AI is changing how we think at work. Is HR paying attention?

Stephen Joyce, chief people and operations officer at Forge Holiday Group, is proud to be leading an AI-forward organisation. The business, which runs holiday brands and platforms including Sykes Holiday Cottages and Forest Holidays, has achieved high AI adoption, even across frontline roles (cleaners use it to decide the most efficient order to clean properties in, for example), and embedded AI into workflows. But as AI changes how work gets done, Joyce has noticed something concerning.

In the firm’s ‘Velocity Lab’, the developer responsible for testing AI workflows is exhibiting signs of cognitive overload. “Come 1pm, he’s shattered,” Joyce says. “We’ve got four or five AI agents, working through stuff, arguing with each other, all playing different roles. Watching those agents and checking [the outputs] is quite draining.”

Joyce’s observations are backed up by research. A study from BCG, published in Harvard Business Review, coined the term “brain fry” to describe the specific cognitive fatigue and overload caused by managing, monitoring and switching between multiple AI tools. Brain fry negatively impacts decision-making and the ability to focus, the study says, as well as causing headaches.

Such side effects raise an urgent question: we may be measuring what AI does for productivity – but who is measuring what it does to our minds? As many businesses rush towards efficiency at all costs, are leaders neglecting the longer term risks of cognitive overload, cognitive atrophy and skill degradation?

Joyce is already considering how to mitigate against cognitive overload, exploring creative solutions. “What if [that developer] doesn’t do a 37-hour week, but still gets paid for it?” he muses. “What if he only does 25 hours, but is at full tilt for those 25 hours? Could he spend time learning, because he still needs to develop his coding skills? Do we give him a day a week to work on something for himself, like a side hustle, as long as he’s improving his skills? This is the real life stuff I’m thinking about right now.”

From cognitive overload to cognitive atrophy

Beyond cognitive overload, there are other risks to cognitive capability that HR leaders need to be aware of, before AI usage becomes even more codified within organisations. Gartner has found that one in five workers already say they can’t do their jobs without AI. Is that a sign of usage supercharging productivity, or of skill loss and overreliance on a tool that thinks for you?

A 2026 paper from The Wharton School suggests that it may be the latter. It refers to "cognitive surrender”, defined by researchers as “a deeper abdication of critical evaluation”, whereby users adopt AI judgement as their own. Another recent study from MIT Media Lab entitled ‘Your Brain on ChatGPT’ found that LLM users exhibited the “weakest [brain] connectivity” when essay writing, compared to those using a search engine or only their brains. “The LLM undeniably reduced the friction involved… However, this convenience came at a cognitive cost,” the paper says. (It is worth noting that this study is yet to be peer reviewed and is currently subject to a formal critique.)

“There's nothing inherently wrong with delegating cognitive work to AI,” says Anne-Laure Le Cunff, a neuroscientist at King's College London, the founder of Ness Labs and the author of Tiny Experiments. “We do this all the time with calculators, search engines, and other tools,” she points out. “The question is what we're choosing to offload. Are we freeing up mental capacity for deeper thinking, or are we outsourcing the very processes that help us learn and develop expertise?” That is arguably the question that HR teams need to be curious and challenging enough to both ask and attempt to answer.

How cognitive offloading can weaken critical thinking

Critical thinking, for instance, is routinely cited as a core and uniquely, irreplaceably human skill. But what if the very way we are using AI is impairing our critical thinking capability? Michael Gerlich is the author of The Convenience Trap: What happens when AI becomes the mind behind our lives? His research has flagged “a significant negative correlation between frequent AI tool usage and critical thinking abilities, mediated by increased cognitive offloading”.

“Too many companies consider AI like a software,” Gerlich explains. “But in using it we tend towards cognitive offloading. You have to put friction in the process. You cannot have AI push the process faster and faster. That leads to efficiency but mediocre results, and it pushes the human out. We need the human to think first, and companies are not doing that.”

Guardrails around usage and training employees to use AI in the most appropriate way are key organisational responsibilities, Gerlich believes. “We have to change the sequence,” he adds. That means managers and leaders guiding people to use AI in a different way. “The sequence is important,” Gerlich says. “The human has to come first and decide what repetitive work AI can do, bringing the critical thinking and questioning.”

CV Viverito, senior director analyst at Gartner, likens using AI in the wrong place in a workflow as “going to the gym and bringing a robot to lift the weights for you: you’re not going to get any stronger.” In the workplace that translates as: “AI is not going to write your first draft. You should write your first draft.” The problem is, in many workplaces this simply isn’t happening, with employees under time pressure to deliver and often turning to AI slop in order to get the job done.

“For the past 15 years in HR we’ve been focusing on making the employee experience as frictionless as possible,” reflects Viverito’s colleague Kaelyn Lowmaster, director analyst at Gartner. “Now we are trying to make things that should be hard easy. That’s where we start to see skill degradation. If our goals are speed-based or output-based, that crowds out some of the tough, frictionful work that we want employees to do.” Gartner research has found that 60% of organisations do not have a strategy in place to prevent AI-driven skills atrophy.

Of course, not all skill degradation or loss is automatically bad. Few people read paper maps any more, for example, relying on GPS instead. “We have got to get focused on which skills are important for us to develop and which skills are okay to atrophy,” Lowmaster says. This presents another opportunity for HR professionals to join the dots between AI, future workforce planning and organisation design.

How HR can build intentional friction into AI workflows

When Ajay Sharma, a global HR executive in a technology firm and a member of the CIPD’s AI Advisory Board, works with teams on AI adoption, he is firm about the need to start in the right place. “Where AI is the right answer, the instinct is to open the tool first and ask it what to do,” he says. “The functional expert then reviews the output, edits a few lines and moves on.” Instead, as Gerlich advises, Sharma guides people to “frame the problem, set out the options and form a first view before AI enters the room”. “A rough brainstorm on a whiteboard is enough,” he adds.

Once the expert has a view, AI can act like a colleague and sparring partner, questioning that view and widening thinking. Sharma uses prompts like “argue the opposite case as strongly as you can”, “what would a sceptical CFO ask?” and “give me three alternatives I have not considered”. He also advises agreeing “AI-free moments” with business leaders, writing these into job descriptions and ways of working, and asking people to note their decision-making processes.

Le Cunff says leaders must role model curiosity, encouraging employees to “use AI to challenge their assumptions, explore alternative perspectives and ask better questions, rather than simply generate answers”. Echoing Gerlich and Lowmaster, she also advocates building intentional friction into workflows rather than completely smoothing off the rough edges and tough challenges that often force people into higher quality thinking. And, she adds, leaders should “reward learning, not just efficiency”..

Viverito suggests that managers use the last few minutes of team meetings to discuss and test how colleagues are using AI. “Have someone plug in how they would typically use AI and dissect the output as a group,” they advise. “Look for its weaknesses and strengths. Discuss: would you send this to a client or not? Build in a layer of AI judgement to stop people simply outsourcing their answers to AI.”

When AI can strengthen cognitive wellbeing

Back at Forge Holiday Group, every department has an “AI genie”, a subject matter expert with an aptitude for AI. “They work with people around their ideas and what they think AI could help with – where they should or shouldn't use AI as part of that process,” Joyce explains. He adds that he is optimistic about AI’s ability to free up capacity, reflecting that embedding agentic AI into HR has freed the team up to exercise higher quality emotional intelligence and judgement. He has been able to expand his role into COO, thanks to AI taking some of the load.

The BCG study on AI brain fry also found that there are cognitive benefits to using AI in the right way. Burnout scores fell and social connection increased when AI replaced routine, repetitive tasks, suggesting that oversight-heavy work is most fatiguing. Viverito cites a study that found practising a hard conversation with AI first makes people 10% more likely to have it, boosting valuable human skills and providing a high quality learning experience – brain gain rather than drain.

The task for HR teams now is ensuring that the need for speed doesn’t crowd out the need to protect those cognitive capabilities that matter most. As Gerlich says: “This isn’t the fastest way, but it will be where AI-plus-humans create the highest quality work.”