Most AI projects in small businesses don't fail because the AI is bad. They fail before anyone opens the tool.

This is my opinion, not a study. If you want the research on why projects stall, with sources, read Di-Hy's piece: Why most AI projects fail, and what the ones that work do differently. What follows is the short version I'd use as an operator, the four things I'd check before spending a dollar.

1. Can two people describe the process the same way?

If the owner and the office manager describe follow-up differently, you don't have a process. You have habits. Automating habits gives you a faster version of inconsistent.

My test is one page. Who does what, in what order, and what counts as done. If it won't fit on one page, the project is fixing the process, and that's fine. It's the real work.

2. Is there one named owner?

A tool is software. A system is software plus an owner plus a routine plus a check. When a field changes in the CRM or a phone number gets updated, an unowned automation quietly stops, and nobody notices because nobody was assigned to notice.

I would not approve a project without a name next to it and a five-minute weekly look.

3. Is there a number, and a date to check it?

Write down two or three numbers before anything goes live, and what they are today. Response time. Booked jobs. Paid hours on the task. Then pick a date to look again.

Without that you get a story ("it feels like it's working") instead of a result. And be careful with hours saved. Time only turns into value if it goes somewhere you can measure. I wrote more on that in hours saved is not profit.

4. Who reads the output before a customer does?

AI is confident whether it's right or not. Anything that touches money or a customer gets a human review until the system has earned trust. After that, spot check it forever.

What I'd skip at the start

Content generators and website chatbots. They're visible, so they feel like progress. For most service businesses the money leaks between the lead arriving and the job getting booked. Start there. I explain how I'd rank the options in what to automate first.

The trade-off

Doing these four checks first feels slow. You will want to buy the tool and see what happens. I'd still do the checks, because the tool will happily run on top of a broken process and charge you monthly for it.

If you'd like a second read on whether your business passes these four, Di-Hy runs a free audit that goes through them.

FAQ

What is the most common reason small business AI projects fail?

An unclear process. If the team can't describe the work by hand, AI automates the inconsistency. Writing the process down first prevents most of it.

What should I check before buying an AI tool?

Whether the process is written down, who owns the result, which numbers should move and who reviews the output before it reaches a customer.


Sources: No outside statistics are cited here. This is opinion. For the sourced research, see Di-Hy's guide on why AI projects fail.