Research · official statistics only
The ABS reports 12% of Australian businesses using AI in 2024–25, up from 1% two years earlier, and 11% among small and micro businesses. A tracker published by a Commonwealth department reports 33 to 40% for the same size bands. Both are official. Neither is wrong. The gap is the story.
There were 2,814,778 actively trading businesses in the Australian economy at 30 June 2026. Of those, 1,729,477 employ nobody at all — the single largest category of Australian business is one person. Another 689,600 employ between one and four people, and 232,912 employ between five and nineteen. ABS, Counts of Australian Businesses, including Entries and Exits, reference period July 2022 – June 2026, released 18 August 2026.
Businesses with fewer than twenty employees make up 97.3% of all Australian businesses. ASBFEO Small Business Data Portal, from ABS Counts of Australian Businesses, June 2025.
Those three sentences are worth sitting with before reading anything else about Australian small business, including the rest of this page. The typical Australian business is not a team. It is one person, or one person and a couple of others, and every claim about what "small businesses should be doing" has to survive contact with that fact.
Just over one in ten Australian businesses reported using artificial intelligence, compared with 1% in 2022–23. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026.
A jump from one per cent to twelve in two years is not a trend, it is a phase change, and it is the single most important statistic on this page. Whatever you think about the technology, the base rate for "does anybody actually use this" moved by an order of magnitude while most people were still deciding whether to have an opinion.
AI use rises steeply with size: 35% of businesses with 200 or more employees, 22% of those with 20 to 199, and 11% of small and micro businesses. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026.
The obvious reading is that size drives adoption, and the obvious reading is only half right. The ABS also splits by whether a business is innovation-active — meaning it was already changing something — and that split is more revealing than size.
Among businesses with fewer than five people, 18% of the innovation-active ones use AI against 6% of the rest — three times the adoption at identical size. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026.
Among businesses with five to nineteen people the gap is wider still: 19% of innovation-active businesses use AI against 4% of those that are not. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026. That is nearly five to one, between businesses of the same size, in the same economy, in the same year.
So the constraint is not headcount and it is not budget — a five-person business and a five-person business have the same of both. The difference is whether the business already had a habit of changing how it works. A business that has changed something recently changes something else. A business that has not, does not.
That is an uncomfortable finding for anyone selling automation, because it means the readiness question is not "can you afford it" but "when did you last change anything". We built the readiness scorecard around roughly that idea before we found this data, and the data is a better argument for it than anything we wrote.
| Innovation-active | Not innovation-active | |
|---|---|---|
| Micro — under 5 people | Partly 18% reported using AI. | No 6% reported using AI. Three times fewer, at identical size. |
| Small — 5 to 19 people | Partly 19% reported using AI. | No 4% reported using AI. Nearly five times fewer. |
| Medium — 20 to 199 | Yes 28% reported using AI. | No 13% reported using AI. |
| Large — 200 or more | Yes 37% reported using AI. | Partly 29% reported using AI — the gap nearly closes at the top. |
Here is the thing nobody puts side by side. A separate tracker published by the Commonwealth reports AI adoption at 33% for businesses with up to four employees and 40% for those with five to nineteen — roughly three times the ABS figure for the same period. National AI Centre, AI Adoption Tracker, Department of Industry, Science and Resources, January – March 2025. Collected by Fifth Quadrant, a private research firm, at around 400 survey responses a month, and not an ABS statistical collection. The definition of "use" differs.
Both are published by the Commonwealth. Both are describing Australian small businesses. They differ by a factor of three, and the difference is method rather than dishonesty: the ABS runs a statistical collection of around seven thousand businesses with a formal definition of use; the tracker is a monthly survey of around four hundred businesses run by a private research firm and published by a department.
Neither number is wrong. They are answers to slightly different questions, and the honest thing to do with them is publish both and say which you would rely on. We would rely on the ABS, because its method is documented and its sample is an order of magnitude larger — and we would still want to know why a tracker with a looser definition finds three times as many people saying yes.
What that gap probably measures, in practice, is the distance between "we use AI" meaning a system in production and "we use AI" meaning somebody in the office has ChatGPT open. Both are real. Only one of them changes what a business costs to run.
Under a third of Australian businesses received orders online — 32%, barely moved from 30% in 2021–22. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026. For all the talk of digital transformation, two thirds of Australian businesses still do not take an order online.
On the ABS digital intensity index, 59% of businesses with up to four people sat in the lowest "baseline" band and 1% of all businesses reached "advanced". ABS, Development of a Composite Indicator for Business Digital Intensity in Australia, 2021–22 data, article released 28 July 2023. The most recent release of this index. Four years old, and we say so rather than implying it is current.
One in five Australian businesses reported experiencing a cyber security incident. ABS, Characteristics of Australian Business, 2024–25 financial year, released 25 June 2026. That one belongs on this page rather than a security page, because it is the counterweight: connecting systems together increases what a single compromised account can reach, and who holds the keys is part of the design rather than an afterthought.
460,461 businesses started in 2025–26 and 375,331 stopped — a 16.9% entry rate against a 13.8% exit rate. ABS, Counts of Australian Businesses, including Entries and Exits, 2025–26, released 18 August 2026. A net gain of eighty-five thousand businesses in a year, and more than a third of a million stopping trading, is the churn any claim about "what Australian businesses do" is averaged over.
We searched the ABS, the ATO, the Small Business Ombudsman, Treasury, the Productivity Commission, business.gov.au and the Payment Times Reporting Regulator. These are the gaps, and they explain why so much of what you read in this market is unsourced.
Three things follow from the numbers above, and none of them is "buy automation".
First, the base rate has moved. At one per cent, doing nothing was the reasonable default and everyone doing something was early. At twelve per cent — and rising steeply — doing nothing is a position rather than an absence of one, and it is worth being able to say why.
Second, the readiness question is about habit rather than size. If your business has changed how it works in the last two years, the data says you are three to five times more likely to make this work. If it has not, the honest first step is a small change that succeeds, not a large one that proves the point.
Third, nobody can tell you what it is worth, including us. There is no Australian government figure for the value of automating small business admin, which is why every number you see quoted for it comes from a company selling the fix. Run the calculator on your own hours and your own wages, and treat the output as a range rather than a promise. Admin hours by trade works through why that figure has to be yours.
Everything above is published by an Australian government agency, fetched and dated. None of it is a Kindra measurement, and this page was originally specced as one — a benchmark drawn from twenty-five to thirty instrumented engagements, where we count the hours going into admin before and after rather than asking anybody to estimate it.
We have not done that yet, so the page does not pretend we have. What would change it is not a bigger survey — surveys are how this field got into trouble — but instrumentation: a timestamp when a job is marked complete and a timestamp when the invoice leaves, on real businesses, for long enough to be boring.
When that exists it gets published here with the sample size, the method and the failure cases, and the industry framing above comes out. If our own numbers turn out to be less impressive than the ones vendors quote, they still get published. That is the entire point of measuring instead of asserting. In the meantime the leak calculator lets you run the arithmetic on your own figures, which are better than anybody’s average.
Because we have not measured one yet, and a benchmark you cannot show the working for is a marketing claim wearing a lab coat. What we could do honestly today was gather the official Australian statistics that do exist, date them, and mark the places where nothing exists at all — which turned out to be more interesting than the page we set out to write.
Partly, and unavoidably. The 2024–25 collection was surveyed between October 2025 and February 2026 and released in June 2026, so it describes a market some months behind the one you are in — and on a number moving this fast, months matter. That is an argument for reading it as a floor rather than a snapshot, not an argument for replacing it with a vendor survey that is timelier and less reliable.
Because the alternative is what everybody else publishes, which is a confident figure with no source under it. A business owner who knows that "hours a week on admin" has no Australian government source behind it is better equipped than one who has memorised a number from a software company’s blog. Knowing which of your inputs are real is most of the work.
20 minutes. We find where the money's going. No pitch.
Book a free Leak Check