Everyone’s telling you to use AI. Almost nobody is telling you what for.

If you run a small business, you have been told at least a dozen times in the past year that you should be using AI. Usually by someone with something to sell, and almost never with a specific, concrete answer to the obvious follow-up question: for what, exactly?

So let’s try to answer that honestly, including the parts that don’t work.

Where it actually earns its keep

The pattern we see is consistent. AI helps most with tasks that are repetitive, language-shaped, and low-stakes if slightly wrong. It helps least with anything requiring judgement, accountability, or knowledge of your specific situation.

That gives you a reasonable filter.

First drafts of anything written. Quote cover letters, service descriptions, follow-up emails, social posts. The draft is rarely good enough to send, but starting from something and editing is considerably faster than starting from a blank page. This is the most common genuinely useful application.

Summarising and reformatting. Long email chains into a summary. Meeting notes into actions. A pile of customer feedback into themes. Turning one piece of content into three formats.

Getting unstuck. Not producing the final answer, but producing ten options when you can’t think of one. Naming things, structuring a proposal, working out what questions to ask a supplier.

Routine admin language. Standard responses to common enquiries, chased invoices, appointment confirmations. Things you write often and slightly differently every time.

Where it doesn’t

Anything where being confidently wrong is expensive. Pricing, legal wording, tax treatment, technical specifications, compliance. These tools produce incorrect information in exactly the same tone as correct information, which is the whole problem.

Anything requiring knowledge of your business it doesn’t have. It doesn’t know your margins, your supplier relationships, or why you stopped taking that kind of job in 2023.

Final client-facing work, unedited. People can tell. Not always, but often enough, and the cost of being caught sending something generic to a client who expected you to think about them is higher than the time you saved.

Anything you can’t be bothered checking. If the review takes as long as writing it would have, the tool hasn’t helped.

The part people skip

This is the bit that gets glossed over in most of the hype, and it’s the bit that can actually cause damage.

Be careful what you put in. Client details, financial records, contracts, staff information, anything covered by privacy obligations, anything commercially sensitive. Free consumer tools and paid business tools often have quite different terms about what happens to the data you enter, and the only way to know is to read the terms for the specific tool you’re using. Assuming is not a strategy.

Check the output before it reaches anyone. Decide who does that, and what they’re checking for. A vague intention to “have a look over it” reliably becomes nothing.

Bring your staff with you. If the first your team hears about AI is a new tool appearing in their workflow, a reasonable number of them will assume it’s the first step in replacing them. That’s not paranoia, it’s a sensible reading of the news they’ve been consuming. Being direct about what it’s for and what it isn’t for costs you one conversation and saves a lot of quiet resistance.

A sensible way to start

Pick one task. One. Something small, repetitive, and low-risk, that you do at least weekly.

Use it for that task for a fortnight. Actually time it. At the end, ask whether it saved you anything real, or whether you spent the saved time fixing the output.

If it worked, add a second task. If it didn’t, you’ve lost two weeks of small experiments and learned something useful. Either outcome beats buying a subscription to something because a podcast told you to.

The businesses getting value out of this are not the ones who adopted the most tools. They’re the ones who found two or three things it does well for them and stopped there.

Hear it from people who’ve tried it

We’re running AI at Work: What Small Businesses Are Actually Using It For as part of NSW Small Business Month 2026.

It’s a free 30-minute panel. Small business operators talking honestly about where AI has saved them time, where it wasted their time, and what they’d do differently starting again. Fast, specific, and no hedging.

It’s vendor-neutral. Nobody is recommending a product and nobody is selling anything.

Wednesday 28 October 2026, 12:00pm to 12:30pm AEDT. Online. Free.

Register for AI at Work — free, Wednesday 28 October