Big tech says AI could boost Australia’s economy by $115 billion a year. Does the evidence stack up?

AI is on the agenda in Canberra. In August, the Productivity Commission will release an interim report on harnessing data and digital technology such as AI “to boost productivity growth, accelerate innovation and improve government services”. Shortly afterward, the government will host an Economic Reform Roundtable where AI policy will be up for discussion.

AI developers are aggressively pursuing influence over the new rules. The Chinese government wants to include AI in trade deals. Meanwhile, as the US government seeks to “win the AI race”, US-based tech companies are making their own overtures.

The most ambitious intervention has come from ChatGPT developer OpenAI, which recently hired former Tech Council chief executive Kate Pounder as its local policy liaison. Pounder is also a former business partner of Assistant Minister for the Digital Economy Andrew Charlton.

OpenAI’s AI Economic Blueprint for Australia makes bold projections about the new technology’s impact on the country’s economy, accompanied by a host of policy proposals. However, these claims warrant careful scrutiny, particularly given the company’s clear commercial interests in shaping Australian regulation.

The gap between promise and evidence

OpenAI claims AI could boost Australia’s economy by A$115 billion annually by 2030. It attributes most of this to productivity gains in business, education and government. However, the supporting evidence is thin.

For instance, the report notes Australian workers have lower productivity than their US counterparts and then claims (without evidence) this is because Australia has invested less in digital technologies such as AI. However, it ignores numerous other factors affecting productivity, from industrial structure to regulatory environments.

The report also describes supposed AI-driven productivity gains in companies such as Moderna and Canva. However, these narratives lack any data about improved organisational or individual performance.

Perhaps more concerning is the report’s uniformly optimistic tone, which overlooks significant risks. These include organisations struggling with costly AI projects, massive job displacements, worsening labour conditions, and concentrating wealth.

Most problematically, OpenAI’s blueprint assumes AI adoption and its economic benefits will materialise rapidly across the economy. However, evidence suggests a different reality.

Economic impact from AI will unfold gradually

Recent evidence suggests AI’s economic impact may take decades to fully materialise. Studies report some 40% of US adults use generative AI yet this translates to less than 5% of work hours and an increase of less than 1% in labour productivity.

AI may not spread much faster than past technologies. The limiting factor will be how quickly individuals, organisations and institutions can adapt.

Even when AI tools are available, meaningful adoption requires time. People must develop new skills, change the way they work, and integrate the new technologies into complex organisations. The economic impacts of earlier general-purpose technologies such as computers and the internet took decades to fully materialise, and there’s little reason to believe AI will be fundamentally different.