AI fatigue: When artificial intelligence makes you tired

Between the promise of productivity and mental exhaustion

Far from freeing workers from the mental load of low-effort tasks, research shows AI is actually making them feel more overwhelmed.

Artificial intelligence should make work easier, speed up processes and relieve people from stress. In practice, the opposite is true: employees feel mentally exhausted, emotionally pressured and cognitively overwhelmed by the technology’s ubiquitous use.

This phenomenon, described as AI fatigue or AI exhaustion, has long since ceased to be a marginal issue, but a structural risk for companies.

What AI fatigue means and how it manifests

AI fatigue describes a feeling of persistent overwhelm that arises from the constant confrontation with AI systems, AI-generated content and the implicit pressure of expectation to constantly use these technologies productively.

Typical symptoms range from mental exhaustion, to poor decision-making and concentration, to frustration and inner withdrawal. The burden is particularly pronounced among younger employees. Professional services firm EY sees the causes as a mixture of the growing fear of job loss and the feeling of having to do more and more despite – or because of – AI support.

A central driver of AI fatigue is the permanent hype: new models, co-pilots and agents appear in ever shorter cycles. Users are under pressure to keep up, often without clear prioritisation or strategic guardrails.

This barrage of hype requires high learning and adaptation efforts as tools, interfaces and best practices are constantly changing. At the same time, the quality of AI outputs suffers. Among other things, this results in so-called workslop – extensive human rework due to erroneous, generic or context-poor results.

The reversal of Parkinson's Law

A study by UC Berkeley, published in the Harvard Business Review, found that while generative AI speeds up individual tasks, it leads to more work overall.

Software engineer Siddhant Khare describes this phenomenon particularly succinctly. In his essay AI-Fatigue is real and nobody talks about it, he reports that he has shipped more code than ever before – and at the same time feels more drained than in any other quarter of his career.

Khare says the reason is not the technology itself, but in changed working patterns. AI lowers the cost of individual tasks but raises expectations for speed and scope.

"Before AI, my job was: think about a problem, write code, test it, ship it. I was the creator. The maker. That's what drew most of us to engineering in the first place - the act of building.

After AI, my job increasingly became: prompt, wait, read output, evaluate output, decide if output is correct, decide if output is safe, decide if output matches the architecture, fix the parts that don't, re-prompt, repeat. I became a reviewer. A judge. A quality inspector on an assembly line that never stops.”

Parkinson's Law states that work exactly to fill the time available. Tasks fit the allotted time frame, regardless of their actual complexity.

With AI, the principle seems to be reversed: AI sets the pace and forces us to adapt the time frame to the completion of the task, and thus complete more tasks in the total time available (aka working day) – a process that basically began with industrialisation.

What does this mean in practice?

It’s not just workloads that are expanding. Contexts are also changing faster, and humans are moving from designers to permanent examiners of a never-ending AI pipeline. The result is cognitive exhaustion due to constant multitasking and a lack of phases of deep concentration.

Khare sums it up: “AI reduces the cost of production but increases the cost of coordination, review, and decision-making. And those costs fall entirely on the human.”

“The cruel irony,” he writes, “is that AI-generated code requires more careful review than human-written code.”

For companies, AI fatigue is more than just an individual well-being problem.

When employees feel dehumanised because algorithms set the pace, motivation, creativity and quality decrease. The Harvard study explicitly points out that the initial productivity boost from AI can turn into burnout, fluctuation and declining decision-making quality if clear rules are not established.

Disillusionment is also growing at management level. According to EY many executives feel overwhelmed by the speed of AI development. They see workforce acceptance declining, despite increasing investments.

Ways out of exhaustion

Practical reports show clear countermeasures. Being transparent about what how you’re using AI – and how you're not – is crucial. Unrealistic promises of salvation increase fatigue, while clearly defined use cases provide orientation.

Targeted training and continuous support are just as important. Employees not only need tool instructions, but also guidelines for sensible use, quality controls and conscious breaks. In this context, the Harvard researchers speak of a necessary "AI practice" – organisational rules that structure and limit the use of AI.

Finally, the focus is on a human-centred approach. AI should complement human capabilities, not displace them.

Khare advocates "sustainable output" instead of maximum emissions: performance that leaves room for thinking, learning and recreation. This is the only way to prevent productivity from increasing in the short term but undermining health and innovation in the long term.

Here are some key questions for IT leaders trying to assess the governance, risk and security factors of AI fatigues:

• Are AI tools, co-pilots and agents centrally regulated (deployment, authorisations, etc.) and inventoried?
• Is AI fatigue part of risk assessments and audits? Are overloads and wrong decisions recognised as safety risks?
• Are there mandatory testing mechanisms for safety-relevant AI results?
• Is error culture anchored in your organisation’s DNA?
• Are teams specifically trained to deal with hallucinations, bias and misinterpretations?
• Is human-in-the-loop an integral part of the AI strategy? Are critical decisions deliberately not fully automated?
• Are there guardrails against permanent alerting, output overload and decision-making stress?

AI fatigue is a warning sign. It shows that technological performance and human resilience do not automatically go in lockstep.

For companies, this means AI’s success is determined less by models and computing power than by leadership, culture and realistic expectations. If you see AI as a tool – and not as a permanent accelerator – you can increase productivity without exhausting people.

This article was originally published on Computing Deutschland.