Where AI creates value, and where it quietly burns money

The question I get most from founders is not "can we use AI here?" It is almost always yes. The better question is "should we?" Value does not come from adding AI. It comes from putting it where the economics actually work.

AI pays off when three things line up

The cleanest wins share a pattern. Look for a task that is repetitive, tolerant of small errors, and expensive in human time. When all three are true, AI compounds quietly, day after day, and the return is obvious within weeks.

• Repetitive: it happens often enough that small savings add up.
• Error-tolerant: a wrong answer is cheap to catch and correct.
• Costly in time: people currently spend real hours on it.

Where it quietly burns money

The expensive mistakes look exciting and fail slowly. A flagship AI feature on a rare, high-stakes decision where one wrong answer is catastrophic. A clever assistant for a task people do twice a year. A model built to impress a board, not to serve a user. None of these show up as a line item called "waste," which is exactly why they persist.

The cost of the wrong AI project is rarely the bill. It is the quarter you spent not building the right one.

A five-minute test before you build

Before greenlighting anything, answer three questions honestly. What task, done how often, costing how much time today? What happens when the AI is wrong, and who catches it? And if this worked perfectly, what number in the business moves? If you cannot answer the third, you have a demo idea, not a value idea.

Start narrow, prove the number, then widen

The teams that get real return do not start with a platform. They pick one painful, high-frequency task, ship a narrow tool, and measure the hours or euros it moves. Once the number is proven, widening is easy and funded. That sequence is the whole strategy: value first, ambition second.

This is most of what I do as a fractional head of AI: help leaders separate the projects that will pay from the ones that will only impress, before the budget is spent.