Most articles about AI and small business fall into one of two camps. Either it’s all breathless hype, or it’s a list of forty tools nobody has time to evaluate. Neither helps much when you’re trying to get quotes out before Friday.
So let me take a different angle. Here’s what AI actually looks like inside a small business on a normal Tuesday — the unglamorous stuff that quietly saves an hour here and there.
I’ll say upfront: I’m not a true believer. I’ve watched plenty of businesses buy a subscription, use it for three weeks, then forget about it. But I’ve also seen it work, and the pattern of what works is fairly consistent.
Where most businesses actually start
Not with strategy. With admin.
The National AI Centre’s tracking of Australian SMEs found content generation and data analytics lead the way, with 54% of adopters using AI for each. QuickBooks’ Australian research points to similar ground — data processing, administration, marketing, customer service and bookkeeping.
That tracks with what I’ve seen. Nobody starts with an ambitious AI transformation. They start because they’re sick of writing the same email forty times.
The jobs AI is genuinely good at right now
Writing the boring things
Quotes. Follow-up emails. Job descriptions. Responses to Google reviews.
This is the single most common use, and honestly the most underrated. A tradie I know uses a chatbot to turn his scrappy phone notes into a tidy quote email. Takes him two minutes instead of twenty. He doesn’t think of it as “using AI.” He thinks of it as not doing paperwork at 9pm.
Meeting notes
If you run client calls, transcription tools like Otter or Fathom are close to a no-brainer. They record, transcribe and summarise in real time, and pull out action items so things don’t get missed.
The value isn’t really the transcript. It’s that you can actually listen to the person in front of you instead of scribbling.
Answering the same customer questions forever
Opening hours. Do you deliver to Geelong. Is there parking. Can I change my booking.
Chatbots and automated replies handle this well now. Modern systems understand what the customer is asking, work outside business hours, and hand over to a human when something gets complicated — with the conversation history intact.
That last part matters more than people expect. The old chatbots dumped you back at square one, which was infuriating. The current ones mostly don’t.
Moving information between systems
This one is less visible but often the biggest time saver. Tools like Zapier sit between your apps and shuffle data around so you don’t have to. They can summarise incoming information, sort support tickets, and kick off actions based on plain-language instructions.
Think: enquiry comes in through the website, gets logged in your CRM, triggers a reply, adds a task to your list. Nobody touches it.
A sector-by-sector snapshot
Different industries land on different uses. A few patterns worth noting:
- Trades and services — quoting, scheduling, chasing invoices
- Retail and e-commerce — product descriptions, FAQs, working out what to reorder
- Hospitality — booking enquiries, review replies, social posts
- Professional services — summarising documents, drafting correspondence, research
- Health — this one’s interesting. Australian GPs spend two-plus hours a day on documentation. Lyrebird Health, built in Melbourne, records the consult and writes the note straight into the patient record.
The Lyrebird example is the one I’d point to if someone says AI isn’t relevant to their industry. It solves one specific, deeply annoying problem. That’s the whole trick.
The relevance gap is real
Speaking of which — this stat from the National AI Centre stuck with me. More than half of non-adopting businesses said AI simply isn’t relevant to them. The Centre’s read is that this isn’t rejection so much as a lack of visible examples for businesses like theirs.
I think that’s exactly right. Adoption is under 30% in construction and agriculture, and above half in health and services. Not because builders are luddites. Because the marketing is all written for marketers.
What separates the businesses it works for
Here’s the uncomfortable bit. Trying AI and getting value from it are different things.
Deloitte’s research found two-thirds of Australian SMBs now use AI, but only 5 per cent are fully set up to get the benefits. Poking at a chatbot occasionally isn’t the same as changing how work gets done.
From what I’ve seen, the businesses that get somewhere do a few things differently.
They pick one workflow, not ten. The failure mode is almost always trying everything at once. One task, done properly, beats five half-finished experiments.
They measure against something. How long did this take before? If you don’t know, you’ll never be able to tell whether it helped.
More than one person knows how to use it. As one Australian consultant put it bluntly, adoption dies when only one person knows how to use the tool. That person goes on leave and the whole thing quietly stops.
They use a paid business tier for anything sensitive. Worth understanding why: business and enterprise plans generally exclude your data from model training, give you admin controls, and clarify where data is stored. Free consumer accounts don’t.
The part nobody wants to talk about
Governance. I know. Stay with me, because this one has teeth now.
Australian research found that about half of AI-using businesses check outputs before they reach customers, but telling customers AI is in use — and giving them a way to raise a concern — lags well behind.
You don’t need a policy document the size of a phone book. A two-page policy your team will actually read covers it: which tools are approved, what data must never be entered, that a human checks anything customer-facing, and who owns the questions.
If you’re in a regulated field, this has already moved from nice-to-have to required. The Tax Practitioners Board published guidance in July 2026 on how the Code of Professional Conduct applies when tax and BAS agents use AI. Others will follow.
A realistic starting point
If you’re at zero and want something concrete:
- Write down your five most repetitive tasks. Not the important ones. The repetitive ones.
- Pick the most boring. Seriously. Lowest risk, clearest benefit.
- Time how long it currently takes. Rough is fine.
- Try one tool for a month. One.
- Compare. If it saved real time, expand. If not, drop it and try the next task.
That’s it. It’s deliberately unambitious, which is the point.
AI hasn’t transformed most small businesses. What it has done is quietly remove a chunk of the admin that used to eat evenings and weekends. That’s less exciting than the pitch decks suggest, but it’s not nothing.
The gap I’d watch isn’t between businesses using AI and businesses not using it. It’s between businesses using it deliberately and businesses using it randomly. MYOB’s data on firms using AI growing faster is often quoted — but I’d read that as organised businesses adopting AI, rather than AI creating organised businesses. Correlation deserves a bit of caution here.
Start small. Measure something. Don’t buy the all-in-one platform in week one.
And if a tool isn’t saving you time after a month, it probably never will.
