Australian businesses adopted AI almost universally and embedded it almost nowhere. So what did Australian businesses get from AI this year? On CPA Australia's numbers, released on 9 October, mostly a subscription. Eighty-eight per cent of Australian businesses use AI. Ten per cent say it is widely used and deeply embedded, and Australia finished last of the five markets surveyed on integration.
That is not a technology problem. It is a decision about where the tool was aimed, and very few businesses aimed it at hiring.
Adoption was never the number that mattered
CPA Australia's Business Technology Report 2026 surveyed 1,679 accounting and finance professionals across Australia, Chinese Mainland, Hong Kong, Malaysia and Singapore in July and August. Australian use came in at 88 per cent, which reads as a win. The figure underneath it does not. Fifty-three per cent of Australian respondents reported full or partial integration, the lowest result of the five markets. Twenty-eight per cent use agentic AI, against 58 per cent in Chinese Mainland. Thirty-six per cent use custom-built tools, against 61 per cent.
Across all five markets the report found that 63 per cent of businesses reporting higher profitability had AI widely used or deeply embedded, against 50 per cent of those whose profitability was flat or falling. That is the only figure in the research that touches money, and it is about depth.
Outstaffer sells an AI recruiter into exactly this gap, which is a disclosure of interest rather than a pitch, and better said before the arithmetic.
Where the money actually sits on an Australian hiring line
Most of the AI a 30-person business bought this year sits on functions that were already cheap. Drafting, summarising, note-taking, first-pass bookkeeping. Useful, and almost none of it attached to a large invoice.
The large invoice is recruitment. Outstaffer publishes the comparison on its own product page as a $10k to $30k agency tax per hire, and publishes Employer of Record cover from $250 a month per employee on its pricing page.
What does one agency fee buy if you spend it somewhere else?
Forty months. One role filled at the bottom of that published range, $10,000, is forty months of Employer of Record cover for the person who fills it. At the top of the range, $30,000, it is ten years.
Why did the AI spend miss it?
Because the hiring line is bought from a third party, so it never looked like a process to automate. It looked like a supplier. A business that threads AI through its back office and keeps paying a percentage of first-year salary for a shortlist has improved the cheap half of its cost base and left the expensive half untouched. The same blind spot shows up in the costs people only count after the invoice lands.
The strongest case against reading it this way
Depth of integration is a self-reported category in a survey of finance professionals, not a measured outcome. A business running one tool well can be better off than one that has threaded AI through nine processes badly, and the 63 per cent profitability finding is a correlation across five markets, not a causal result and not Australian.
Fair. It still does not rescue the adoption number, because 88 per cent was never evidence of anything except that somebody signed up. The question of what an AI recruiter actually does to a hiring process is a better use of ten minutes than another adoption statistic.
Related reading
What I would do before the end of the year
Stop counting tools. Take the three largest discretionary people costs in the business and ask which of them any AI you bought this year has touched. For most Australian businesses that list starts with recruitment and the answer is none.
Then test it on one role. If you want the honest version of what your current hiring arrangement costs, sign up and run a single role through it and compare it against the last agency invoice you paid. Or keep the subscription and change nothing, which is also a decision.
So here is the real question, and I would like to be argued with on it: what did Australian businesses get from AI this year that actually showed up in the accounts, rather than in the stack?