Glean: AI creating hidden workload for Australian employees

Australian knowledge workers are spending an average of 6.5 hours per week overseeing AI systems rather than using them to perform productive work, according to new research from Glean's Work AI Institute.

Glean research, based on a survey of 6000 knowledge workers, found Australian employees spend around 38 per cent of their AI-related time on 'bot sitting' - the human labour required to make AI tools usable.

This hidden workload is emerging as a significant source of frustration for workers across the nation.

Much of the problem is not necessarily AI use as a concept, but rather the way organisations are implementing and managing the technology.

Employees are increasingly required to monitor AI outputs, correct mistakes and oversee automated processes, with much of that work going unrecognised by employers, according to Head of the Work AI Institute at Glean, Rebecca Hinds

"When workers are forced to use technology in a way that isn't recognised, that is frustrating, it is overwhelming, and often becomes a driver of burnout," she said.

Glean found that 43 per cent of Australian employees surveyed said they felt worn out by AI tools, a higher proportion than in the US.

Hinds argued that organisations need to recognise the human work required to make AI effective rather than assuming automation will simply eliminate labour.

"We need to be investing in technologies that actually work," she said. 

"We need to build the infrastructure within our organisations to acknowledge this, train employees on how to do this well, and reduce the unproductive bot sitting that's also going on."

AI policies could contribute to staff churn

The research also identified a potential link between the amount of bot sititng employees undertake and their likelihood of wanting to leave their organisation.

Hinds clarified that the data could not establish a causal relationship between AI-related workloads and employee turnover, but confirmed the research showed some correlation.

"When employees are reporting higher levels of bot sitting, they're also reporting a higher likelihood to want to leave the organisation," she said.

She argued that excessive AI oversight could undermine employee confidence in an organisation's broader technology strategy.

"If you're spending the better part of a full work day auditing the technology, and you're not seeing those gains translate into real business results, then it's not surprising workers want to look elsewhere," Hinds said.

The research also examined what Hinds described as 'bot sh--ing', which is the production of low-value AI-generated content, or AI slop, within organisations.

The amount of AI oversight, as well as proliferation of low-value AI-generated work, within a company could have significant implications for employee retention.

Potential for skills gap in graduates and young staff

The increasing use of AI also presents a challenge for organisations training and developing graduates and early-career employees.

AI could inadvertently remove the 'good friction' involved in the formative stages of one's career, Hinds explained.

"There are parts of work that are meant to be messy. They're meant to be full of friction," she said.

"Often, the process of doing something matters more than the outcome, in particular for new hires who are trying to build new skills and capabilities."

Automating those processes could mean younger employees reach an answer without developing the underlying skills needed to arrive at it themselves.

But the same technology could also significantly improve onboarding if used deliberately.

Hinds pointed to the ability of AI systems with access to an organisation's internal context to help new staff understand how a company operates.

It comes down to well thought out implementation of automation, while supporting new or younger workers in the early stages of their career.

Australian AI nuances

While some of the challenges identified in the research were consistent across markets, Hinds noted several regional differences in how workers interact with AI.

One of the more unusual findings was what Australians do when AI fails.

Australian workers were more likely than their counterparts in other markets to respond to an AI failure by treating the technology like a person, including saying 'please' or apologising to it.

Workers in other regions were more likely to respond by redoing the work themselves or rephrasing their requests.

Hinds attributed the Australian behaviour partly to cultural differences, but also to weaker governance and training around AI within organisations.

"If workers aren't sure how to leverage this technology effectively, they default to what they know, and that is treating the technology more like a human," she said.

This highlights a broader psychological difference in how employees across markets are adapting to AI.

Automation could reshape established corporate org charts

Looking ahead, the biggest opportunity for AI may not be individual productivity but changing how teams and entire organisations work together.

Instead of simply using AI to boost individual and collective productivity, Glean expect organisations to increasingly use the technology to measure and improve collaboration.

AI could, for example, analyse meetings to understand the distribution of speaking time or identify imbalances in participation.

Hinds posited that AI could even challenge the traditional corporate hierarchy.

"The org chart is something that has remained unchanged for decades and decades within our organisation," she said.

AI systems with access to rich organisational data could potentially analyse employees' skills, experience, relationships, interests, career ambitions and availability to determine which combination of people is best suited to a particular project.

Glean is already seeing customers use its technology to dynamically staff teams rather than relying on traditional reporting structures.

"You can start to put together really exciting and complex calculus in terms of what is the exact right mix of people from my potentially thousands of employees who are best suited to execute this specific task or project or team," Hinds said.

Glean expect that could result in significantly more dynamic organisational structures over the short to medium-term.

"There's always going to be a need for hierarchy, but I think we're going to see much more dynamic structures within our organisations," Hinds said.