Help us AI, you’re our only hope.
Treasury has set a stark productivity range for Australia's AI future — and the gap between the low end and the high end depends almost entirely on what businesses do next.
Help us AI, you're our only hope.
Australia's productivity growth has slowed to a crawl, public debt has climbed past a trillion dollars, and the latest Treasury advice makes plain that artificial intelligence is being asked to carry more economic weight than any single technology has borne in decades. Businesses have begun experimenting with the tools, yet the depth of adoption remains so modest that the promised surge in output per hour worked may never arrive at the scale required to lift wages, ease cost-of-living pressures, and keep debt servicing costs from crowding out other spending.
Widespread use, shallow roots: why dabbling won't move the dial
The gap between dabbling and transformation is the heart of the problem. Treasury's note to Treasurer Jim Chalmers describes current AI use as "widespread but shallow." Finance, insurance and professional services lead the pack, which is unsurprising given their information-heavy workflows. Yet large swathes of the economy, health, education, construction and tourism, are structurally harder to reshape with chatbots and pattern recognition. Access to the technology is not the constraint; what matters is whether organisations rewrite their operating models around it. That requires spending on training, redesigning job descriptions, and accepting short-term disruption for long-term gain. Most small businesses, which employ nearly half the private-sector workforce, lack dedicated technology staff and face daily cash-flow pressures that make experimentation feel like a luxury.
This is not a new pattern. Australian firms have form for slow diffusion of productivity-enhancing tools. The data-centre construction boom now under way, forecast to be worth $150 billion by 2030, shows capital is flowing into the foundations of the AI economy. Yet as earlier analysis noted, that surge in information and telecommunications investment is largely masking weakness in broader private capital expenditure. Bricks, servers and cooling systems are easier to install than the organisational capital Treasury says is essential. Without the latter, the hardware simply hosts other people's productivity gains.
The fiscal stakes make shallow adoption unaffordable
The stakes are sharpened by Australia's fiscal position. Annual interest payments on gross debt already run into the tens of billions. Growing the economy's productive base is the least painful way to keep those payments sustainable. Treasury still judges its long-run assumption of 1.2 per cent annual productivity growth achievable, largely on the strength of AI's potential. The plausible range it sketches, 0.5 to 0.8 per cent on the downside, 1.5 to 2 per cent on the upside, reveals how much hinges on execution. The lower path is what you get if Australia treats AI as a clever spreadsheet rather than a reason to rethink entire value chains.
The United States confronts an even sharper version of the same wager. American productivity growth has been equally disappointing for most of the past two decades, while federal debt has climbed faster and higher. US firms have moved more quickly on frontier AI models, yet the broader diffusion problem remains. Economists watching both countries see a common risk: the technology delivers concentrated gains to a handful of large technology vendors and early adopters while the rest of the economy experiences little measurable uplift. The feared mass job displacement has not yet appeared in either labour market. Studies cited by Treasury find only "limited impacts" so far, with officials paying particular attention to entry-level roles overseas for early signals. That absence of disruption is welcome, but it also means the complementary gains, higher wages from higher output per worker, have yet to materialise either.
The confidence gap is costing small business more than the skills gap
Small-business hesitation is instructive. Enterprise Tasmania chief executive Brian Collins describes a "confidence gap" rather than a pure skills gap. Owners worry AI is for big companies, will cost too much, or will add one more task to an already overloaded list. The proliferation of hundreds of AI products creates choice paralysis. Judo Bank economic adviser Warren Hogan, whose institution lends heavily to smaller firms, warns Australia risks "getting the worst of AI, the displacement, without the best, which is the broader productivity lift." His point echoes Treasury's: the technology is no substitute for structural reform. Red-tape reduction, planning-system streamlining and migration policies that bring in scarce skills all remain necessary companions. An earlier look at federal efforts to cut land-use definitions in Western Australia illustrated how even modest bureaucratic simplification can matter; the same logic applies to removing barriers that slow AI integration.
The technology delivers concentrated gains to a handful of large technology vendors and early adopters while the rest of the economy experiences little measurable uplift.
Hosting the infrastructure is not the same as capturing the dividend
Global private investment in AI reached nearly half a trillion Australian dollars in 2025, up 128 per cent on the previous year. That capital is betting on transformation, not experimentation. Australia's appeal as a data-centre host gives it a seat at the table, yet hosting the infrastructure is not the same as capturing the productivity dividend. The difference will show up in wages, in the budget bottom line, and in whether younger workers inherit an economy capable of delivering rising living standards.
The evidence is not yet conclusive. AI could prove the first credible accelerant to global growth in nearly twenty years, as Treasury suggests. It could equally become another overhyped general-purpose technology whose gains are delayed by organisational inertia. The distinguishing factor will not be the sophistication of the models but the willingness of Australian managers, large and small, to treat AI as a prompt for genuine reinvention rather than a productivity placebo. The debt trajectory and the productivity numbers leave little room for half measures. If businesses do not move beyond shallow adoption, the cavalry on the hill may turn out to be a mirage.
Sources
ABC News — Australia risks missing out on AI economic boom due to 'slow' business uptake
The Bearing — Data centre boom masks weakness in broader capital investment
The Bearing — The debt trap: why governments can't just print their way out
The Bearing — Let's hope the fed's $30m red tape scissors for WA will work
Frequently Asked Questions
Why isn't AI boosting Australia's productivity if so many businesses are using it?
Two-thirds of Australian businesses report using AI in some form, but fewer than one in ten describe that use as significant. Using AI for minor tasks like drafting emails or summarising documents does not restructure workflows or lift output per worker — and that organisational transformation, not access to the tools — is what actually moves productivity numbers.
What productivity growth rate is Treasury assuming for Australia?
Treasury's long-run baseline assumption is 1.2 per cent annual productivity growth, which it still considers achievable largely on the basis of AI's potential. The realistic range it sketches runs from 0.5 to 0.8 per cent if adoption stays shallow, up to 1.5 to 2 per cent if firms genuinely restructure around the technology.
Is AI causing job losses in Australia?
Not yet in any significant measurable way. Studies cited by Treasury find only limited labour-market impacts so far, and mass displacement has not appeared in either the Australian or US labour markets. The problem at the moment is the opposite: the productivity gains that should accompany AI-driven efficiency — higher wages from higher output — have also not yet materialised.
Why are small businesses slow to adopt AI in Australia?
Enterprise Tasmania's chief executive identifies a confidence gap as the primary barrier: small business owners believe AI is designed for large companies, worry about cost, or feel they lack the bandwidth to add another operational change. The proliferation of hundreds of competing AI products compounds the problem by creating choice paralysis rather than clear options.
Does Australia building data centres mean it will benefit from AI economically?
Not automatically. Australia's data-centre construction boom — forecast at $150 billion by 2030 — funds the physical infrastructure that AI runs on, but the productivity dividend flows to whoever uses the technology to restructure their operations, not necessarily to whoever hosts the servers. Capturing the economic gains requires organisational change inside Australian firms, not just real-estate investment in compute.