How should a small business actually start using AI?

Why most of the advice doesn't fit your business

Nearly everything written about AI is written for companies with an IT department. Main-street owners read it and reasonably conclude AI is for someone else. It isn't. A five-person shop benefits more than a five-hundred-person one, because every missed call is a bigger share of the week's revenue. But you have to come at it like an operator, not a hobbyist.

In my consulting work the same pattern shows up in almost every business we map. The marketing is fine. The craft is fine. The follow-through leaks. The phone rings during a job and nobody calls back. The estimate goes out and nobody nudges. That, not a robot receptionist, is where AI earns its keep first.

A sample, with real-shaped numbers

Picture a three-truck plumbing outfit. It misses six calls a week while the crew is under sinks. Two of those six would have booked, and the average job runs $400. That is $800 a week walking away, call it $40,000 a year, from a problem nobody can see because it never makes a sound. Now the fix is obvious and small: a system that answers the missed call by text in seconds and books the job. The leak had a dollar figure, so the tool has a job description.

The field sequence

1. Pick one leak. Missed calls, slow follow-up, no reviews, invisible online presence. One. Not five.

2. Measure it for two weeks. Count the missed calls. Count the estimates that never got a second touch. You cannot manage what you refuse to count, and you cannot know if AI helped without a before number.

3. Fix that one thing. A missed-call text-back. An automatic follow-up sequence. An agent that answers after hours and books the appointment. Small, boring, specific.

4. Keep score for thirty days. Compare against your before number. Real dollars, not vibes.

5. Expand or stop. If it paid for itself, add the next leak. If it didn't, you spent one month and one small bill finding out. Not a year and a five-figure retainer.

In the end there is doing, and there is trying. A thirty-day pilot with a scoreboard is doing. Five subscriptions and a hope is only trying.

What to watch out for

Be suspicious of anyone selling AI with income promises, pressure funnels, or a subscription before they have asked how your business actually runs. The tools are getting cheap and generic. What you are really buying is the configuration, the integration with how you work, and a human who stays accountable after launch. If a pitch skips straight past understanding your business, so should you.

Do it yourself, or done for you

Both are legitimate. If you have the appetite, pick one tool and genuinely learn it. If you don't (and most owners we sit with don't have the hours), have the system designed, built, and run for you, and judge it by the scoreboard. Either way the philosophy is the same one we use for everything in this practice: understand first, evaluate second, design the solution, then re-evaluate it.