Insights
Should my business be using AI? A plain answer.
Probably — but not for the reason the pitches give. The useful question was never "should I use AI"; it's "is there a problem in my business worth solving, and is AI the cheapest way to solve it."
Most small businesses already touch AI somewhere — a scheduling feature, something on the phone line, a tool inside software you already pay for. So the adopt-or-fall-behind framing is mostly a way to sell more software. You don't fall behind by skipping a tool. You fall behind by leaving the same money on the table every month — the calls that ring out, the estimates nobody chases, the invoices that sit — while somebody down the street quietly stops leaving it.
That's the answer. The rest of this piece is the part the answer usually skips: how to actually tell whether your business has a problem AI is the cheapest fix for, and how to do that sorting yourself before anyone sells you anything.
Where is AI already in my business?
Start by taking inventory, because "adopting AI" usually means noticing what you already have. Check four places: your phone system (many now transcribe voicemails or offer auto-replies), your scheduling or booking software (reminder logic, smart slotting), your accounting package (categorization, invoice chasing features you may never have turned on), and your website or listings platform (chat widgets, auto-responses). Spend an hour flipping through the settings screens of software you already pay for. It's common to find real capability sitting there switched off — capacity you've been paying for and not using.
That hour matters for a second reason: it recalibrates the pitches. Once you know what your current stack can do, "revolutionary AI feature" often translates to "the thing your CRM already includes."
How do I tell a real problem from a manufactured one?
A real problem has a number attached, or can get one in an afternoon. Count last month: how many calls came in after hours or while everyone was on a job, and how many of those callers you ever heard from again. How many estimates went out, and how many got a single follow-up. How many invoices sat past thirty days. How many hours you personally spent on paperwork someone — or something — else could have handled. That's the whole method: find the leak before you shop for the plug.
A manufactured problem arrives from the outside in: a pitch tells you competitors are "leveraging AI," and the discomfort is about falling behind in general rather than any line on your P&L. The test is simple. If you can't say what the problem costs you monthly — even roughly — you don't have a problem yet. You have a mood, and moods make expensive software decisions.
When is AI the cheapest way to solve it?
Sometimes it isn't, and it's worth being plain about that. If your phone rings out because you're understaffed on the counter at lunch, the fix might be a schedule change, not a subscription. If estimates die because your pricing is high for your market, follow-up automation will just collect the no's faster. Process problems, pricing problems, and people problems all wear the same costume — "we're losing jobs" — and AI only fixes the first kind reliably.
Where AI genuinely is the cheap fix: work that's repetitive, rule-bound, and time-sensitive, happening in the gaps between your systems and your people. Answering and qualifying the after-hours call. Chasing the quiet estimate on day three and day seven. Nudging the unpaid invoice. Reminding the no-show. These are cheap to automate because they're boring — the patterns are stable, so the build is straightforward and the arithmetic is easy to run.
What would a sensible first step look like?
Not a purchase. A week of counting. Track the four numbers above for seven working days — misses, unchased estimates, aging invoices, your own paperwork hours. Put rough dollar figures on each using your own average ticket and your own sense of close rates. Rank them.
Then, and only then, look at fixes — and look with the wariness the counting just bought you. You now know what the leak costs, so any quoted price has something to be compared against. That single move — leak first, price second — filters out most of the bad decisions available in this category, in both directions: it stops the panicked yes to a slick demo, and it stops the reflexive no that quietly costs more every month than the tool would have.
What about "falling behind"?
Here's the part that doesn't make the pitch: the technology is a commodity. Anyone can buy the same tools you can, and better versions show up every quarter. What isn't a commodity is knowing which problem in your specific business is worth pointing them at — and which ones aren't worth touching at any price.
That cuts both ways, and it's worth sitting with. Your competitor buying AI tools isn't the threat. Your competitor plugging the same leak you both have is the threat — and the advantage goes to whoever finds their leak first, not to whoever subscribes first. Adoption is not the race. Diagnosis is the race.
So: should your business be using AI? If a leak is costing you real money and an automated fix clearly costs less, yes — and it won't feel like a leap of faith. If nobody's found that leak yet, that's the work to do first. It's a judgment call about your business, not a technology decision.
Everything you need to run that diagnosis yourself is above, and it's a good week's homework. The part that's harder to do alone is the ranking — which leak, which fix, in what order, for your shop specifically. That call is the one that pays, and it's the work I do. If you'd rather make it with someone who does this daily, book a conversation — bring your numbers and I'll tell you straight what I'd build, and whether I'd build anything at all.