AI fatigue at work: Why it happens and how to manage it

Artificial intelligence is proving to be a great productivity tool that helps to manage routine work and free professionals up for the tasks that require a human mind. But what if the tools built to lighten cognitive load are adding to it?

Harvard Business Review calls the mental fatigue that builds up from spending too much time interacting with such tools AI “brain fry”. For finance and business leaders who have been pushing AI adoption as a productivity lever, poorly implemented strategies may be creating a new form of overload — one where employees are switching between tools and managing increasingly complex workflows.

Dr David Tuffley, technology communicator and senior fellow of the Higher Education Academy at Griffith University, Queensland, Australia, says the problem often starts with volume. Trying to fold AI-generated material into a final piece of work, he says, is where the exhaustion sets in.

“It can very easily turn into an overwhelming amount of data, and if you think you should be bringing all of it into the discussion, or into the outputs you are creating, it can be stressful,” Tuffley says. “If people go about it the right way, then there are real gains to be made here. But equally, going about it wrong way is counterproductive.”

Workplace expert Michelle Gibbings points to a potential downside of more tech tools: increased neural switching. Despite a professional’s confidence in their own multitasking, the research clearly shows that humans can lose effectiveness when they constantly jump between tasks, she argues.

“When we are simultaneously managing multiple AI tools, toggling between tasks and constantly reviewing outputs, we are placing greater demands on our attention,” says Gibbings. “We are touching more elements and working at a much faster pace than in the past, and this constant switching can increase cognitive load and leave us feeling mentally fatigued by the end of the day.”

When AI works

AI can produce in seconds what once took hours, but a fast answer is not always a finished piece of work. The risk is greatest in organisations that treat the technology purely as an accelerator, with no thought given to how it should be used, says Tuffley.

Without guardrails — for example, which tools to use for what purpose and what good output looks like — AI becomes another layer of work rather than a replacement for it. “People end up wading through AI-generated material, cross-checking outputs and comparing versions, all while under pressure to deliver faster than before,” says Tuffley.

“If it is taking people longer to do their jobs by using the technology, then that is a recipe for overload.”

When AI removes work rather than merely speeding it up — for example, summarizing large datasets, sorting through volumes of information and spotting patterns — the payoff is real. That frees professionals to do the parts of the job that are distinctly human, he says, such as relationship-building, judgement and higher-order analysis.

The challenge, says Gibbings, is that some tools are too agreeable. “If you go back and ask it to check something, it can sometimes reinforce your view rather than genuinely challenge it.” Therefore, it is important that professionals craft the correct prompt and also understand what they can do to challenge the output rather than take it at face value.

“There is no doubt that AI can help,” she says. “There are things that used to take hours that you can now do very, very quickly.” However, the real work is deciding where AI belongs in a given process and where human judgement and oversight are still needed. “That is a conversation for each organisation, because guardrails are really important in all of this.”

For leaders, that means being deliberate about that they discuss with their teams. How people approach it varies, she says. Some organisations are being explicit about when and where employees should use the technology; others have people who are simply nervous to use AI at all.

The role of leaders is to build an environment where people feel comfortable experimenting with AI, Gibbings says, while ensuring appropriate human oversight to verify the output is legitimate and accurate.

“If you are not using it, ask ‘Why not?’,” she says. “How can you experiment in a way that is safe and in line with company guardrails, to see where it might help you with your work?”

Ask the right questions

Being overwhelmed by the possibilities AI offers is easily avoided, says Tuffley. He outlines a four-part framework for constraining outputs and steering clear of a data rabbit hole: “Tell the tool who you are, who the audience is, what you want and what level of language to use.

“Asking the right questions is absolutely critical,” he continues. “You do not get good answers unless you ask good questions first.”

That means keeping the scope narrow.

“If the use of AI is disciplined and very specific, then yes, there are big benefits,” Tuffley says. “But if it is adopted in an undisciplined way, then it tends to create more problems than it solves.”

 

_{Production credit
Banner image Moor Studio/DigitalVision via Getty Images}