A whole-of-department AI Strategy, built on four pillars

Context

Home Affairs and the Australian Border Force operate at a scale that makes manual process untenable: around 9.5 million visas, 45 million travellers and 148 million cargo consignments a year, growing faster than the workforce that handles them.

The department was not new to AI. It had used forms of it for two decades, but that capability had grown up fragmented across the border, visa and corporate domains, with no single view of where AI was deployed, what it was doing, or who was accountable for it.

At the same time the obligations were unusually heavy. National security, legal risk, public trust and Ombudsman scrutiny all bear on decisions in this portfolio, and a mandate to substantially increase the use of AI arrived alongside an expectation of demonstrably responsible governance. Those two pressures pull in opposite directions unless the strategy resolves them deliberately.

Our approach

We built the strategy around four pillars: Data, Trust, Tools and People, developed in parallel. This matters more than it sounds. Departments that treat data readiness as a precursor to AI work spend two years on foundations and deliver nothing; departments that skip it deploy AI onto data that cannot support it. Running the four together forces the trade-offs into the open.

Underneath the pillars we evidenced seven strategic drivers using both internal operational data and external benchmarks, so the case for change rested on the department's own numbers and not on vendor claims about the technology.

We then defined six guiding principles, each tied explicitly to the AI Plan for the APS and the whole-of-government AI Policy. Tying principles to published policy, instead of inventing local ones, meant the strategy could be read directly against the frameworks the department is measured on.

AI systems are only as reliable as the data they learn from. Strengthening our data foundations is not preparatory work for AI adoption. It is AI readiness itself.

The structural decision that distinguished this engagement was linking strategy, governance framework and implementation plan into a single architecture. These are usually three documents written by three teams at three different times, which is why they so often contradict each other. Here there is a traceable line from a named mission risk, to the principle that addresses it, to the milestone that implements it, to the measure that evidences it.

Outcomes

The Secretary endorsed the strategy

SecretaryThe Secretary endorsed the strategy

The framework names the CDAO as Accountable AI Official

v1.6The framework names the CDAO as Accountable AI Official

Each milestone has an accountable owner

18Each milestone has an accountable owner

The strategy was endorsed by the Chief Data and Analytics Officer and taken to the department Executive. Alongside it, Governance Framework v1.6 established the CDAO as the Accountable AI Official, the role the whole-of-government policy expects, made concrete with decision rights instead of left as a title.

Eighteen Horizon 1 milestones were scoped, costed and assigned to accountable owners, which is the step most strategies skip and the reason most of them stall. An AI register, ethical review process and AI Impact Assessment pathway were defined for operational use rather than described in principle.

The result is a clear line from mission risk through to measurable outcome: a strategy a Secretary can defend, an auditor can follow, and a delivery team can actually execute.