I have now spoken to several hundred business owners about their experience with AI tools. The pattern of disappointment is remarkably consistent. They read about a tool. They saw a demonstration. The demonstration was impressive. They subscribed. They used it three or four times with genuine enthusiasm. Then they used it less and less, until it became one of six monthly subscriptions that they vaguely intend to cancel but haven't got around to yet.

When I ask what went wrong, the answers cluster around a theme that people rarely articulate directly but which is consistently present once you listen for it. They were impressed by the tool's capability without having first identified the problem they needed it to solve. The tool could do many things. They needed it to do one specific thing that would genuinely improve their business. They had not defined what that thing was before they bought it, and so they could not evaluate whether the tool actually did it.

This is not a technology problem. It is a procurement problem dressed up as a technology problem. And it happens with AI tools at a rate that suggests something specific about how AI is marketed, and how business owners are making decisions in response to that marketing.