AI is often sold on the promise of increased productivity. Companies claim the technology can assist with, or even replace, white collar roles. Indeed, adoption has been quick for many businesses. However, research shows that a lack of AI workplace trust means the technology might not save as much time as you would think.
Six in ten Australian workers have personally caught an AI tool getting something wrong at work, and only 14% say they fully trust what the technology produces. The extra legwork to verify AI output, rework mistakes, and take turns to get the desired result robs much of the time savings.
These numbers raise the question, does AI really save time in practice or does it simply reframe the work employees have to take on? Here’s what the research has to say about AI productivity.
Checking output takes time
Appian, which sells AI governance and automation software, commissioned Zoho Research to survey 500 Australian workers in Q3 2026. The research found 60% of respondents identified an error made by AI at work, and more than 20% say employees at their workplace are discouraged from challenging AI’s recommendations or decisions.
The high rate of identified errors, extremely low trust, and discouragement around employee oversight makes clear that there’s fundamental issues with adoption of the technology in business organisations.
Professionals Australia conducted an AI At Work survey which polled more than 2000 of its own members, weighted towards engineering and ICT. It found 63% of respondents say AI decision making at their workplace lacks transparency, 71% had not been consulted before AI tools were introduced to their role, and fewer than 20% had received any formal training on AI usage.
“It is like managing a junior colleague who never learns from their mistakes,” said one engineer responding to the survey.
Together these studies show AI use is widespread in workplaces, but trust in its output quality is low. Left on their own, employees using the technology demonstrate how a lack of guidelines defeats the purpose of the technology in the first place.
Compliance issues are widespread
When AI output is treated as authoritative with minimal oversight as shown above, risk to the business itself is high.
Another independently run study by KPMG and the University of Melbourne found 59% of Australian respondents had made mistakes at work because of AI, and 57% had relied on AI output without evaluating its accuracy. 46% admitted using it in inappropriate ways, and 44% had used it in ways that contravened workplace policies.
As long as there’s a continued lack of regulation from government and private business alike, the technology will cause issues inside workplaces.
Time saved does not equal productivity
Global research from workplace AI company Workday, which surveyed more than 3200 full-time employees at large organisations around the world, shows usage is high, but the time savings don’t add up.
Its report found that nearly 40% of time AI saves is spent reworking AI outputs, like correcting errors, rewriting content, and verifying output for accuracy. Only 14% of employees reported a clear, positive net result from using AI.
AI workplace trust issues
Across the Australian data, there is demonstrable demand for better infrastructure around AI. 94% of Appian’s respondents said AI should not be allowed to make decisions affecting employees or customers without limits or human oversight, and 66% want such decisions to require human review and approval.
Keeping a person in the workflow isn’t enough on its own. The people responsible for oversight need the training and authority to act on problems, and clear rules are necessary on when, how, and why oversight action happens.
AI productivity gains might look great on paper, but in practice the improvements are overstated. Until organisations start implementing better training, guidelines, and metrics on AI usage, AI can’t deliver on the promises it’s sold on.
To keep up with the latest news, advice, and reviews, consider subscribing to our newsletter.
Comments