Despite 84% of Australians now using AI at work, a new report shows most remain stuck at beginner level, leaving billions in productivity and wage gains, as well as major ethical and legal risks, on the table
More than five million Australian workers remain stuck at beginner level when it comes to using artificial intelligence, stalling productivity and wage growth and exposing businesses to costly mistakes, a new report from RMIT Online and Deloitte Access Economics has found.
The report estimates that if just half of all beginner-level AI users lifted their skills to an intermediate level, the Australian economy could unlock an $18.9 billion dividend. Improving literacy from beginner to intermediate is associated with a 6.2% wage uplift – roughly $7,000 a year for full-time workers – rising to almost $11,000 for those who progress to advanced AI literacy.
Yet despite 84% of Australians now using at least one AI tool at work, only 7% have reached an advanced level of AI literacy, while 54% remain beginners. The findings raise serious questions about whether workers are equipped to use AI safely, ethically and strategically enough to create meaningful value.
‘Shadow AI’ and the skills that are missing
In conversation with HRD, Rima Das, workforce development director at RMIT Online, said the current approach to workplace AI skills is leaving too many employees to figure it out alone.
“To move the needle on AI literacy in Australian workplaces, we need to avoid shadow AI, where 49% of workers are currently teaching themselves through ad-hoc trial and error instead of treating it for what it is – a hard and soft skill that needs to be learnt,” she said.
Das warned that AI literacy is not just about learning prompts and tools, but building the judgement to use them safely.
“As employees are twice as likely to be advanced in technical skills over critical evaluation, improving their AI literacy will require moving beyond basic technical prompting toward structured learning prioritising critical judgement,” she said.
“For businesses, improved literacy means transitioning from cursory adoption to impactful value creation. While AI currently saves the average worker nine hours a week, nearly half of that time is simply being reallocated to more tasks rather than strategic skill development.
“By facilitating greater competency, businesses reduce the immense risk of AI-related incidents, which have already impacted 95% of companies with average losses of $800,000. Enhanced literacy ensures teams use AI safely, ethically and strategically, turning this tech tool into a driver of sophisticated, high-value output and long-term institutional growth.”
Judgement lagging behind technical skills
The report highlights a growing imbalance between employees’ ability to operate AI tools and their capacity to critically assess what those tools produce.
Workers are twice as likely to be advanced in technical AI skills (21%) – such as prompting and task automation – as they are in judgement and evaluation skills (11%), which include ethical awareness, risk assessment and the ability to spot hallucinations or errors.
Rhiannon Yetsenga, associate director at Deloitte Access Economics, said the “how” of AI use has overtaken the “should”.
“Right now, there is a clear opportunity to upskill employees where it matters most, focusing on critical evaluation and decision-making. Most workers are using AI, but far fewer are using it well,” she said.
“Improving AI literacy isn’t just about knowing how to use tools or prompt effectively — it’s about critically evaluating outputs and understanding when AI should, and shouldn’t, be used. Our research shows a clear gap: Australian workers are twice as likely to be advanced in technical AI skills as they are in judgment skills.”
Yetsenga said the economic upside of closing that gap is already showing up in pay packets.
“The average worker saves around 9 hours per week using AI, with that time largely reinvested into higher-value tasks. Moving a full-time employee from beginner to advanced AI literacy is associated with an $11,000 wage uplift, reflecting the value of these skills,” she said.
“But this is just the start: stronger AI judgement reduces errors and overreliance on flawed outputs, while enabling businesses to transform processes, drive innovation, and lift productivity — turning AI into a reliable engine for growth.”
Generational and gender divides
The report also reveals sharp generational differences in how AI is adopted and understood.
Younger workers, particularly Gen Z and millennials, are more likely to be confident and technically capable AI users, but also more prone to overestimating their literacy. Around 21% of Gen Z and 17% of millennials overstate their AI skills, compared with 10% of Gen X and 8% of Baby Boomers, increasing the risk of misuse and overreliance on flawed outputs.
Older workers, meanwhile, are more cautious adopters. More than three-quarters (76%) of Baby Boomers possess only beginner-level AI literacy, compared with 43% of millennials. Yet boomers are also more likely to hold senior decision-making roles, meaning their comfort level with AI can determine how – or whether – the technology is rolled out across the organisation.
If boomers were to catch up to millennials in AI literacy, the report estimates it would unlock a collective $3.1 billion “boomer bonus” in additional economic value.
There is also a gender literacy gap with 9% of men at an advanced level of AI literacy compared with 6% of women, and men are about a third more likely to overstate their abilities. Meanwhile, 26% of men are advanced in technical AI skills versus 16% of women, though the gap narrows significantly when it comes to judgement skills (11% versus 10%).
Productivity gains, but risks rising
On paper, . The average worker saves nine hours per week by using AI; millennials save 11 hours, while boomers save six. However, 47% of workers reallocate that time to simply doing more tasks, while only one in four (25%) invest it in further skill development.
At the same time, the lack of judgement skills is fuelling costly errors. Previous research cited in the report shows that nearly all (95%) of companies have experienced an AI-related incident over the past two years, incurring average losses of $800,000. More than half of workers have made mistakes due to AI use, and a similar proportion admit to passing off AI-generated content as their own.
Workers with advanced AI literacy are also more likely to feel the pressure of AI-fuelled productivity expectations: 52% of advanced users say they feel pressure to work faster due to AI, compared with 26% of beginners.
Call for structured, lifelong AI learning
Despite the scale of the opportunity and the risks, AI training remains patchy. Only 48% of workers receive any AI training from their employer, and just 11% receive structured, ongoing support. Younger workers spend an average of 45 hours per year on AI training, compared with 35 hours for boomers.
RMIT Online CEO Nic Cola said relying on informal learning is no longer enough.
“The research makes it clear that self-guided, ad-hoc experimentation is not enough to move the needle on national productivity. We are seeing a landscape of ‘shadow AI’ where nearly half (49%) of workers are teaching themselves through trial and error, often building surface-level technical skills while critical judgement lags,” he said.
“For employees and the economy to benefit, employers must move away from informal guidance and invest in structured, accredited learning that prioritises critical thinking and strategic application.”
Das echoed the need to embed AI literacy across entire careers, not just among early adopters.
“We also need to embed AI literacy across all career stages, but particularly for senior leaders, as they are more likely to influence AI adoption across the organisation,” she said. “What’s more, by progressing half of AI beginners to an intermediate level, we can unlock a massive $18.9 billion dividend for the economy.”
The report concludes that Australian businesses now have a narrow window to turn widespread AI use into genuine competitive advantage by targeting the skills that matter most: critical evaluation, ethical decision-making, transferability of skills across tasks, and clear rules on when and how AI can be used.
With AI adoption racing ahead of capability, the message from RMIT Online and Deloitte is clear: without structured, judgement-focused training at all levels – especially in the C-suite – Australia risks paying a growing “AI beginner tax” in lost productivity, lower wages and rising organisational risk.
