AI use is rising in Australia, but most businesses are still early in adoption.
The latest ABS business survey provides a useful counterpoint.
The Australian Bureau of Statistics reported that 12% of businesses said they used AI in 2024–25, compared with 1% in 2022–23. The survey did not measure the intensity or quality of that use.
That is extraordinary growth.
It is also a reminder that most businesses in that 2024–25 survey did not report AI use. It is not a live measure of adoption in September 2026.
It means the market does not only need advanced agent orchestration and model optimisation. Many organisations still need help understanding what AI is appropriate for, which products are safe to use, how staff should be trained and where automation will produce a real return.
There is a danger that the technology industry designs for itself.
We talk about autonomous agents negotiating with other agents while a typical small business is still deciding whether employees should paste customer information into a public chatbot.
The opportunity is therefore broader than selling sophisticated AI.
At one end, a business may start with controlled productivity use: drafting, summarising and research.
The next stage might connect AI to structured internal knowledge.
Then AI becomes part of workflows with approvals, permissions and auditability.
Eventually, some systems may act more autonomously.
For many organisations, moving straight to autonomous action without testing data, controls and ownership would be a risky shortcut.
Australia’s National AI Plan reflects this need to spread adoption while maintaining safety. The Government is also supporting AI Adopt Centres and publishing practical guidance for businesses.
For agencies, this suggests a strong service opportunity.
Do not sell “AI transformation” as an abstract promise.
Help the client identify one valuable use case.
Build the approval process.
Measure whether it works.
Then expand.
That may sound less exciting than autonomous everything.
Expand only after the first use case has shown a useful result. Check whether staff can review outputs, correct mistakes and explain where the data went.
Document what worked, what failed and who owns the next stage. That gives a business a repeatable path to wider adoption without treating autonomy as the starting point.


