Globally, there have been concerns about the likelihood of artificial intelligence (AI) taking people’s jobs and reducing job opportunities for university graduates.
AI software companies, such as OpenAI, Google DeepMind and Anthropic are pursuing AI systems capable of matching or exceeding human performance across a broad range of intellectual tasks.
What does this mean for graduates and universities?
Which jobs are most at risk?
A federal government report, published in July, rated jobs typically involving cognitive tasks as those with the highest exposure to AI automation. This includes accountants, clerks and receptionists.
Jobs that were rated as the lowest exposure to AI automation tended to be those requiring more physical work, such as tradies, carers and truck drivers.
Much of the government narrative has been that jobs will more likely be changed or “augmented” by AI, rather than be cut by the technology.
The report also found that as of February this year, AI had not had significant impacts on the Australian job market. This included job opportunities for graduate students, despite international concerns AI will reduce the number of entry-level roles.
This article is part of a series on how artificial intelligence (AI) is reshaping higher education.
We take a deeper look into how AI has disrupted degrees, how universities are responding to these new challenges and what the technology means for the sector’s future. You can find more stories here.
However, things might be changing. This month the Reserve Bank of Australia noted there are already some signs younger workers in more AI-exposed industries are struggling to find work in Australia.
The number of graduate jobs in some law firms has been slashed, with low-level tasks passed onto AI. Meanwhile, consulting firms have also cut graduate roles in recent years, but claim its not because of AI.
Just last week, Qantas announced it will outsource 850 jobs to India while investing more in AI.
What does this mean for desk jobs?
AI models have been improving fast and can now automate many tasks in university-qualified jobs.
Still, some experts argue reasoning and critical thinking skills are valuable for graduates in the AI job market.
Research shows AI can still compete.
In a 2024 study, researchers compared the reasoning skills of humans and AI language models (systems programmed to mimic human speech and writing) through a series of tests.
The tests included maths puzzles designed to assess whether people and AI models could recognise that an answer, which seems right at first, is actually wrong. For example, a scarf costs 210€ more than a hat. The scarf and the hat cost 220€ in total. How much does the hat cost? The correct answer is 5€.
People and AI models were then tested on how they reasoned their way to the correct answer.
Researchers found earlier AI systems performed worse than human participants in the study’s reasoning test, while more recent AI models performed better.
If this trend continues, AI could eventually outperform people on a much wider range of reasoning tasks. In one sign of that progress, an OpenAI model recently disproved a longstanding conjecture in geometry dating back to 1946.
My own research reflects similar findings.
We analysed AI’s responses on university criminal law exams in 2023 and 2025. We found its average exam scores improved from 52.5% in 2023 to 76.3% in 2025. The most striking finding was in critical thinking capabilities. Unlike the 2023 study, the 2025 models, on average, performed better than an average university student.
These were specific tests and do not prove AI could displace humans and take over an entire profession. But the results challenge the assumption AI weaknesses will remain unchanged.
What about trades?
Physical work is a challenge for AI. But if a sufficiently capable AI could control reliable, affordable robots then electricians and plumbers could also face competition.
We’re already seeing signs of this shift. Robotaxis carry passengers in a growing number of cities, delivery robots bring meals to customers and autonomous excavators dig on construction sites.
Despite significant progress in robots’ ability to carry out complex physical tasks it’s still unclear how soon they will be able to do more of the work tradies do. For now, students who choose vocational pathways, such as apprenticeships, may face a lower risk of AI-related job disruption.
Working with AI
This doesn’t mean it’s over for white-collar workers.
One likely possibility in an AI job market is that a person working with the technology will achieve a better result than AI alone. This is a form of AI augmentation, which can keep workers relevant and useful.
But research shows collaborating with AI is not an easy skill. Getting useful results takes judgement.
In a 2026 pilot study health-care professionals (mostly nurses) accepted half the deliberately wrong AI recommendations and passed over about one in six correct ones. And in a study looking at how the UK public uses AI for medical advice, researchers found people sometimes left out important details when consulting chatbots. As a result, the chatbots sometimes mixed useful suggestions with misleading ones.
How can unis prepare graduates well?
The possibility of humans working with AI should be central to how universities prepare students for the workforce.
If a worker can’t improve the quality or efficiency of AI’s output, an employer may have little economic reason to involve them.
To add value to the future workforce graduates may need to:
-
understand what a good result looks like when using AI
-
recognise when AI falls short
-
have the knowledge and skills to improve AI’s output.
They also need to learn how to collaborate with AI, how to direct it, question its output and persist until the result is better than what it produced initially.
Universities can’t simply replace disciplinary knowledge with AI skills. Students will need the knowledge and skills their degrees have always required, as well as the ability to collaborate effectively with AI.
The irony is that a technology designed to make our lives easier may actually require us to work harder to stay relevant in the job market.