Twelve months ago, every mid-size company in Australia was chasing the same fifty machine learning engineers. Salaries spiralled past $350,000. Recruiters camped outside university labs. It felt like the only path forward was winning a bidding war for a handful of specialists.

That chapter is closing. And the next one is far more consequential.

The Hiring Frenzy Ran Into Reality

The numbers tell the story clearly. According to the Tech Council of Australia, the technology workforce grew by 53,000 roles in 2025, but demand for dedicated AI research positions actually flattened in the second half of the year. Companies that spent 2024 competing for PhDs started realising something uncomfortable: hiring three brilliant researchers doesn’t help if the other 500 people in your organisation can’t work with what they build.

The bottleneck was never the models. It was the gap between a prototype in a Jupyter notebook and an organisation that knows how to act on AI outputs. That gap doesn’t get closed by adding more specialists to the payroll. It gets closed when your product managers, operations leads, and customer service teams can actually integrate AI into their daily workflows.

From Talent Acquisition to Capability Building

The companies pulling ahead right now aren’t the ones with the biggest AI teams. They’re the ones running structured internal programs to bring everyone else up to speed.

ANZ Bank launched what it calls an “AI fluency pathway” in late 2025. Rather than funnelling all AI work through a centralised team, they trained 2,400 employees across business units on prompt engineering, output evaluation, and knowing when to trust (or override) model suggestions. Their internal metrics show a 34% increase in AI tool adoption within the first quarter.

It’s a pattern showing up globally. Microsoft’s 2026 Work Trend Index found that organisations with formal AI upskilling programs saw three times the productivity improvement from AI tools compared to those relying solely on hiring. The difference wasn’t the technology. It was whether people knew how to use it.

Jobs and Skills Australia has flagged this as a national priority. Their 2025 capacity study estimated that 3.7 million Australian workers will need some form of AI-adjacent skill development by 2028. Not to become data scientists. To become competent collaborators with AI systems already embedded in their tools.

What Effective Upskilling Actually Looks Like

The worst version of this is a mandatory two-hour webinar where someone explains what a large language model is. That’s box-ticking, not capability building.

The programs that work share a few characteristics. They’re role-specific. A financial analyst needs different AI skills than a marketing coordinator. They involve hands-on practice with the actual tools the company uses, not abstract tutorials. And they’re ongoing, not one-off.

Atlassian’s approach is instructive. They embedded what they call “AI coaches” into each product team. These aren’t external consultants. They’re existing employees who received intensive training and now spend 20% of their time helping colleagues integrate AI tools into specific workflows. The cost is a fraction of hiring dedicated AI engineers, and the results compound because knowledge spreads organically through teams.

There’s also a critical evaluation component that most programs miss. Teaching people to use AI tools is necessary but insufficient. People also need to develop judgment about when AI outputs are reliable and when they’re confidently wrong. The MIT Technology Review documented multiple cases in 2025 where organisations faced costly errors not because the AI failed spectacularly, but because employees lacked the training to spot subtle inaccuracies in otherwise plausible outputs.

The Uncomfortable Truth About Unicorn Hiring

Here’s what the talent war really taught us: hiring a few exceptional people is satisfying but structurally limited. One brilliant machine learning engineer can build an incredible model. But if the sales team doesn’t trust AI-generated leads, if the operations manager doesn’t understand why the demand forecast looks different, if the customer service reps don’t know how to use copilot tools effectively, then that model sits in production generating outputs nobody acts on.

The organisations that spent 2024 and 2025 in hiring wars are now sitting on expensive AI teams producing work that the broader organisation can’t absorb. Meanwhile, companies that invested the same budget into broad-based skill development are seeing returns across every function.

This isn’t an argument against hiring AI specialists. You still need them. But the ratio matters. For every AI engineer you bring in, you probably need to upskill forty existing employees to make their work effective. Most companies have that ratio inverted.

What Comes Next

The skills war won’t be as dramatic as the talent war. There won’t be headlines about record salaries or poaching scandals. It will be quieter and slower, playing out in training budgets, internal program design, and the unglamorous work of changing how people do their jobs day to day.

But it will determine which organisations actually capture value from AI and which ones just talk about it. The technology is increasingly accessible. The competitive advantage now lives in whether your people can work with it, not whether you managed to hire the right handful of specialists.

The talent war made for good stories. The skills war will make for good companies.