AI is intensifying work – we need a new approach to recovery from burnout

by Nicola Turner, Founder of BodyMindBrain

AI is changing the way we work, pushing up productivity – and significantly increasing the risk of burnout. As humans’ working relationships with machines intensify, roles and workflows need to be redesigned to help us deal with stress. Nicola Turner, founder of BodyMindBrain, offers some insights for HR.

AI has been part of working life for some years, but until recently few studies showed its impact on everyday work. The conversation has shifted from fears of mass job losses to understanding how AI is changing how we work. Our working patterns now include a wider set of tasks and activities, and the productivity gains are clear. But so are the dangers: intensified work, longer hours, unclear boundaries around work schedules, a faster pace and greater risk of burnout.

Research from the Upwork Research Institute found that employees reported a 40% increase in productivity from using AI tools. The same study found that workers who got the most done with AI were 88% more likely to report burnout and twice as likely to consider leaving their job. Those engaging most and producing more with these tools are the ones fatiguing fastest.

A 2026 Harvard Business Review article, in conjunction with Berkeley Haas, found that AI expanded work in ways that led employees to work faster, take on a broader scope of tasks and push work into more hours of the day. Crucially, much of this was unprompted because AI made doing more feel possible, accessible and, in many cases, enjoyable.

It can seem counterintuitive to slow down voluntary productivity, but this form of overload invites us to rethink how we interact with AI at work.

Why intensification stays invisible

Intensified work over longer hours, however enjoyable, is still stress. While we are adaptive beings designed to handle stress, the problem takes a different form here – prolonged activation with few breaks and insufficient downtime.

Mobilising your body for stress is an energetic and metabolic process that draws on the same resources needed to power down and repair. Think of it like a balance sheet: when stress and recovery work together, the account is replenished. When stress activation becomes chronic, sustaining it drains the very resources repair depends on.

AI has quietly removed the friction that once enforced a natural rhythm of activation and recovery: time to think without relentless output, meetings at a human pace, natural stops for refuelling and movement or the point when people become too tired to continue. Moments of this kind allow people to slow down and regulate towards recovery. The stopping points exist because human capacity runs out. But now that AI extends what we can produce, those natural stopping points have become less clear.

A suggestion from AI research is that, because the work feels novel and enjoyable, the brain classifies the stress as positive. The same stress experienced as pressure in a difficult week is felt as engagement in a productive one. The cost is the same, so the body pays the same price whether activation is strain or flow.

People seek recovery because sustained activation eventually feels unpleasant. When it is seen as rewarding instead, that prompt disappears. The person does not just fail to notice the cost; they lose the cue that would have prompted action.

So while enthusiastic uptake and the increased productivity of using AI can look like increased engagement, the question now becomes: is the engagement sustainable?

Designing recovery and connection back in

Henley Business School’s work on AI leadership concludes that value depends on redesigning roles and workflows, not simply deploying tools. It also names the broader challenge: we have a human-machine interaction problem, not a technology problem. This distinction shifts the focus from managing AI to designing the conditions in which people and AI work together.

I see that redesign as threefold:

• The first is to build recovery into the working day. Organisations need to recreate the rhythms that human energy and focus limitations once provided naturally. Individuals need to read their own states to know when to take breaks and recalibrate their stress response. Recovery can no longer be left to chance; it must be actively designed and managed.
• The second is to protect the human interactions that AI can never replace. Conversations with colleagues, shared problem-solving, formal and informal check-ins and moments of mutual support can sometimes be seen as interruptions to productive work. In reality, they are essential to stress management and engagement.
• The third follows naturally from the first two. If work is designed without opportunities for recovery and positive human connection, AI risks becoming more than a productivity tool. It becomes the easiest relationship at work. Work design should protect the human friction worth keeping: the challenge and collaborative problem-solving that produce critical and creative thinking. That way AI remains a tool that supports how people work well together, rather than becoming a substitute for it.

One of the Upwork research findings was that many employees reported better, more trusting relationships with AI than with colleagues. This may point to something more fundamental. When people are fatigued, overwhelmed or in environments where human interaction feels effortful, they naturally gravitate to the easiest interaction. AI becomes a refuge from the fatigue and frayed relationships that better work design might have prevented. Ironically, spending more time in that refuge can deepen the cycle of prolonged activation, reduced recovery and growing disconnection.

AI is changing how work gets done. The organisations that stand to benefit most will be those that redesign work from the top down and equip people to manage recovery from the bottom up, so that wellbeing and connection remain central to sustained performance. In the age of AI, recovery is no longer simply about wellbeing. It is becoming part of how performance is designed.