Treasury: Australia's AI Use Is Wide but Shallow

Two-thirds of Australian businesses now use AI. Fewer than one in ten use it in a way that changes anything. That gap is the finding in Treasury's own analysis prepared for Treasurer Jim Chalmers, reported by the ABC on 31 August — and it is the clearest signal yet that adopting AI and profiting from it are two different projects.

For a trades or professional-services owner, that is the more useful headline than any "AI will add $X billion" forecast. It says the crowd has already picked up the tool. Almost none of them have turned it into money. Which means the advantage is still there for the taking — but not by signing up for one more app.

The document is a detailed economic assessment of AI prepared by Treasury for the Treasurer. Its numbers are blunt: around two-thirds of Australian businesses report using AI in some form, but fewer than 10 per cent describe their adoption as significant. The ABC characterised the pattern as "widespread but shallow." Adoption is led by finance, insurance, information and telecommunications, and professional services.

Treasury's central estimate is that AI supports about 1.2 per cent annual productivity growth over the long run — with an optimistic case of 1.5 to 2 per cent and a realistic downside of 0.5 to 0.8 per cent. The difference between those scenarios is not the technology. It is whether businesses actually change how they work. In Treasury's words, realising the gains "requires investment in organisational capital, including changes to processes, business models, management practices and workforce skills." Access to AI alone, it warns, will not deliver economic gains. On jobs, Treasury found limited employment impact so far, while flagging a potentially profound effect on cognitive, white-collar work over time.

Want to know what this means for your business? We'll map it in one call.

Here is the distinction that decides whether your AI spend returns anything. A tradie asking a chatbot to word a quote, or a bookkeeper pasting a client email in for a summary, is the two-thirds. It is genuinely useful, and it costs almost nothing. But it does not move the P&L, because nothing about how the work flows has changed. The tool got sprinkled on top of the old job. Treasury's point, translated into plain English: the return lives in redesigning the job around the tool.

This is a Cost Intelligence problem before it is a technology one. At a loaded rate of $65 to $95 an hour, the money is in the hours you hand back — and shallow use hands back minutes. Take an accounting firm running an AI summariser ad hoc: it saves a few minutes here and there and nobody can find it in the accounts. The same firm that rebuilds one workflow end to end — BAS-prep drafts, or client onboarding — around AI, with a human signing off before anything goes out, recovers hours per week per person. Same tool. Different depth. Entirely different result. It lines up with what we found when Australia's AI ROI problem turned out to be the data and the process, not the AI. If you're not sure which of your workflows is even exposed to AI yet, see where it touches your business before you buy anything.

Shallow is the default for an obvious reason: signing up is free and instant, while redesigning a process is real work. That is also why the depth is a durable advantage rather than a fad — most of your competitors will stay in the shallow end. Parliament is now running a national inquiry into exactly why smaller firms lag, with submissions closing 14 September. On the government's own numbers, the lag is less about access than about depth.

Don't add another AI tool this quarter. Take one workflow — the one where your team loses the most billable time, whether that's quoting, scheduling, BAS prep or client intake — and redesign it around AI from end to end, keeping a human on anything that moves money or reaches a customer. Run it for a fortnight and measure the hours saved against the errors caught. That single deep build will teach you more than ten shallow trials, and it is where Treasury's 1.2 per cent actually shows up on your side of the ledger. The businesses that pull ahead this year won't be the ones using the most AI. They'll be the ones using one thing properly.

Key takeaways

Common questions

Related notes

Using AI, but not seeing it in the numbers?

That's the shallow-adoption trap Treasury just described. A short call picks the one workflow worth redesigning around AI — and tells you what depth actually returns for a business like yours. Book a call.