What AI can actually do for a small business right now

Not the sci-fi version. Five jobs it handles well today, three it's still bad at, and where to start without wasting money finding out yourself.

Every week somebody asks me some version of the same question: is AI actually useful for a business like mine, or is it all hype? Fair question. Most of what gets written about AI is either breathless or terrified, and almost none of it is about a five-truck plumbing company in Brandon.

So here's the plain version. Five things AI does genuinely well for a small business today, and three it's still bad at.

What it's good at

1. Answering people when you can't

This is the big one, and it's not really about intelligence — it's about being awake. A lead fills out your form at 11:40pm. You're asleep. AI reads what they wrote, texts back something that actually addresses their problem, asks the two questions you'd ask, and hands you the whole thing in the morning.

Most jobs go to whoever answers first, not whoever is best. That's not a theory, it's just how people hire. If you're the only one who can answer, you're losing to whoever automated it.

2. Writing the first draft of anything

Quotes, proposals, follow-up emails, the description on your Google profile, the page about a service you've been meaning to add for two years. AI won't write it the way you would. It will write something 80% right in nine seconds, which turns a task you avoid for a month into ten minutes of editing.

3. Turning one thing into six things

You finish a job. You take four photos and record a thirty-second voice memo about what you did. That becomes a social post, a caption, a short video, a blog post, and a before-and-after for your Google profile.

You already did the work. The content problem was never a shortage of material — it was that repackaging it takes an hour you don't have.

4. Reading things you don't have time to read

A stack of invoices. Two hundred customer reviews. Six months of job notes. AI is genuinely good at reading a pile of unstructured mess and telling you what's in it — which complaint keeps repeating, which service makes the most margin, which month you're always slow.

5. Being the thing between you and a spreadsheet

Instead of your office manager filtering columns, she asks "which customers haven't been back in six months and live in South Tampa" and gets an answer. This is where AI has quietly gotten very good, and it's the least talked-about use.

What it's still bad at

Anything where being wrong is expensive

AI will state something incorrect with total confidence. For a first-draft email, who cares. For a price you're quoting a customer, or a legal term in a contract, or a medication dosage — a human checks it or you don't do it. I build everything with a human approval step for exactly this reason.

Knowing your business

It doesn't know that you don't take jobs south of Gandy, or that a particular customer always pays late, or that you charge more for two-story houses. It only knows what you tell it. Getting that context in is most of the work, and it's most of what I actually do.

Replacing the reason people hire you

Nobody in South Tampa hires a mobile mechanic because his emails are well written. They hire him because he showed up when he said he would. AI can make sure you never miss a call. It can't be the guy who shows up.

Where to start

Pick the one thing that costs you the most and is the most repetitive. For most trades that's lead response. For most offices it's a report somebody rebuilds by hand every week. For most restaurants it's posting.

Don't buy a platform. Fix one thing, see if it holds up for a month, then fix the next one.