4 in 10 Australian workers say AI tasks fail outright, forcing major rework: Report

Nearly four in 10 Australian workers say their AI sessions "fail outright", forcing them to restart tasks, undertake substantial rework or abandon the output altogether, according to new research from Glean’s Work AI Institute.

The findings raise questions about whether widespread AI adoption is delivering the productivity gains many organisations expected.

The study found that 90% of Australian digital workers use AI for work tasks. Yet only 10% believe the technology has significantly improved organisational performance.

Hours spent supervising

Australian workers spend an average of 6.5 hours a week "botsitting" AI, the report found.

Botsitting includes providing missing context, checking outputs, fixing errors, rerunning prompts and correcting inaccurate responses.

Some 73% of workers said they have corrected or redone AI-assisted work in the past month, while 30% do so at least weekly. Another 45% admitted they had submitted AI-generated work they could not fully explain.

"If we’re having to babysit (AI) for six and a half hours to make sure it does the right thing, you’re probably using it for the wrong thing," said Work Futurist and Work AI Institute expert Dom Price.

"If something used to take me seven hours and now it takes me half an hour, but I then spend the other six and a half hours looking after the bot, I’ve saved no actual time,” he added.

Organisation lag

While 72% of Australian AI users said the technology improves their personal productivity, only 10% reported a significant improvement in organisational performance.

More than half of respondents, 56%, said important workplace information is not connected to or accessible through AI tools.

Price said many organisations are introducing AI without redesigning the processes around it.

"Australia has always been good at moving quickly and making new technologies work in practical ways," he said.

"But AI is different because it doesn’t just ask organisations to adopt another tool – it asks them to change how work gets done. Right now, too many companies are trying to push AI-speed change through legacy-speed systems."

He warned that businesses risk accelerating inefficient processes rather than improving them.

"They say, ‘We have this terrible process that we’ve never liked. Let’s put AI all over it and let’s do it faster’.

"My advice is: Don’t do the stupid thing faster. Instead, stop doing the stupid thing. Find a better way of doing it and then use AI as the accelerant."

Growing fatigue

The report also highlighted rising frustration among workers using AI.

Some 43% of Australian digital workers said they feel worn out by AI tools. Frequent botsitters were 59% more likely to be actively looking for a new job, while workers who submitted unchecked AI-generated work were 3.4 times more likely to be job hunting.

The findings suggest that while AI is becoming a standard workplace tool, many organisations have yet to bridge the gap between individual adoption and meaningful business outcomes.