AI procurement across the Australian Public Service has matured noticeably through 2024–2026. The early-period AI experiments funded through small operational budgets have given way to more structured procurement processes, clearer assurance frameworks, and more coordinated supplier engagement. Worth a working read of where the picture sits in May 2026.

The procurement landscape.

The Commonwealth coordinated procurement arrangements for AI-related services have continued to develop. The Buy ICT arrangements and the various coordinated panels that cover AI services have been refined through the period. The agencies have a clearer picture of which arrangements to use for which categories of AI procurement than they did two years ago.

The standalone agency procurement of AI services continues alongside the coordinated arrangements. The agencies with the most active AI programs — typically the larger service-delivery agencies and the specific policy agencies with AI-relevant policy responsibilities — have built internal procurement capability that handles AI engagements without leaning heavily on the coordinated arrangements.

The major Microsoft-related AI procurement has been a particular focus area. The whole-of-government Microsoft licensing arrangements have evolved to cover the relevant AI components. The Microsoft Copilot, Microsoft 365 AI-related capabilities, Azure AI Foundry, Microsoft Copilot Studio, Microsoft Fabric, and the broader Microsoft AI stack are all now visible categories within the procurement and operational arrangements.

The custom AI development engagement.

The procurement of custom AI development services — the engagement with specialist AI consultancies for bespoke AI development work — has continued at a measured pace across the APS. The agencies running production AI builds in 2026 are mostly using a small number of trusted suppliers rather than open-market tendering for every engagement.

The market for AI consulting services to the APS has narrowed. The early-period proliferation of AI vendors competing for APS work has given way to a more consolidated supplier landscape. The serious AI consulting providers to the APS have built specific public-sector capability — security clearances, security accreditation pathway, public-sector delivery experience, and the contracting and assurance capability to work inside the APS operating environment.

The assurance framework.

The Department of Finance, the Digital Transformation Agency, the Australian Signals Directorate, and the relevant individual agencies have all contributed to the maturing AI assurance frameworks. The standardised assurance language has continued to develop. The required documentation for AI implementations now typically includes use-case assessment, risk classification, data governance arrangements, model behaviour documentation, evaluation and monitoring plans, and assurance reporting cadence.

The security and information protection assurance work has been a focus area. The data residency, the model behaviour, the access controls, and the audit trail capabilities have all received more rigorous treatment than they did in earlier AI engagements. The high-security agency AI work has had particularly developed assurance practice.

The legal and ethical assurance frameworks have continued to develop in parallel. The Australian AI Ethics Framework and the various agency-specific responsible AI practices have been more visibly integrated into the procurement and implementation work.

The skills and capability picture.

The internal APS capability to specify, procure, and govern AI services has continued to develop. The number of APS staff with deep AI expertise remains constrained but the broader staff understanding of AI considerations has expanded. The cross-agency professional networks for AI practice — internal to the APS and across the broader public sector — have been a useful capability-building mechanism.

The reliance on external specialist support for AI engagement design and implementation remains real. The pattern at most agencies is a small internal AI capability supported by external specialist suppliers for the deeper technical and delivery work. The relationship between internal capability and external supplier engagement is one of the more important operating considerations.

The cross-government coordination.

The cross-APS AI capability program has produced shared capability assets — common evaluation frameworks, shared training material, shared supplier reference points, and common operational templates. The agencies that engage with the coordinated capability program are spending less time reinventing the operational basics than the agencies that do not.

The international engagement on AI policy and AI government procurement has continued. The Australian participation in OECD, Five Eyes, and other relevant international AI policy fora has been steady. The international comparisons of AI procurement and assurance practice are useful inputs to the domestic operating environment.

A note on the supplier conversation.

The AI consulting firms that work effectively with the APS in 2026 share several characteristics. They have built specific public-sector capability rather than treating APS engagements as commercial-sector overflow. They have engaged with the assurance and security frameworks rather than complaining about them. They have invested in delivery teams with public-sector experience. They have built long-term relationships with their APS clients rather than chasing transactional engagements.

The supplier conversation is one of the more interesting indicators of the maturity of APS AI procurement. The arrangement between the APS and its specialist suppliers has become a more grown-up working relationship than it was two years ago.

The outlook for the rest of 2026.

The procurement pace through the rest of the year is expected to be steady. The major coordinated arrangements have most of the operational tooling they need. The agencies are continuing to scale their AI programs at a pace consistent with their capability development.

The realistic assessment is that the APS in 2026 has a workable AI procurement environment that is continuing to develop. The arrangements are not perfect but they are operating. The next twelve months should see continued operational maturation rather than dramatic shifts.