AI Policy and Productivity: What the Economists Are Actually Saying About Australia
|

AI Policy and Productivity: What the Economists Are Actually Saying About Australia

BySmallbizai.au

Australia’s Productivity Problem Is Real and Getting Worse

Jim Chalmers says AI will fix Australia’s productivity problem. The RBA says it’s complicated. Deloitte says SMBs could unlock $44 billion – but only 5% of them are actually equipped to do it. Here’s what the real economics actually say.

Start with the baseline. Australia’s labour productivity has fallen 5.7% since its peak in March 2022. The non-market sector is running at near 20-year lows. We’ve had effectively zero productivity growth since 2016, and the slide since 2022 has dragged Australia down to 18th of 69 nations in the IMD World Competitiveness Yearbook 2025. For corporate use of digital tools specifically, we rank 54th globally. (AICD, September 2025)

That’s not a minor blip. That’s a structural problem – and it’s the frame you need to understand why AI has become central to economic policy, not just a tech trend.

Deloitte Access Economics put it plainly in July 2026: the economy is expected to “limp along at less than 2.0 per cent annual growth for the next two years.” The diagnosis is broadly shared. The arguments are about what to do about it and how fast to move.

What Chalmers Is Arguing

Treasurer Jim Chalmers has been unusually direct. AI is, in his words, “central to the work that I do as Treasurer” and “central to trying to turn around a couple of decades of ordinary productivity performance.” He said this on ABC’s That’s Business podcast on 17 July 2026 – the same week as the government’s big AI policy announcement. (ABC News, 17 July 2026)

The framing matters: this is a Treasurer who met with Anthropic in April 2026 and is building AI policy into the intergenerational report due later this year – a document that sets Australia’s economic framework for the next four decades. AI isn’t a footnote in this government’s economic strategy. It’s the headline.

Chalmers has pointed to compliance-cost cuts and regulatory reform as part of the productivity strategy – the 2026 Budget forecast $10 billion in annual compliance-cost cuts to lift GDP by $13 billion. (Budget 2026–27: Productivity) The logic is simple: remove friction, increase output per unit of input. The AI play is meant to be the next layer on top of that.

The government position is: we have a productivity problem, AI is part of the solution, and we need to move fast enough to capture the opportunity without getting left behind by the UK, US, and South Korea, which are already “forging ahead.” (McKinsey, cited by AICD)

What the RBA Is Actually Saying

The Reserve Bank’s position is more measured – which is exactly what you’d expect from an institution whose job is to not get caught out by optimistic assumptions.

In its November 2025 Bulletin, the RBA published research based on liaison with over 100 medium to large firms across industries. The findings were broadly positive but came with important caveats.

The positive: “Firms anticipate these investments, particularly in artificial intelligence tools, to be labour-saving and productivity-enhancing over the long term.” (RBA Bulletin, November 2025) The historical evidence also suggests that labour-creating effects of past technologies have generally outweighed labour-replacing effects in aggregate.

The caveats are worth reading carefully. The main barriers firms cited to realising productivity gains from AI were not technology limitations – they were the regulatory environment and access to suitable labour. The technology exists. The surrounding conditions to deploy it effectively are what’s holding things back.

The RBA also found that AI adoption takes several years to show up in productivity statistics – a normal technology diffusion lag documented in the academic literature (Brynjolfsson, Rock and Syverson 2021, cited by the RBA). This matters because it means you can’t look at current productivity numbers and conclude AI isn’t working. The timeline is longer than a single budget cycle.

There’s also a monitoring note. RBA research found that for occupations with AI exposure one standard deviation above average, employment was about 2% lower by February 2026 than it would have been under pre-ChatGPT trend. That’s a real effect, and the RBA is watching it. But the aggregate employment outlook – accounting for new roles created by productivity gains – may still be net positive. “May” is doing important work in that sentence.

Deloitte’s SMB Numbers – and Why 95% Matters

The most striking data on this topic comes from Deloitte Access Economics in their November 2025 report “The AI edge for small business,” commissioned by Amazon.

The headline figure is $44 billion: if 1 in 10 SMBs currently at basic or intermediate AI use advanced just one rung on the adoption ladder, that’s the estimated annual GDP contribution. (Deloitte, November 2025)

The numbers on what that rung-up looks like are significant: moving from basic to intermediate AI use correlates with a 45% increase in profitability. Moving from intermediate to “fully enabled” AI correlates with a 111% increase.

But here’s the problem: only 5% of SMBs using AI are currently “fully enabled” – meaning they’ve embedded AI into core processes with training and centralised data access. The other 95% are getting partial benefits at best.

“SMBs contribute more than half of Australia’s private sector GDP and generate 60% of company profits. However, they also lag larger enterprises in productivity per hour worked. This productivity gap drags on Australia’s broader economic performance,” said John O’Mahony, Deloitte Access Economics Partner. (November 2025)

The $44 billion figure is real potential. But it requires most of the SMB economy to move from dabbling to doing – and that’s not a given.

The Productivity Commission Said Go Slow. The Government Went Fast.

Here’s the tension that doesn’t get enough attention.

In December 2025, the Productivity Commission released its final report on “use Data and Digital Technology.” Its recommendation on AI regulation was clear: exercise care, implement only measures proven necessary by gap analysis, rely on existing technology-neutral laws rather than creating new AI-specific obligations. (Productivity Commission, December 2025)

Six months later, in July 2026, the Albanese government announced Australia’s first mandatory AI standards – establishing a new Office of AI in PM&C and signalling the move from voluntary guidance to regulatory obligations. (Office of AI | PM&C) The PC’s recommended cautious, evidence-based approach was effectively set aside.

The government’s argument is roughly: Australia can’t afford to wait. Other countries are moving. The productivity crisis is too serious for a slow-walk approach to regulation. Get the framework right, get it in place, and let business adjust.

The Productivity Commission’s argument was: premature regulation risks locking in the wrong rules before we know what’s actually needed. Let the evidence accumulate first.

Both positions are coherent. The government’s won – for now. But the tension between “move fast to capture the opportunity” and “don’t regulate before you know what you’re regulating” is going to play out over the next 18 months as draft legislation is developed and tested against real business feedback.

What This Actually Means If You Run a Small Business

The economics here are not academic. They point to a few practical things worth understanding.

The productivity gains from AI are real – but uneven. The businesses capturing the full benefit are the 5% that have embedded AI into their core operations. If you’re using AI for the odd task here and there, you’re getting something – but not the 45% profitability lift Deloitte is measuring at the intermediate level.

The regulatory barriers are real too. The RBA found that the main obstacles to AI productivity gains aren’t the technology – they’re the rules around it and the people skilled enough to use it. This is partly why the July 2026 policy announcement matters: getting the regulatory framework right (or wrong) will affect how fast businesses can realistically move.

The adoption lag means the payoff is slower than the headlines suggest. If you invest in AI tools and systems now, the measured productivity benefit may not show clearly for a year or two. That’s not a reason not to invest – it’s a reason not to expect instant results.

The $44 billion opportunity is contingent. It requires not just businesses starting to use AI, but businesses moving from starter-level to intermediate-level use at scale. The 1 in 3 non-adopters who say they “don’t know where to start” are part of the problem – and there’s no sign that the knowledge gap is closing fast enough on its own.

The economists broadly agree: Australia has a productivity problem, AI is part of the solution, and the opportunity is biggest in the SMB sector. They disagree about how fast to regulate and what the employment effects will look like. If you’re a small business owner reading the news, that honest summary is more useful than either the “AI will fix everything” pitch or the sceptics who say it won’t matter for years.

The data says it matters now – just not equally, not automatically, and not without doing the work to embed it properly.

Sources

📖 More deep dives: News Deep Dives

Related reading:

Sources

📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs

📬 The SmallBizAI Brief

One practical AI tip for Australian small business. Every Tuesday. Free.

Join business owners getting smarter about AI every week.

Subscribe free →

📖 More deep dives: News Deep Dives →

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *