The Australian Public Service AI Centre of Excellence has been part of the federal AI capability building story for several years. The 2026 view of its operational reality is more nuanced than either the enthusiastic announcements or the cynical commentary suggests. The Centre has delivered meaningful capability, has had real impact on how AI is deployed across the public service, and has also been constrained in ways that are worth understanding honestly.

This is a working observer’s view of where the Centre actually sits, what it does well, and where the gap between aspiration and delivery is most visible.

What the Centre Has Delivered

Several genuine capabilities and contributions are now operating that the Centre has been central to building:

The whole-of-government AI assurance framework that several agencies now operate within. The framework provides a consistent reference point for AI capability assessment, risk management, and operational governance that didn’t exist before.

A capability uplift program that has actually moved agency capability beyond what individual agency efforts could have produced. The shared training, the rotation programs, and the practical secondment arrangements have built skills across the public service that wouldn’t otherwise have developed.

Reference architectures and reusable patterns for common AI deployment scenarios. The work to define repeatable patterns for things like document processing, citizen interaction, internal automation, and analytics has reduced the time and risk of individual agency deployments.

Procurement guidance and supplier evaluation frameworks that have helped agencies make better-informed AI procurement decisions. The advice about evaluating vendor claims, structuring contracts, and managing supplier relationships has been substantively useful.

Whole-of-government engagement with the international policy and standards conversations on AI governance. Australia’s contribution to international AI governance development has been more coordinated and substantive because of this work.

These contributions are real and they wouldn’t have happened at the same scale through individual agency efforts.

Where the Gap Between Aspiration and Delivery Is Visible

Several areas where the operational reality has fallen short of the original aspirations:

The pace of capability development has been slower than the original program timeline suggested. The work has happened, but the timeline has slipped repeatedly. This pattern is familiar from many cross-government capability programs but it remains a real limitation.

The whole-of-government shared services that were envisioned as a major part of the Centre’s contribution have been more limited than originally planned. The infrastructure investment required to operate shared AI services at scale across diverse agency needs has been harder than projected.

The intended coordination of agency-level AI investments has been partial. Some agencies have engaged actively with the Centre’s coordinating role. Others have continued to make significant AI investments largely independently, with limited reference to the Centre’s frameworks or guidance.

The thought leadership ambition — Centre as a generator of significant research and analysis on AI governance, public sector AI use, and related topics — has been less prominent than the original positioning suggested. The output has been more operational and less intellectual than initially envisaged.

The Funding and Capacity Constraint

The honest reality is that the Centre’s funding and headcount have not matched the breadth of its mandate. The scope of activity it’s expected to support is broader than the resources allow. This is partly the consequence of competing budget pressures and partly the consequence of optimistic original scoping.

The implications are visible. Some priority areas receive substantial attention. Other areas receive less than they would benefit from. The prioritisation choices being made are reasonable given the constraints, but the constraints are real and shape what the Centre can deliver.

The resourcing question has been debated periodically. Whether the Centre should be expanded, refocused on a narrower mandate, or operated through different mechanisms remains an open governance question. The current arrangement is functional but stretched.

The Agency Engagement Patterns

The pattern of how individual federal agencies engage with the Centre is informative:

Agencies with limited internal AI capability tend to engage more actively with the Centre’s resources and guidance. The Centre provides capability uplift that these agencies couldn’t develop independently.

Agencies with substantial internal AI capability tend to engage more selectively. They contribute to the Centre’s work, draw on specific resources, but maintain independent capability for the work they prioritise.

Agencies in central policy or coordinating roles tend to engage with the Centre’s policy and framework work but less with the operational delivery side.

Agencies in service delivery or operational roles tend to engage more with the reference architectures and reusable patterns and less with the policy work.

This pattern is reasonable — different agencies have different needs and the Centre’s offerings are matched to different types of engagement. The aggregate effect is that the Centre is a meaningful but not dominant influence on how AI capability develops across the public service.

The Capability Development Programs

The capability development programs the Centre operates have been genuinely valuable. The structured training, the secondment programs, the rotation arrangements, and the cross-agency communities of practice have built skills and connections that wouldn’t have developed otherwise.

The challenge has been scale. The demand for these programs exceeds the Centre’s capacity to deliver them. Waiting lists for the most valuable programs are real. The agencies whose staff are accepted into the programs benefit substantially. The agencies whose staff aren’t have to develop capability through other means.

The longer-term capability building question is whether the Centre’s programs are producing a sustained capability pipeline or whether the impact is more episodic. Sustaining a capability uplift requires continued investment over years. The episodic project funding that has supported much of the Centre’s work makes sustained programs harder.

The Procurement Reform Impact

The Centre’s contribution to AI procurement reform has been one of its more concrete impacts. The procurement frameworks that have been developed are being used by agencies. The supplier evaluation tools are reducing the procurement risk that agencies face in AI engagements.

The market for AI services and products to the federal government has matured partly in response to the procurement frameworks. Suppliers know what’s expected and have adapted their offerings accordingly. The market behaviour is more professionalised than it was a few years ago.

Several Australian AI consultancies and integrators have engaged with the procurement framework development process and have built their service offerings around the requirements. Some of these firms — Team400 among them — have become regular partners on federal AI projects in part because their service models fit the framework requirements naturally.

The procurement reform work continues. Several refinement initiatives are in progress that should further reduce the friction in federal AI procurement over the coming year.

The International Coordination

The Centre’s role in coordinating Australia’s international AI governance engagement has been substantively important. The work with OECD, G20, and other international forums has been more coordinated and more credible because of the Centre’s coordinating capability.

The international counterparts — the equivalent AI centres in other governments, the international standards bodies, the multilateral working groups — generally find the Centre a useful interlocutor. Australia’s voice in international AI governance is louder and clearer than it would otherwise be.

This is a less visible part of the Centre’s work but it has long-term significance for Australia’s position in the global AI governance conversation that increasingly affects domestic policy choices.

What’s Next

Several initiatives are progressing that should shape the Centre’s role over the coming year:

A review of the Centre’s mandate and resourcing in the context of the broader federal AI strategy update. The outcome will determine whether the Centre is expanded, refocused, or maintained at current arrangements.

Continued development of the whole-of-government AI capability framework, with planned expansion into additional capability areas.

Refinement of the procurement frameworks based on early experience.

Expansion of the capability development programs subject to resourcing.

Continued engagement with the international AI governance development.

The trajectory is generally positive but the pace is constrained by the funding and capacity realities.

The Honest Mid-2026 Position

The APS AI Centre of Excellence in 2026 is a meaningful contributor to how AI capability develops across the federal public service. It has delivered real value through frameworks, capability development, procurement reform, and international engagement. The scope of its impact is bounded by the resourcing available and by the choices individual agencies make about engagement.

The Centre is not the dominant force in federal AI capability development. That distinction probably belongs to the cumulative individual agency investments in AI capability. The Centre’s role is more catalytic and coordinating than commanding.

This is probably the right framing for what the Centre realistically can be in the Australian federal governance environment. The agencies retain accountability and ownership of their AI capability. The Centre supports, coordinates, and catalyses without commanding.

Whether the current arrangement is optimal depends on what one thinks federal AI governance should look like over the coming years. There are reasonable arguments for expanding the Centre’s mandate and resourcing. There are equally reasonable arguments for keeping it in its current catalytic role and continuing to invest in individual agency capability. The debate will continue and the resolution will come through the broader federal AI strategy decisions over the next year or two.