On 15 July, the Australian Government established a new Office of AI inside the Department of the Prime Minister and Cabinet, with a remit to coordinate AI policy across government and develop Australia’s new national AI standards.
Just over a month later, National Cabinet backed a nationally consistent framework including mandatory standards for large data centres covering their energy, water and land-use impacts. The Commonwealth intends to legislate the standards in early 2027.
That tells us something about where AI has landed in Australia. It now cuts across enough portfolios, industries and parts of the economy that coordinating it has become a job in its own right.
As a Chief Technology Officer, I think that makes sense.
AI doesn’t live in one particular silo. It’s pretty much ubiquitous. It reaches into technology, privacy, workforce, infrastructure, economics and the way people actually get work done. An Office of AI can speak to AI wherever AI happens to be relevant, rather than trying to wedge it into a single existing portfolio.
I spend a lot of time thinking about this as part of my role, and even then I still pick up new ways of using AI almost every week, often through conversations with people about how they’ve approached a problem or found a new use for a tool.
For leaders, I think the Office of AI raises an interesting question closer to home: if AI is broad enough to need cross-government coordination nationally, how well are we coordinating it inside our own organisations?
Start with what is already happening
AI is unusual because governments, businesses and individual consumers are adopting essentially the same technology at the same time.
While an organisation is still working out procurement, governance and investment decisions, the people inside it are already experimenting and developing their own understanding of what AI can do for them.
I think it would be a very rare CTO who imagined they were seeing all of that activity.
That makes visibility an important starting point, although I would be cautious about treating every instance of “shadow AI” purely as a compliance problem.
It may sometimes be evidence of successful adoption. If people have found genuinely useful applications and are going outside formal systems to access them, that may tell you the organisation hasn’t adapted quickly enough to support how people want to work.
There are obviously real risks when customer information, intellectual property or contractually protected data is involved, so guardrails matter.
But if I wanted to understand AI use across an organisation, I would begin somewhere very simple: I would ask my people.
What are you using? What are you using it for? What works? Where has it saved time? What has produced a poor result?
Technical telemetry can add to that picture, but nothing is going to beat talking to people and encouraging them to share what they are learning.
Education is a governance control
This is where I think the Office of AI could provide something particularly useful.
Education is often placed in the softer change-management bucket, while policies and technical restrictions are considered the serious parts of governance. With AI, I think that’s probably the wrong way around.
If people understand why certain information shouldn’t be pasted into a model, when hallucination matters, what data provenance means and when a human needs to remain in the loop, you’ve created a workforce capable of making better decisions across situations a policy could never anticipate.
At a national level, I would like to see the Office of AI help build that shared understanding.
Practical governance baselines, education resources, guidance on cost control and useful ways of assessing return on AI investment would give organisations something concrete to work from.
That could be especially valuable for smaller businesses without a CTO or equivalent capability. For government and larger organisations, a shared vocabulary and baseline could also make conversations between agencies, Customers and technology providers much easier.
I’d also hope that conversation flows both ways, with organisations already experimenting heavily with AI able to inform the Office about what they are seeing in practice.
Responsibility should follow where AI is used
The same cross-cutting principle applies to accountability.
When AI use is distributed across hundreds or thousands of employees, responsibility cannot meaningfully belong to a single technology executive.
Employees have responsibilities, managers have responsibilities, and so do security teams, technology leaders and data owners. Executives remain accountable for making sure an appropriate governance system exists.
You can’t tell 5,000 employees that AI is everyone’s responsibility and walk away thinking the problem is solved. Equally, you can’t make the CIO personally responsible for every AI-assisted decision those 5,000 people make.
Responsibility can be distributed without becoming ambiguous, provided people understand what is expected of them and have the tools and knowledge to make good decisions.
Then ask what you’re actually getting from it
The final question I think leaders should be asking is whether all this AI activity is translating into value.
AI usage is not itself productivity.
Token consumption can tell you about usage and cost, but it doesn’t necessarily tell you whether the work is better. If the underlying process is poorly thought through, AI may simply help people travel down an inefficient pathway faster.
As the economics of AI mature, I expect this question to become much more important. Organisations will have to become more deliberate about what they are spending and what they are receiving in return.
The conversation will gradually move beyond “Are we using AI?” towards “Where is AI actually making us better?”
That is why I think the Office of AI matters beyond the standards it eventually produces.
Its creation recognises something organisations need to recognise too: AI now cuts across enough of what we do that treating it as a contained technology issue no longer reflects reality.
So while we watch what comes next from the Office, I’d be asking four questions internally: Do we know how AI is already being used? Do our people understand how to use it responsibly? Is responsibility clear? And do we know whether it is actually creating value?
The answers will probably tell you more about your organisation’s AI readiness than the number of AI tools you have deployed.
The Prime Minister’s address, “AI in Australia’s interests”, is worth reading in full. You can find it here.
Continue the conversation at the Government Innovation Showcase
I’ll be at the Government Innovation Showcase alongside the Mojo Soup team, so if you’re working through these questions in your own organisation, come and have a conversation with me about what you’re seeing.
Our CEO, Dave Lockie, will also be moderating the panel “Agile Innovation in Government: Is It Really Possible?”, bringing together government leaders to explore what needs to change around delivery, governance and innovation for agile approaches to work in practice.
