Highlights

  • AI adoption is spreading rapidly across Australian businesses, but economy-wide productivity gains remain difficult to identify.
  • RBA liaison suggests many firms are still experimenting with AI rather than comprehensively redesigning operations.
  • The timing and scale of potential productivity improvements remain uncertain.

Artificial intelligence is moving from experimentation into everyday Business activity across Australia, yet its wider economic payoff remains uncertain. In a 9 November 2026 speech, Reserve Bank of Australia Assistant Governor (Economic) Sarah Hunter described AI as a general-purpose technology that could eventually reshape production, employment and living standards. The immediate challenge is that adoption does not automatically translate into measurable productivity growth.

Why adoption is accelerating

The RBA’s business liaison evidence indicates a sharp shift in the breadth of AI use between 2025 and 2026. The share of firms reporting no AI adoption fell from around one-third to almost zero. Limited use became the largest category, rising from roughly 40 per cent to about 60 per cent, while moderate adoption increased from around 20 per cent to 30 per cent. Deep or transformative use remained much less common, at approximately one-tenth of firms.

These figures point to a distinction between trying a technology and integrating it deeply into business systems. Employees may use AI to summarise documents, draft material or support customer interactions without changing the structure of the organisation. Such improvements can matter at the task level while remaining too small or uneven to shift aggregate economic statistics.

Why measurable gains may lag

The RBA’s liaison findings show a mixed picture of productivity benefits. Around 35 per cent of firms reported no gains from AI, about 30 per cent minor gains, 23 per cent moderate gains and roughly 11 per cent significant gains. The evidence therefore does not yet establish a broad, material lift in aggregate productivity.

Hunter compared the process with the productivity J-curve observed during earlier technological transitions. Businesses initially devote resources to training, experimentation, infrastructure and redesign. Those costs may precede the gains. Electricity, for example, became most productive when factories changed how they organised production, rather than simply replacing one power source with another.

AI could follow a similar path if companies redesign workflows and enable workers to take on activities that previously required more time or were not practical. However, implementation brings additional concerns about reliability, cyber risk and the quality of AI-generated output.

What a productivity breakthrough could mean

The RBA outlined a wide range of estimates for AI’s potential effect on annual productivity growth over the next decade, from around 0.1 per cent to more than 2 per cent. This wide spread is a measure of uncertainty, not a forecast that either outcome will occur.

More effective research, software development, design and experimentation could eventually expand the amount of output produced with existing resources. If those improvements become widespread, they could support higher incomes and living standards. The extent of that benefit will depend on organisational change, the speed of diffusion and whether businesses can manage new operational risks.

Key Insights

Australia’s rapid AI adoption is clearer than its aggregate productivity effect. The RBA is monitoring whether limited use evolves into deeper organisational transformation and whether early task-level improvements ultimately appear in national productivity data.