JACKSON HEWETT: Can AI fix Australia’s great productivity puzzle?

Artificial intelligence is being touted as a cure-all for pretty much every ill, from drug discovery and new materials to the drudgery of emails.

But can it drag Australia up from its bottom-of-the-table position in global productivity?

OpenAI — the firm behind ChatGPT — certainly thinks it can.

The company commissioned economic modelling that suggests AI could unlock $115 billion in value by 2030 if Australia gets behind the technology by training its workforce, overhauling public service delivery, incentivising business adoption through tax breaks, and investing in the digital infrastructure to make it all work.

OpenAI and its large language model peers are setting records at both ends of the ledger. The good includes becoming the fastest technology in history to reach 100 million users. The bad includes burning cash at a pace never seen before, with the top four AI firms raising almost $US100 billion for just $US11 billion in revenue so far.

As tech analyst Mary Meeker puts it, sky-high valuations, sucking in investor funds on the promise of future returns are “good news for consumers… others TBD.”

Consumers are seeing the benefit of venture capital subsidising mass adoption. All AI needs now is a use case.

At a recent roundtable attended by senior figures from government, banking, tech, law and academia, optimism about AI’s potential ran high. Assistant Minister for Science, Technology and the Digital Economy Andrew Charlton was among those grappling with how to turn that potential into policy.

Given the technology’s rapid emergence, experts differed widely on what it could deliver and how. Participants were presented with a range of productivity scenarios based on how deeply AI is integrated into workplace routines.

At the low end, AI that simply helps summarise emails would lift productivity by around 1 per cent. But at the high end — where businesses use AI to create autonomous agents that manage entire workflows — the gains could reach between 7 and 18 per cent. Compounded annually, that would be transformative.

Estimates of the impact on jobs were just as varied. Some modelling suggests that around one in five roles are highly exposed to AI. Goldman Sachs puts the figure closer to two-thirds.

The government is still weighing how directly it should intervene. Australia remains reliant on imported models and has limited influence on global standards. By contrast, Singapore launched a national AI strategy in 2019 and is now developing its own large language model.

Other nations are moving faster. The UK has committed more than £1 billion to AI research and regulatory readiness. The US, while fragmented, is leveraging its computing advantage and private capital to drive AI adoption across sectors. Australia risks becoming a passive consumer of offshore innovation unless it develops targeted areas of local strength — whether in agtech, trusted AI, or public service delivery.

Mr Charlton said Australia’s approach should not just be shaped by deep-pocketed first movers but by broad national interest. “If there is a big productivity dividend from AI,” he asked, “how can Australia ensure that the value from that productivity dividend accrues to Australians?”

Part of the challenge is that the technology is still in its messy middle: adoption is surging, but the impact is uneven and difficult to measure. That’s not unusual for transformative technologies — it took nearly two decades from the invention of the microchip to widespread adoption of personal computers.

So far, roundtable participants noted, AI has not led to large-scale job losses. But that has not eased public anxiety. One contributor warned that people felt “very disquieted” by an industry dominated by large, opaque companies, many of them offshore. There was concern that Australians had little visibility into how these tools were being deployed — and even less say in their governance.

Trust emerged as a key theme. Some argued that while corporate Australia was taking a “thoughtful, not fearful” approach to AI, parts of government remained cautious. Leadership alone wouldn’t be enough without a shift in institutional culture.

Among large corporations, there is growing momentum. Australia’s major banks and telcos are already integrating generative AI into customer service and internal systems. While it might signal the next wave of digital competitiveness, participants worried it would become inequitable if it failed to extend beyond the big end of town to small and medium-sized businesses.

Productivity potential in the public service was even harder to generate. Much of that hesitation was traced back to Robodebt, the automated welfare program that, despite not being AI-driven, has cast a long shadow over public sector innovation.

Several attendees said it had created a chilling effect, with agencies wary of taking risks for fear of reputational or political damage. AI might offer real productivity gains — but not if systems are designed to minimise blame rather than deliver results.

Several participants stressed the need for civic engagement, not just technical capability. Improving AI literacy across the community — from school curricula to workforce training — was seen as critical to building both capability and consent.

As the Government prepares for its productivity summit next month, the potential of AI will increasingly be part of the national conversation, and one that will need to extend well beyond the experts who live near its bleeding edge.