Aqera
Talk to us
Menu

Services

Resources

Government buyers

Company

AqeraOne

Talk to us

Research · Part 5 of 6

AI transformation in Australia: the problem is the funding model

Australia regulates AI through policy and existing law, has dated public-sector obligations, and funds technology as projects. The last of these matters most.

Aqera Research4 min read

Australia combines middling adoption, a large estimated productivity prize, regulation through policy and existing law rather than an AI statute, and a public funding system built around projects. The last of these is the least discussed and the most consequential for transformation.

Adoption and the productivity prize

Adoption is broad but shallow. The National AI Centre's tracker found 43% of SMEs reporting some AI adoption in December 2025 to February 2026; among non-adopters, 54% said AI is not relevant to them and 19% did not know how to use it (National AI Centre, May 2026).

The Productivity Commission's final report (19 December 2025) judged multifactor productivity gains above 2.3% likely over the next decade, about A$116 billion of extra GDP, while stressing the uncertainty (King & Wood Mallesons). That estimate sits well above Acemoglu's global ceiling of 0.66%; the difference is largely an assumption about how much work organisations actually redesign.

Jobs and Skills Australia expects generative AI to augment more work than it replaces, with the highest automation potential in middle-skill clerical work, and notes that outcomes depend on workplace decisions rather than exposure alone (HRD, on the Gen AI Capacity Study, August 2025).

Regulation through existing law

Australia has chosen not to legislate an AI act. The National AI Plan (2 December 2025) relies on technology-neutral existing laws with targeted reforms, creates an AI Safety Institute, and commits to Chief AI Officers in every agency and the GovAI hosting service (Bird & Bird). The Productivity Commission recommended that AI-specific regulation be "a last resort", after gap analyses of existing frameworks.

The most concrete new obligation is in privacy law. From 10 December 2026, APPs 1.7 to 1.9 require privacy policies to disclose the kinds of decisions made, or substantially assisted, by computer programs that could significantly affect individuals. The test is wider than the GDPR's, because a human in the loop does not by itself take a process out of scope (Allens).

Public-sector obligations with dates

Government is where AI governance is mandatory and dated. The DTA's Policy for the responsible use of AI in government v2.0 took effect on 15 December 2025, requiring accountable officials, an accountable owner for each in-scope use case, an internal register, training, and reporting of high-risk use, with the remaining requirements due in December 2026 (DTA).

In NSW, the Audit Office examined 10 agencies with significant AI use in 2026. All kept an AI register, but only four recorded what they had formally assessed; half had an AI strategy, half had an agency AI policy, and most did not track AI spending. It recommended assessing all required AI solutions against the state framework by 30 June 2027 (IDM, August 2026).

Trust: the shadow of Robodebt

Australia's public debate about automated decisions is shaped by Robodebt. The Royal Commission (7 July 2023) found that an automated debt scheme using averaged income data was "a crude and cruel mechanism, neither fair nor legal". It raised about 470,000 wrongful debts, and the Commission's 57 recommendations included a statutory body to monitor and audit automated decision-making (JURIST).

Trust in suppliers has been tested too: Deloitte partially refunded a A$440,000 government contract after its report contained AI-generated fabricated citations (CFO Dive). For Australian buyers, the governance of AI, including a supplier's own use of it, is now a condition of purchase.

The structural constraint: funding by project

Australian public investment in technology is approved and funded as projects. The Commonwealth's Digital and ICT Investment Oversight Framework runs from strategic planning and prioritisation through contestability, procurement and delivery to benefits realisation, and high-cost, high-risk investments go through a two-pass business case (DTA).

That structure suits systems with a finish line. AI systems do not have one: they need standing owners, continuous evaluation, re-validation when models change, and eventual decommissioning. A model funded as a project loses its owners at go-live, exactly when the evidence says value is decided.

The organisations that have moved furthest stopped funding change as projects and funded standing teams instead. Australian organisations must find a way to do the same inside rules that assume projects. This, more than regulation or technology, is the distinctive Australian problem.

Sources

AI transformationAustraliaPublic sectorRegulation