Skip to content
← Joining The Dots
AI News4 min read

Twelve Per Cent or Forty-Four? Reading Australia's AI Adoption Numbers

Why the wildly different AI adoption figures landing on your desk are measuring different things, and how to build a number you can actually defend.

Depending on which report crossed your desk this quarter, AI adoption in Australia is either barely off the ground or nearly half the market. Both numbers are real. Neither is lying to you. The trouble is they count different things, and if you take either one into a board meeting without understanding the difference, you will end up complacent or panicked for no good reason.

The gap is a definition, not a contradiction

Start with the official figure. The Australian Bureau of Statistics found that around 12 per cent of Australian businesses reported using AI in the workplace in 2024-25, up from just 1 per cent in 2021-22 before generative tools were widely available (Source: ABS, 'Business adoption of Artificial Intelligence accelerates in 2024-25', 25 June 2026). That survey went to nearly 7,000 businesses and counts AI running in day-to-day operations.

Now the higher number. The National AI Centre's SME AI Pulse put adoption at 44 per cent in February 2026, its strongest result in several months, and 43 per cent across the December to February quarter (Source: National AI Centre, 'AI adoption insights: December 2025 to February 2026', 3 June 2026). That tracker surveys at least 400 business owners and decision-makers each month and counts any level of AI use, from a single ChatGPT prompt to a production system.

So the honest read is this. About 44 per cent have touched AI in some form. About 12 per cent use it in a way the national statistician counts as genuine workplace adoption. The gap between them is not a measurement error. It is the distance between having tried something and having built something.

That distinction matters the moment you report it upward. Quote the 44 and your board relaxes. Quote the 12 and they worry. Neither reaction tells you anything about your own business.

What the money says

There is a third signal, and we like it because it is harder to fudge. Weel tracks card and expense transactions across more than 4,000 Australian and New Zealand businesses to see who is actually paying for AI tools. The share paying rose from 22.1 per cent in January to 27.5 per cent in March and 30.8 per cent in June 2026, crossing 30 per cent for the first time in the history of the index (Source: Weel, 'The Weel Australian AI Spending Index, Q2 2026 Update').

Transaction data removes the self-report problem. Nobody is telling a surveyor what sounds good. Somebody signed off an invoice. That makes roughly 31 per cent a useful middle marker between casual exposure and embedded use.

It is not perfect. Weel cannot see AI that comes bundled inside a Microsoft or Google subscription a business already pays for, so it will understate genuine reach. But as a floor under the noise, paying is a stronger tell than saying.

Put the three together and you get a rough ladder. Around 44 per cent have tried it, around 31 per cent are paying for it, around 12 per cent have it running in operations. That is the shape of the market, and it is far more useful than any single headline.

Measure your own adoption before you benchmark against theirs

Here is the practical part. All three numbers describe the market. None of them describe you. Benchmarking your business against a national percentage tells you almost nothing, because you do not know whether you are being counted the way the ABS counts or the way a vendor survey counts.

A more honest internal measure ignores self-reporting entirely and asks three questions.

First, what are you paying for. Pull your actual spend. Which AI tools appear on the card or the invoice, in which teams, and did anyone approve them. Quiet spend is common, and it is where your governance risk lives.

Second, what is actually in a workflow. A licence sitting unused is not adoption. Count the processes where AI is a standing step that work depends on, not a thing one enthusiast opens occasionally.

Third, what is it worth. For each of those workflows, name the hours saved, the errors avoided or the revenue touched. If you cannot, you have activity, not adoption.

Run those three questions and you will land on your own number, and it will usually sit somewhere between the survey figure and the ABS figure. That number is defensible in a board pack because you can show the working. It also tells you where to spend next, which a national average never will.

The uncomfortable point in all of this is that the size gap is real. The ABS found 35 per cent of large businesses using AI against 22 per cent of medium businesses and around 11 per cent of small and micro businesses (Source: ABS, 'Business adoption of Artificial Intelligence accelerates in 2024-25', 25 June 2026). The businesses with the most to gain are often moving slowest, and the ones already ahead are compounding. Knowing your own honest number is how you decide which group you intend to be in.

The right next steps for you depend on where you are right now. Our AI Maturity Assessment is an easy place to start. After that, let's have a chat about your next steps.

Want to know where your team actually stands?