The AI workforce debate in Australia is centered around an odd irony. The nation is spending a lot of money—according to recent ADAPT research, businesses alone invest an average of $28 million a year in AI—and Canberra policymakers have made workforce capability a top priority. Curricula at universities are being redesigned. There are announcements of industry partnerships. Microcredentials are becoming more and more common. And yet, the people who go through all of that preparation, the ones who actually come out the other side genuinely skilled, keep leaving.
It’s not a secret. Anyone who has spent time around tech hiring in Sydney or Melbourne has probably noticed it. Graduates who can bridge the gap between technical AI fluency and practical business application are highly sought after, and their options typically extend far beyond Australia’s boundaries. And increasingly, they’re taking them.
Compared to five years ago, this is more important now. The employment market is changing in ways that penalize nations that lag behind. Entry-level roles are being hollowed out by automation, and the work that remains is demanding more — more judgement, more domain knowledge, more ability to apply AI responsibly rather than just operate it. According to a recent study, employees are about twice as likely to have trouble using sound judgment as they are with the technical aspects of AI tools. That’s not just a training issue; it’s a pipeline issue.
Canberra’s response, broadly speaking, has been to lean into education. That makes sense. Developing skills is necessary to close a skills gap. Collaboration between universities like the University of Western Australia and multinational tech companies suggests something genuine: industry-informed curricula, applied learning, and credentials that represent how work is actually done today rather than how it was done when curricula were last updated. These factors are not insignificant, but they take time to manifest in the labor market.
The retention aspect of the problem has not been adequately addressed in the policy discourse. Training someone to be genuinely capable in AI, to understand its limits, its risks, its application in complex environments — that takes years. Recruiters in markets with deeper capital pools, higher salaries, and, to be honest, more intriguing problems to work on become aware of that individual as soon as they are able to add significant value. It seems as though Australia is serving as a testing ground for talent that is tested both domestically and internationally.
It’s also worthwhile to consider the gender aspect of this. The roles that women and young workers have traditionally used to enter the workforce are being disproportionately displaced by the adoption of AI. For a sizable section of the population, clerical work, administrative duties, and junior-level coordination are the beginning of their careers. The inequality story gets much worse if those entry points are closing while the more advanced, higher-paying positions covertly relocate abroad.
Canberra’s understanding of the distinction between retaining and expanding a workforce is still lacking. The language around sovereign capability is everywhere right now — and it’s right to be there. Without local talent that stays, digital infrastructure, critical systems, and AI governance cannot function. However, announcements regarding curriculum reform and partnerships do not always translate into retention. They develop into credentials, which subsequently serve as leverage for a person’s subsequent overseas application.
Australia has previously created skills gaps in engineering, medicine, and research, and it has seen other nations profit from its investments. It’s possible that the AI version of that pattern is coming sooner than anyone anticipated. Whether the skills plan is intended to support Australia’s economy or unintentionally support someone else’s is an important question.
