The AI licences renewed. Executive dashboards showed adoption climbing through the first two quarters. Then the curve flattened. Some teams kept using the tools; others quietly returned to how they worked before. Nobody escalated it, because nobody was asked to. A year into the investment, the productivity case that justified the spend has not held, and the people who were meant to benefit are working around the technology rather than with it. This is the shape AI adoption failure takes inside Australian enterprises, and the reason it stalls has almost nothing to do with the tools themselves.
In Brief
- AI adoption stalls on unchanged workflows, incentives and trust conditions, not on the technology executives assumed was the barrier.
- In Prosci's 2026 study, user proficiency accounts for 38% of implementation difficulty, while technical issues account for only 16%.
- McKinsey finds high performers are 2.8 times more likely to redesign workflows, the strongest predictor of AI earnings before interest and tax (EBIT) impact.
- Australian employees use AI regularly, yet only 36% trust it and 48% breach their organisation's AI policies (KPMG & University of Melbourne, 2025).
- Executives who treat AI as an operating model shift capture value the more-than-80% still measuring licence utilisation miss.
The Australian evidence puts the shape of the problem into view. In the Australian workforce, 65% of employees report their employer uses AI, and 49% say they intentionally use AI regularly at work (KPMG & University of Melbourne, 2025). Yet only 36% of Australians are willing to trust it, and 48% admit to using it in ways that contravene their organisation’s policies. This is the Australian AI trust gap in numbers: employees are inside the tools without being inside the change, and the executives who signed the business case are usually the last to see that gap.
AI tools are the easy part
The dominant explanation for stalled adoption is that AI is complicated, unfamiliar or immature. That explanation does not survive contact with the evidence. Prosci’s study of 1,107 change practitioners found that 98% of organisations rate AI as valuable and 94% describe the tools as easy to use (Prosci, 2026). User proficiency, meaning how people learn to prompt the tools and rely on them for real work, accounts for around 38% of reported implementation difficulty. Purely technical issues account for 16%. The barrier is more than twice as human as it is technical, which is why AI adoption is a change management problem before it is a technology problem.
The structural pattern beneath this repeats across sectors. Executives approve an AI investment against a productivity thesis. The tools are deployed to individual desks, training is offered and policies are published. Then the work itself, meaning the tasks people are actually measured on and the approval cycles they operate inside, is left exactly where it was. The pilot succeeds. Enterprise adoption stalls.
McKinsey’s State of AI finds that 88% of organisations are experimenting with AI, but more than 80% report their organisations are not seeing tangible enterprise-level earnings before interest and tax (EBIT) impact from generative AI (McKinsey & Company, 2025). Workflow redesign has the single largest effect on EBIT impact from generative AI, and high performers are 2.8 times more likely to fundamentally redesign their workflows than low performers. The gap between experimenting with AI and getting value from it is not a gap in capability. It is a gap in what the organisation was willing to redesign to accommodate what it bought.
The AI investment gets governed as a technology decision. The business case is written by the Chief Information Officer (CIO) or Chief Data Officer (CDO), procurement runs through IT, and training is delivered as feature enablement — how to use the tool, not how the work is now expected to run. The performance framework and the approval workflows sit with different accountable owners on different governance cycles, sometimes in different divisions. The tool arrives. The work does not move to meet it.
Trust conditions produce the same effect through a different route. Where employees are unclear about what output they are allowed to rely on, which decisions the AI is permitted to influence and what happens if the AI is wrong, the rational response is to keep the AI at arm’s length. In the same KPMG and University of Melbourne survey, 57% of Australian employees rely on AI output without evaluating it for accuracy, and 59% report making mistakes in their work because of AI (KPMG & University of Melbourne, 2025). The adoption number and the mistake number are the same problem: people using tools without a work system that tells them what safe use looks like.
Adoption requires redesigning the work
If the constraint is structural rather than technical, that reframes what the AI investment actually was. The organisation did not buy productivity. It bought a set of capabilities that will produce productivity only when the work has been redesigned to use them, and until that redesign happens, the investment is producing volatility rather than value. That reframe also changes what “adoption” means. Adoption is not the licence utilisation rate on the CIO’s dashboard. It is the point at which the work looks materially different because the AI is present.
The cost of leaving the constraint in place is where the executive stakes become concrete. Each cycle the work architecture stays where it was, three things get worse together. The productivity thesis in the original business case loses more of its credibility with the board because the numbers do not move. The workforce becomes more polarised between those experimenting with the AI in undocumented ways and those refusing to touch it, so risk exposure and inconsistency grow at the same time. And the operating model that would have to be redesigned to unlock value becomes harder to change, because more workarounds and shadow practices have accumulated around the current shape.
Writer’s 2026 enterprise survey reports that 79% of organisations now face AI adoption challenges, a double-digit increase on 2025 (Writer, 2026), and Forrester’s 2026 predictions find that 21% of AI decision-makers cite employee experience and readiness as a primary barrier (Forrester, 2026). The discourse is moving away from which AI to buy and towards whether the organisation is built to use the AI it already has.
Value comes from redesigning the work
The question the executive committee needs to answer is not “how do we increase adoption?” but “what does the work look like when the AI is present, and what has to change in the operating model, the incentives and the trust conditions to get there?” That question sits at the executive layer, not the technology layer. It cannot be answered by a procurement decision or a training uplift. It requires a decision that names how the work is structured today, what it needs to become, and who is accountable for making that change.
The organisations that are getting AI value are the ones whose executives treat the deployment as an operating model shift with a technology enabler, rather than a technology deployment wrapped in change communications. That posture is what McKinsey’s high performers share. It is also what the 79% currently struggling with adoption have not yet adopted. The adoption gap is a governance signal, and reading it correctly is what makes the next 18 months of AI investment worth the spend it will consume.
The executive who sees that the AI investment was a work-design decision from the beginning will be running a different organisation eighteen months from now than the one still measuring licence utilisation and calling it adoption.
References
- Forrester. (2026). Predictions 2026: AI moves from hype to hard hat work. Forrester Research.
- KPMG & University of Melbourne. (2025). Trust, attitudes and use of Artificial Intelligence: A global study 2025 — Australia snapshot. KPMG Australia & Melbourne Business School.
- McKinsey & Company. (2025). The state of AI: How organizations are rewiring to capture value. McKinsey & Company.
- Prosci. (2026). AI in change management. Prosci.
- Writer. (2026). Enterprise AI adoption in 2026: Why 79% face challenges despite high investment. Writer.
