By Darian Macey, March 26 2026
A quiet shift last week
Last week, Defence quietly removed the biggest excuse holding this sector back from using AI.
The release of the Policy Settings for Responsible Use of Artificial Intelligence did not make headlines. But it marks a clear shift. For the first time, Defence has stated in plain terms that AI is expected to be used across its functions — not avoided.
The real opportunity is not where people are looking
For all the noise around AI in Defence, the biggest near-term opportunity is not autonomy, robotics, or advanced design tools.
It is paperwork.
Across the Defence estate system, a large share of effort is consumed by routine, high-volume tasks:
- drafting procurement documents
- preparing bid responses
- writing business cases
- compiling reports
- summarising meetings
- managing correspondence
- reviewing submissions
This is the daily workload of Defence staff, Commonwealth representatives, consultants and contractors.
It is also where the system slows down.
The real blocker: perception, not capability
Most organisations are not blocked by lack of use cases.
They are blocked by fear.
Across the sector, there is a consistent pattern. Firms are experimenting with AI internally, but deliberately keeping it away from Defence-facing work.
The concern is simple: If Defence thinks we are using AI, we will be seen as careless with security or cutting corners.
So people hold back.
They recreate documents manually. They absorb avoidable effort. They slow themselves down.
All in the name of protecting trust.
That logic made sense six months ago.
It does not anymore.
What the policy actually says
The new policy is clear about Defence’s intent.
AI is expected to:
- improve the speed and quality of decision-making
- increase productivity in repetitive tasks
- optimise logistics and maintenance in support of operations
Just as important is what it does not say.
It does not say avoid AI.
It does not say wait.
It does not restrict use to isolated trials.
It says use AI — but use it responsibly.
That comes with conditions:
- humans remain accountable
- outputs must be explainable
- systems must be reliable and secure
- risks must be managed through proportionate controls across the lifecycle
This is not a barrier.
It is a framework for use.
Two attitudes, one organisation
You can already see how this plays out in practice.
In March 2026, two very different attitudes emerged inside Defence.
At Collab X, SEG leadership pointed to cyber security concerns when asked about AI in evaluations. The implication was that the environment is too sensitive, and the risks too high, to move quickly.
That view is common across the estate sector.
But at the JP8218 industry brief earlier in the month, the posture was different.
Within a CASG-led program under pressure to deliver, the project team made it clear they intend to compress a major tender evaluation down to around two weeks – something that would normally take three to six months – and use AI to help do it, with human validation over the top.
Same department. Same policy environment.
Two completely different responses.
One uses cyber risk as a reason to pause.
The other treats time as the bigger risk – and moves.
The policy released last week clearly supports the second approach.
Where AI actually delivers value
The policy’s reference to optimising logistics and maintenance is instructive.
In estate terms, that is not abstract. It is about:
- work orders
- asset condition
- contractor performance
- maintenance scheduling
- prioritisation of limited resources
All of this depends on turning information into decisions.
And most of that work is still manual, document-heavy, and inconsistent.
The same applies across:
- procurement documentation
- business case writing
- monthly reporting
- contract correspondence
- meeting summaries and decision briefs
These are not edge cases.
They are the system.
AI is well-suited to this kind of work:
- drafting first-pass procurement and evaluation documents
- structuring business cases from existing data
- summarising submissions and identifying risks
- generating reports from live inputs
- capturing and organising decisions from meetings
This does not replace people.
It removes friction around them.
The implication for industry
The signal from Defence is now clear enough.
The question is whether industry is prepared to act on it.
If you continue to treat AI as something that must be hidden from Defence, you will:
- fall behind firms that are using it properly
- continue absorbing unnecessary cost
- and deliver slower, less consistent outputs
All while telling yourself you are being cautious.
You are not.
You are preserving inefficiency.
The right approach is straightforward:
- use AI in controlled, secure environments
- keep sensitive data out of uncontrolled systems
- maintain full human accountability for outputs
- ensure outputs are traceable and reviewable
That is exactly what the policy requires.
Bottom line
Last week’s policy did not introduce AI to Defence.
It legitimised it.
The Defence Estate is not blocked by lack of use cases.
It is blocked by hesitation — driven by fear of how AI use will be perceived.
That hesitation is now out of step with Defence’s direction.
The real opportunity is not advanced autonomy.
It is reducing the administrative burden that slows procurement, reporting, maintenance planning and decision-making across the system.
The organisations that move on this now will have a structural advantage.
The rest will keep waiting for permission that has already been given.
Author: Darian Macey
Darian is a former Engineering Officer with the RAE with tertiary qualifications in Engineering and Project Management. He is a leading practitioner within Australia's NSI sector with previous employment in Government, project management and building contractor organisations.
