For most of the past three years, artificial intelligence has been treated in Canberra as a specialist concern, something to be handled by a scattering of departments, advisory groups and consultation papers. That framing has now shifted. By locating a new Office of AI inside the Department of the Prime Minister and Cabinet, the federal government has effectively declared that the technology is no longer a niche portfolio issue but a whole-of-economy agenda that belongs at the centre of power.
Machinery-of-government decisions rarely make headlines, yet where a function sits says a great deal about how seriously it is taken. Placing AI coordination alongside national security, economic policy and cabinet business puts it in the same building as the levers that actually move money and people. The symbolism is deliberate, and it lines up with a growing recognition across business and the public service that AI is bleeding into productivity, employment, skills, procurement and service delivery all at once.
As the Australian Financial Review argues in its analysis of the move, the structure is the easy part. The genuinely difficult work sits downstream, in whether Australia has the people who can build, buy, govern and safely use these systems. On that measure, an AFR Technology piece makes the case that workforce capability, not org charts, holds the key to the country’s AI success.
The news beneath the announcement
The creation of the Office of AI has been read in several ways. Supporters see it as overdue central coordination, a single point of accountability that can stop the current patchwork of agencies from tripping over each other. Critics, including the Coalition, have argued that a dedicated office arriving now is late, given how quickly generative tools have moved through workplaces since late 2022. Both readings can be true at once, and neither settles the core problem.
That problem is capability. A national AI ambition depends on a supply of people who understand the technology deeply enough to deploy it responsibly: data engineers, machine-learning specialists, procurement officers who can interrogate a vendor’s claims, managers who can redesign a process rather than simply bolt a chatbot onto an old one, and frontline staff who trust the tools enough to use them. Right now that pipeline is thin, and it is thinnest in exactly the places where the payoff would be largest, such as the public sector, health, education and small business.
Two views on what capability actually means
There is a live debate about where the workforce gap really sits. One camp frames it as a shortage of elite technical talent, the researchers and engineers who can train models and stand up sovereign infrastructure. On that view, the answer is migration settings, university funding, and competing harder for the small global pool of specialists who can command salaries that dwarf Australian public-service pay scales.
A second camp argues the more urgent gap is broad rather than deep. In this reading, the country does not need everyone to build models, it needs millions of workers who can use AI competently and skeptically inside their existing jobs. That means practical literacy: knowing when an output is likely to be wrong, understanding what data should never be pasted into a public tool, and being able to judge whether a task is suited to automation at all. Skills recognition bodies and industry groups have pushed this line for months, warning that without shared standards, employers cannot tell the difference between genuine competence and a weekend spent watching tutorials.
The two views are not mutually exclusive, but they compete for the same money and attention. An Office of AI sitting inside the prime minister’s department is well placed to arbitrate that trade-off, provided it is resourced to do more than write frameworks and convene roundtables.
The Australian stakes
The stakes here are unusually concrete for Australia. The economy is heavily weighted towards services and the public sector, which is where AI’s productivity gains are supposed to land, yet those are also the sectors with the least in-house technical depth. Treasury and the Reserve Bank have both leaned on AI-driven productivity as part of the story for how living standards improve from here, which means a capability shortfall is not just a tech problem, it is a growth problem.
There is also a sovereignty dimension. Much of the debate over the past year has centred on whether Australia is genuinely building AI or merely hosting and renting it from overseas providers. Data centres and cloud contracts are visible and countable, but they do not create capability on their own. If the workforce cannot design, audit and adapt these systems locally, the country risks importing dependency along with the software, and paying a premium for the privilege.
Small and medium businesses, which employ the bulk of Australian workers, sit at the sharp end of this. They lack the budgets to hire specialists and the time to run their own training programs, so they will rely disproportionately on whatever public scaffolding governments put in place. A central office that ignores them in favour of large enterprises and agencies would miss most of the actual workforce.
What happens next
The test for the Office of AI will be whether it can convert a strong signal into visible delivery. Three things will indicate seriousness. The first is money and mandate: whether the office can direct skills funding and set standards, or whether it is limited to coordination and advice. The second is measurement, because without a credible picture of where AI skills exist and where they are missing, any workforce plan is guesswork. The third is speed, given that the private sector is already moving, often through informal and unsanctioned tool use that outpaces official policy.
None of this is guaranteed. Australia has a long record of announcing strategies that outrun the capability to execute them, and AI is a harder case than most because the technology keeps shifting under the policy. Still, moving the function to the centre of government at least creates the conditions for a serious answer. Whether that answer materialises will depend far less on the office’s letterhead and far more on the depth of the workforce standing behind it.
Sources: AFR Technology.
