Australia will not get an AI productivity boom by training workers alone

Australia's artificial intelligence debate keeps circling around skills. Train more workers. Teach people to prompt. Build digital literacy. Those are sensible goals, but they leave out the group that decides whether new technology changes a business at all: managers.

Last week's Atlassian results are a useful reminder of how quickly the AI story swings between anxiety and optimism. The Australian-founded software company reported strong growth while its leadership argued that AI can deepen the value customers get from its products. That sits beside a wider news cycle full of layoffs, automation claims and warnings about white-collar work.

Neither the optimistic nor pessimistic headline tells an ordinary Australian business what to do on Monday morning.

The harder question is whether managers know how to redesign work around AI without simply adding another tool to an already crowded workflow.

An Australian government economic reform paper has already identified managerial capability as a barrier to productivity and technology adoption. That point deserves more attention. The National AI Plan also says small and medium enterprises are central to productivity gains from AI, while noting a sizeable gap between metropolitan and regional adoption.

Training workers without upgrading management practice risks widening that gap. Employees can learn how to use an AI assistant in an afternoon. A manager still has to decide which tasks are appropriate, what quality standard applies, when a human must intervene, how errors are reported and whether the system is producing a measurable business result.

Buying licences is easy. Redesigning work is the expensive part.

Consider a customer-service team. A worker can learn to generate draft responses quickly. But somebody must decide which inquiries the system can handle, which require escalation and what happens when a confident answer is wrong. The same pattern appears in finance, marketing, legal work, human resources and operations. Technical fluency matters, but the productivity gain depends on management choices around the tool.

That suggests Australia should treat AI management capability as a distinct economic skill.

For every significant AI deployment, managers should be able to answer five practical questions. What problem are we solving? What part of the workflow changes? What result will we measure? What mistakes require human intervention? What will we stop doing if the new system works?

That last question is especially important. Organisations often add AI without removing meetings, approvals, reports or old software steps. Workers then complete the old process and the new one. Usage rises, dashboards look impressive and productivity barely moves.

The same mistake can happen in training. A company celebrates completion rates for an AI course but never checks whether employees can apply the skill to real work. A better test is observable performance: can a worker complete a defined task more accurately, faster or with less rework? Can a manager explain where the tool should not be used? Can the team recover when the output fails?