Why AI projects stall in small businesses, and the fix is smaller than you think

Most AI projects in small businesses don't fail. They stall. And they nearly always stall for the same boring reason: they were scoped like a big IT project when they should have been scoped like a single job. The good news in that is the fix is smaller, cheaper and faster than the version most owners are dreading.

Pilot purgatory is a scoping disease, not a tech problem

There is even a name for the stall: pilot purgatory. A project gets started, shows a flicker of promise, and then just sits there, never quite finished, never quite killed.

This isn't only a small-business affliction. A widely-read 2025 report from MIT's NANDA initiative, "The GenAI Divide: State of AI in Business 2025," found that around 95% of enterprise AI pilots were delivering little to no measurable impact on the bottom line, while only about 5% were driving real value. These were big companies with budgets and staff to throw at it. What the report pinned the failures on is the important part: not the quality of the AI, but how clumsily it was bolted into real workflows. The tools mostly worked. The projects around them didn't.

Sit with what that means for a smaller business. If organisations with whole teams and deep pockets stall at that rate, a small business scoping its AI the same way, big and broad, has even less room to absorb the stall. The lesson isn't "AI doesn't work for us." It is "don't scope it like they did."

Big scope is the trap

Here is how it usually goes wrong. The project starts too big. "Let's automate our operations." "Let's do a proper AI rollout." Scoped like that, it needs months, a budget, and someone senior to own it. So it quietly slides down the list behind the actual running of the business, and it dies there. Not with a decision, just with neglect.

The size is the problem, not the ambition. A goal that needs months before it shows anything is a goal that competes with everything else on your plate, and loses.

Start small enough to actually finish

The businesses that get real value do the opposite. They pick one specific, painful, repetitive task, the kind you sigh about every week, and they get that one thing working end to end in a couple of weeks. Small enough to finish. Real enough to matter.

Not "transform how we operate." Something more like: every quote that comes in gets logged, acknowledged and filed without anyone touching it. One job. Narrow edges. A result you can see by the end of the fortnight.

The first win is the real product

That first win does something a big plan never does. It builds trust. Once the people in your business see one boring job just handled, quietly, every day, without drama, the second one stops feeling like a leap. Momentum comes from a real result, not from a roadmap.

That is deliberately how we work at Handiwork: one narrow useful thing, quickly, before anyone says the word transformation. If you want a sense of the kinds of jobs that make good first targets, our use cases are a practical place to look.

Name the one task

So don't plan the whole thing. Name the one task you would most love to never do again. That is your first AI project. Not the whole business, not a platform, not a rollout. One task, working in weeks, not months. Get that done, and the next one will be easy.

Ready to find out where you stand?

If you want help picking that first task, the free AI Readiness Check is a good place to start. It takes about five minutes and there is no pitch at the end of it.