Somebody has probably told you that AI is about to change your business. A vendor, a nephew, a guy at the supply house. Maybe all three.
Here is the most useful thing I can tell you up front, and I sell AI advice for a living: most small shops I talk to should not buy anything yet. Not because AI does not work. It works fine. But the math has to work first, and for a lot of businesses it does not.
You can run that math yourself in about ten minutes. Three questions.
1. Where do the hours actually go?
AI is good at a narrow kind of work: tasks that repeat, follow rules, and mostly involve reading or writing. Quotes. Follow-up texts. Chasing invoices. Turning a phone call into a work order. Summarizing paperwork you have to touch but do not want to read.
It is bad at judgment calls, one-off problems, and anything that requires hands. Nobody is automating the actual pipe fitting, and anyone who implies otherwise is selling something.
So take one normal week and write down every task that repeats. Not the big dramatic problems, the boring ones. The quote you rebuilt from scratch for the fourth time. The same three questions you answer on every call. The invoice reminders you send at 9pm.
If nothing on that list repeats at least weekly, stop here. You do not have an AI problem, and you just saved yourself real money.
2. What is that time worth in owner dollars?
Now put a number on it. Take the repeating task, count the hours it eats in a week, and multiply by what your time is actually worth. Not minimum wage. What an hour of your attention earns when it is pointed at billable work or at growing the business.
Say quotes take you two hours each and you do five a week. That is ten hours. If your time is worth $85 an hour, that task costs you around $44,000 a year. That is a real problem, and it might justify real tooling.
Now run the same math on something small. A task that takes twenty minutes a month is a rounding error. It does not matter how impressive the demo looks. Nothing that costs thousands to build is worth buying to recover pocket change.
The size of that number is the size of your problem. Write it down. You will want it for question three.
3. What is the cheaper fix?
This is the question the people selling you AI skip, because the honest answer often does not pay them.
Fixes come in an order, cheapest first:
- Fix the process. Missed calls are not an AI problem, they are a phone problem. A price list that lives in your head is not an AI problem either.
- Use a spreadsheet. A quoting template with your real prices in it costs one afternoon and beats a surprising amount of software.
- Buy something off the shelf. A lot of repeating problems already have a tool that costs $20 to $50 a month. Scheduling, review requests, invoice reminders. Boring, proven, cheap.
- Build something custom. Last resort, not first. Custom makes sense when the problem is big (see your number from question two), specific to how your shop runs, and not solved by anything you can rent.
Anyone who starts you at custom without walking the first three steps is selling, not advising.
The ten-minute answer
That is the whole exercise. Find the repeating work, price the hours, rule out the cheaper fix. If the number is big and nothing cheap solves it, AI might genuinely be worth your money, and it is worth doing right. If the number is small, you get to ignore the hype with a clear conscience. Most shops land there, and that is a fine place to land.
If you want a second opinion on your answers, that is what the free call is for. Fifteen minutes, one problem, and if the answer is no, I will say so on the call.