Australia leads use of AI agents, but managers feel strain

Australia ranked highest among seven markets for the use of AI agents in multi-step work in new research commissioned by monday.com. The survey also found that, as AI use rises, many Australian managers are spending more time supervising output.

The Nielsen study, which surveyed 1,701 Directors and Senior Managers in large companies, found that 46% of Australian respondents use AI agents to complete multi-step work, compared with a global average of 38%. It also found that 58% expect to manage more agents than people within the next 12 months.

monday.com used the term "agent sprawl" to describe what it sees as a growing build-up of AI agents, software tools and automated hand-offs inside organisations without clear ownership, controls or measures of value.

While Australian businesses appear to be adopting the tools quickly, the research suggests managers are taking on more of the surrounding work. Some 59% of Australian leaders said they now spend more time reviewing and guiding their team's output since AI use increased, the highest result across the seven markets surveyed. A third said AI adds hours to their working day.

The findings point to a gap between deploying AI and reorganising work around it. In Australia, 62% of respondents said identifying and testing where AI creates value requires more effort than it saves.

Manager burden

The extra workload often comes from tasks around the technology rather than from operating the tools themselves. Managers still need to decide what work to delegate, provide context, connect systems, check results and correct errors, while remaining responsible for the final output.

Australian respondents also reported broader changes in their software environments. Some 65% said they are using more software tools since adopting AI, the highest share among the seven markets, while 58% said their tools have become more interdependent. More than a third, or 36%, said manual steps in workflows have increased.

That pressure comes despite relatively high confidence in AI familiarity. More than two-thirds of Australian respondents described their teams as advanced or expert AI users, and 69% said their organisation provides sufficient AI training.

"Agent count is fast becoming a vanity metric. Giving one employee ten agents doesn't make them ten times more productive if they inherit ten streams of work to supervise. Businesses need to measure whether the entire workflow moves faster, with less rework and fewer hand-offs. Every agent also needs a job description, a defined outcome, its own permissions, a clear point of human escalation and one accountable owner. Without that discipline, agent adoption simply creates more work the organisation can't see," said Gavin Watson, Regional Director, AI & Industries at monday.com.

Changing roles

The survey also indicates a shift in how managers allocate work between people and AI. Only 14% of Australian Directors and Senior Managers said they mainly assign tasks to people and use AI as a secondary tool. By contrast, 45% said they reserve work requiring human judgement for staff and shift the remaining tasks to AI, while 26% said they assume AI will handle most execution and employees will focus more on editing and strategy.

As that balance changes, concerns about oversight and accountability are rising. Nearly half of respondents, or 44%, said they fear agents will not receive enough human oversight or control. Almost one in three, or 31%, said they worry about being held accountable for an agent's work. A larger majority, 84%, said good management now requires different skills in the age of AI.

The issue is attracting wider attention in Australia as policymakers and public agencies examine how organisations use AI systems and how much autonomy they should have. The survey results come amid closer scrutiny of agent access, human oversight, productivity effects and workforce implications.

Performance reviews

The research also suggests AI is changing how employers assess staff performance. Four in five Australian organisations surveyed said they have changed the way they evaluate employees since adopting AI, and 37% said they have significantly changed their evaluation criteria, the highest result among the seven markets.

Nearly six in ten respondents said an employee's original contribution has become harder to evaluate when work passes through several AI tools or AI-supported processes. The survey also found that 40% of organisations use AI to draft performance reviews.

Watson said those changes could make individual judgement less visible inside businesses.

"Agent sprawl makes work harder to coordinate and talent harder to see. When several agents help research, draft and assess the same output, polish and speed stop being reliable evidence of ability. Businesses risk rewarding the best-presented work rather than the strongest judgement, while overlooking the people who set the direction, challenged the answer, caught the mistake and took responsibility for the result. Unless performance systems capture those decisions, companies will lose sight of where human capability actually sits," Watson said.

The Nielsen survey covered respondents in the US, UK, Australia, Germany, France, Brazil and Singapore, and focused on companies with at least 500 employees.