Think your staff are using AI well? Think again...
Corporate Australia is telling itself a comforting story about artificial intelligence.
Policies have been drafted. Enterprise licences signed. Training rolled out. Usage numbers, at least on paper, look healthy. Many executives are convinced their organisations are well on the way to becoming “AI‑enabled”.
The workforce is telling a different story.
A new report based on a survey of 5,000 knowledge workers in large companies across Australia’s peer markets – the U.S., U.K. and Canada – suggests that while AI tools are now widespread, genuine proficiency is vanishingly rare. Employees are using AI, but mostly in shallow, low‑value ways that barely move the needle on productivity.
For Australian HR leaders, the message is blunt: this is no longer an IT issue. It is a people, capability and work‑design problem, and it now sits squarely in HR’s in‑tray.
A Proficiency Bar Most Workers Are Nowhere Near
Until recently, “AI proficiency” meant something fairly basic: knowing what these tools are, understanding the risks, and being able to write a passable prompt. Over the past year, many organisations have focused on exactly that. Staff can now ask AI to summarise emails, tidy up a draft or provide a quick answer. They know, roughly, what not to paste into a public chatbot.
But the report argues that in 2026 the bar has shifted again. Proficiency now means something more demanding: being able to fold AI into meaningful, value‑adding work tasks every week. Not just a one‑off shortcut, but a consistent part of how real work gets done – in hiring funnels, call centres, project delivery, finance, legal and operations.
By that standard, the numbers are sobering.
According to the survey, 97 per cent of the workforce are either using AI poorly or not using it at all. Only 2.7 per cent qualify as “AI practitioners” – people who have embedded AI into their workflows and see significant productivity gains. A microscopic 0.08 per cent are classed as “AI experts”.
The vast bulk of employees sit in two groups. Some 69 per cent are “AI experimenters”: they use AI for basic, surface‑level tasks like summarising meeting notes, rewriting emails and getting quick answers. Another 28 per cent are “AI novices”: they rarely or never use AI, or tried it a few times and bounced off.
The impact on time is exactly what you’d expect. One in four workers (24 per cent) report saving no time at all from AI. Another 44 per cent save less than four hours per week. Only a minority get beyond that: 23 per cent report saving two to four hours, 18 per cent report four to eight, 8 per cent report eight to twelve – and just 6 per cent say they save 12‑plus hours per week.
For employers quietly banking on AI to claw back the equivalent of a day or more per employee, those figures should be alarming.
Stuck in a “Use Case Desert”
If you ask Australian executives why AI isn’t delivering more, many will reach for a familiar answer: people need more training.
The report points somewhere else. The real bottleneck, it suggests, is not the ability to prompt. It is knowing what to use AI for.
Across thousands of workers, a pattern emerges. Employees understand, at least in theory, what a large language model can do. They know how to ask for a summary or a rewrite. But when they look at their actual job – a queue of customer emails, a pipeline of candidates, a monthly reporting cycle, a backlog of contracts – they struggle to see where AI can do more than polish the edges.
The survey calls this the “use case desert”. A full 26 per cent of respondents said they don’t have a single work‑related AI use case. Sixty per cent said the use cases they do have are beginner‑level. When researchers analysed thousands of reported use cases, they judged that only 15 per cent were likely to generate any real return on investment for the business.
The rest? Largely busywork.
Eighty‑five per cent of knowledge workers either have no AI use case or only beginner ones. A quarter say they never use AI for work. And 40 per cent go further, saying they’d be perfectly fine never using AI again.
For HR leaders used to seeing new technologies move through a familiar adoption curve – from early resistance to eventual indispensability – this is a warning sign. The workforce isn’t just slow to adopt; a large portion is unconvinced that AI is actually useful in the jobs they do.
What People Actually Use AI For
Look at the detail of what workers call their “most valuable” use cases and the gap between promise and reality widens.
The single most common use is as a Google search replacement, cited by 14.1 per cent of respondents. Draft generation comes next at 9.6 per cent, grammar and tone editing at 5.7 per cent, basic data analysis at 3.8 per cent and code generation at 3.3 per cent. Meeting support (such as note‑taking) clocks in at 2.7 per cent, document summarisation at 2 per cent, learning and skill development at 1.6 per cent, and task and process automation – the sort of thing leaders dream about – also at just 1.6 per cent.
When you group use cases by category, writing (18.1 per cent) and research (19.6 per cent) dominate. Both are largely being used at the beginner level, for one‑off copy suggestions and basic information‑gathering. More advanced categories such as data analysis (6.6 per cent), code (6.2 per cent), customer service (2 per cent) and strategy (7.1 per cent) are comparatively rare.
The report’s analysis is blunt: 59 per cent of reported AI use cases are basic task assistance. More than a quarter had no meaningful role in broader processes or workflows. Only 2 per cent were judged to be genuinely advanced.
In other words, AI is functioning as a convenience layer – a smarter spellcheck, a faster Google – not as a serious productivity engine.
A Leadership Blind Spot Measured in Double Digits
Perhaps the most politically sensitive findings in the report sit at the top of the org chart.
C‑suite respondents are overwhelmingly positive about the state of AI in their companies. Seventy‑five per cent say they’re excited about its implications. Ninety‑four per cent say they trust its contributions. The majority (57 per cent) use AI for work daily and only 2 per cent say they don’t use it at all.
They also believe, in large numbers, that their organisations have done the right things. Eighty‑one per cent of C‑suite leaders say their company has a clear, actionable policy that effectively guides AI use. Eighty per cent say tools exist with a clear access process. Seventy‑one per cent say there is a formal AI strategy. Sixty‑six per cent say they’re encouraged to experiment and build their own solutions. Just under half (48 per cent) believe there is widespread adoption with open sharing of use cases and best practices.
Ask individual contributors – the employees without direct reports, doing much of the day‑to‑day work – and the answers change sharply.
Only 28 per cent of ICs agree their company has a clear, actionable AI policy. That’s a 53‑point gap from the C‑suite. Just 39 per cent believe tools exist with a clear access process (versus 80 per cent of executives). Only 32 per cent say there is a formal AI strategy (against 71 per cent). Just 25 per cent feel encouraged to experiment and build their own AI solutions (compared with 66 per cent of the C‑suite). And a mere 8 per cent think there is widespread adoption with open sharing of best practices, versus nearly half of leaders.
For HR chiefs reporting up to optimistic executive teams, that disconnect matters. If decision‑makers are looking at adoption dashboards and policy documents while frontline staff feel unclear, under‑equipped and unconvinced, there is a risk of complacency at the very moment when course‑correction is most needed.
Individual Contributors: Least Supported, Most Affected
The report is equally clear about who is being left behind.
Individual contributors – those without direct reports – are the least likely to benefit from their company’s AI resources, despite often doing the most repetitive, automatable work.
Only 32 per cent of ICs say they have clear access to AI tools, compared with 57 per cent of managers, 72 per cent of directors and VPs, and 80 per cent of C‑suite executives.
Just 27 per cent of individual contributors have received company AI training. For managers that figure rises to 48 per cent; for directors, 68 per cent; for VPs, 70 per cent; and for the C‑suite, 81 per cent.
When it comes to reimbursement for AI tools, the numbers are even starker: only 7 per cent of ICs are reimbursed, versus 26 per cent of managers, 44 per cent of directors, 33 per cent of VPs and 63 per cent of executives.
It’s little wonder, then, that ICs report more anxiety and less enthusiasm about AI. While many leaders are bullish, 32 per cent of individual contributors describe themselves as anxious or overwhelmed. Only a tiny slice – 7 per cent – say their managers expect daily AI use, and only around a third receive active encouragement to use AI at all. Since May 2025, manager support for AI use among ICs has actually fallen by 11 per cent.
For Australian employers in sectors under strain – healthcare, education, retail and logistics among them – starving frontline knowledge workers of tools, training and support while showering them on senior staff is an expensive way to go nowhere.
Where Industries and Functions Lag – And Why That Matters Here
Although the survey focuses on North America and the U.K., its pattern is recognisable in Australia.
Technology leads with an AI proficiency score of 42 out of 100, followed by finance at 36 and consulting at 35. Manufacturing (34), media (33) and real estate (32) sit in the middle of the pack. Food and beverage and education both score 29. Healthcare comes in at 28, retail at 27.
In each case, the leading industries are more likely to have a formal AI strategy, clear policies and accessible tools. The lagging ones are more likely to lack those fundamentals or to have policies that are unclear or unhelpful.
By function, engineering or tech roles top the table with a proficiency score of 41, followed by strategy at 39 and business development or sales at 37. Human resources and marketing both sit at 36–37. Finance or legal score 35, product 34, operations 32, and customer service or support just 27.
The details are particularly telling. Fifty‑four per cent of engineers don’t use AI for writing or debugging code, scripts or formulas. Fifty‑six per cent of marketers don’t use it for creating first drafts of content. Eighty‑seven per cent of product managers don’t use AI to create prototypes.
In other words, even in functions where the high‑value use cases are obvious, most employees are not using AI for them.
For Australia – where labour markets are tight, wage growth is under pressure and service expectations keep rising – that represents not just a skills issue, but a missed structural opportunity.
Why Current Training and Investment Aren’t Closing the Gap
To their credit, organisations are not standing still. Since March 2025, access to a formal AI policy is up 17 per cent. Clear guidelines for AI usage are up 16 per cent. Investment in AI tools and platforms has inched up a further 2 per cent.
Those investments do matter. The report finds that employees in companies with an AI strategy are 1.6 times more proficient than those without one. Workers with access to AI tools are 1.5 times more proficient than those with no access. Those who have been trained are 1.5 times more proficient than those who haven’t. Employees whose managers expect AI usage are 2.6 times more proficient than those whose managers discourage it.
And yet, even in these “higher proficiency” groups, most employees are still only at the “experimenter” level. On average, workers who have undergone AI training score just 40 out of 100 on AI proficiency.
The likely explanation is simple: most training programs are still focused on access, safety and basic prompting. They teach how to use AI safely and how to craft a decent prompt. They do not equip staff to identify bottlenecks in their own workflows, design higher‑order use cases or re‑engineer processes so that AI delivers real leverage.
In short, training is getting people to the starting line, not across the finish.
What Australian HR Leaders Need to Do Differently
If the report has a single message for Australian HR leaders, it’s this: stop treating AI as a generic technology rollout and start treating it as a deep change in how work is done.
That implies six shifts.
First, stop measuring success purely by access and adoption. If 55 per cent of your workforce uses AI weekly but only 15 per cent have value‑driving use cases, your metrics are lying to you. HR should push for measures that track time saved per employee, the quality and maturity of use cases, and tangible business outcomes.
Second, treat use case development as a managed competence, not a personal side project. The workforce isn’t stuck because it can’t prompt; it’s stuck because people don’t know what problems AI can solve in their specific role. HR can drive the creation of function‑specific use case libraries, role‑based playbooks and clear expectations that team leads will develop, test and share AI use cases – and be held accountable for doing so.
Third, close the gap for individual contributors. These are the workers doing the most repetitive, automatable tasks, yet they have the least access to tools, training and manager support. This is backwards. HR should standardise access, design fair reimbursement policies and set clear expectations that every manager will identify and track at least three AI use cases for each direct report.
Fourth, redefine what “training” means. A workforce that scores 40 out of 100 after training is not one that needs more of the same. Learning programs need to move from “how to use AI safely” to “how to identify and redesign workflows where AI can eliminate or transform work”.
Fifth, close the executive awareness gap. If C‑suite members believe deployments are succeeding while frontline staff report minimal impact, there is a data visibility problem – and a cultural one. HR can institute regular skip‑level conversations focused on AI barriers, require executives to shadow employees using AI in their daily work, and ensure that reporting includes uncomfortable truths as well as adoption wins.
Finally, accept that the bar will keep rising. The gap between AI experimenters and practitioners will widen as capabilities accelerate. That means building continuous learning infrastructure now – ongoing coaching, communities of practice, internal certifications – rather than treating AI as a one‑and‑done training topic.
For Australian organisations, the stakes are clear. The tools are here. The policies are proliferating. The training calendars are full. Yet 85 per cent of workers still lack a value‑driving AI use case, 25 per cent don’t use AI for work at all, and most of those who do are skimming the surface.
The question facing HR is not whether employees can use AI.
It is whether the organisation has shown them what to do with it – and whether leaders are willing to measure, and act on, the answer.
