Somebody has already pitched you an AI tool this year. Maybe it answers the phone. Maybe it writes the estimate. Maybe it reads the tech's notes and drafts the invoice. The demo looked good, the rep was friendly, and the question on your desk is simple: do I buy it?
Here is how I answer that question for a shop, and how you can answer most of it yourself before anyone sends you a contract.
Start with the job, then the tool
Every tool pitch starts from the tool. The rep shows you what it does and asks you to picture it in your office. That is backwards for a shop your size, because the tool only earns its keep at one place: where a job moves from one person to the next.
So start there. Pick one ordinary job from last week. A service call, a replacement, a small commercial repair. Write down, in a paragraph, every person who touched it and how it reached them.
"A call comes in, Dana writes it up, Mike quotes it from the truck, if they say yes Dana puts it in the schedule, the tech does it and texts Dana, Dana invoices, Sarah does the books on Fridays."
That paragraph is a job journey. It names the texts. The texts are where the hours go, because every text is something somebody reads and then types again somewhere else.
Now hold the tool up against that paragraph. Which of those handoffs does it touch? If the answer is "none of them", the demo was about a different shop.
Four answers, and only one of them is buy
When I finish mapping a shop, every leak I find gets one of four verdicts in the blueprint. They are worth knowing before you buy anything, because they are the options the rep will never put on the table.
Configure. The software you already pay for can do it, and nobody turned it on. This is the most common finding and the least exciting one. The field service product has a mobile form. The accounting package can take the invoice straight from it. Somebody set it up in a hurry three years ago and the office has typed around the gap ever since. The fix is a settings page and an afternoon of training.
Connect. Two tools each do their half well and nothing passes the job between them, so a person does. The fix is a connection, and it is usually small: when this field changes over here, write it over there. This is where a lot of what gets sold as AI actually lives, and it often needs no AI at all.
Build. There is a real gap no product fills for the way your shop runs, and it costs enough hours to pay for closing it. This is the one answer where a custom tool or an AI piece makes sense. It is also the rarest of the four, and it should be, because it is the one you maintain.
Leave it alone. The leak is real, and it is small, or it only happens in March, or the person who handles it is fast and happy doing it. Writing that down matters as much as the other three. It is the answer that keeps you from paying to fix something that was fine.
A tool pitch only ever offers the third answer. The shop usually needs the first two.
Count the hours before the dollars
The question "is it worth it?" gets answered in dollars too early. Dollars are easy to argue about and hard to check.
Hours are easier. Ask each person in the office what their typical week looks like, in rough percentages. Payroll, reconciling the card transactions, internal meetings, re-typing the tech's notes, chasing a signed estimate. You are looking for order of magnitude here, and a guess is fine. Once you know which handoffs eat the most hours, you can turn hours into dollars with one loaded rate per role. That is the whole calculation. It does not need your P&L.
Then go back to the tool. Does it touch the handoff that eats the most hours? If it touches one that eats very few, it can work perfectly and still change nothing you would notice.
Three questions to ask the rep
If you are still weighing a specific product, these three get honest answers quickly:
- Where does the job come from, and where does it go next? You want a named field in your current system on both ends. "It integrates with everything" answers a different question.
- Who in my office types less after this goes in? A name and a task. If the answer is a new screen someone has to check, the tool has added a step.
- What happens the week the tech doesn't use it? Every tool depends on the field filling it in. If the office can't tell the difference between "no data" and "no work", you'll find out at month end.
When to get help
You can do everything above yourself, and for a lot of shops it's enough. The paragraph, the hours, the four verdicts. It takes an afternoon.
When it isn't enough, it is usually because the job lives in people's heads and nobody can write the paragraph without arguing about it. That disagreement is a finding, and it's the reason the AI Audit exists. I map how a job actually moves through the shop, interview the people who do the work, and hand you a blueprint that ranks every leak with one of those four verdicts. You keep the blueprint. Building anything is a separate decision, made after, with the hours in front of you.
If you want to see what I'd ask for first, the audit checklist is ungated. If you want to talk it through, book a call. If it's not a fit, I'll tell you.