. Having to cut something shows people the choice is serious, and it frees up the people and attention the chosen project needs. A plan that funds every idea avoids that cost. Everyone can agree to it, and nothing else in the company has to change.
The uncomfortable question
Pick any AI project on your list. What was cut so it could be funded? If the honest answer is nothing, it is not a priority, however well it scored.