AI helps and speeds things up, but what is the impact on employees?

Exponential change often only feels urgent to people when it is almost too late. That is the lesson of the well-known French children’s riddle about the pond with water lilies, cited by Donella Meadows: on the day before the pond is completely covered, only half of it is covered. AI increasingly works exactly like that in the workplace. For a long time, little seems to be happening, until the acceleration is suddenly felt everywhere: in pace, expectations and mental load.

This is no longer a theoretical point. Deloitte reported in early 2026 that organisations increased employee access to AI by 50% in one year, and that 34% of companies now use AI to transform their business in depth. The World Economic Forum describes the same movement: AI is shifting from separate experiments to integration in core processes, decision-making and operating models. AI is therefore no longer a separate tool, but more and more a system layer underneath the work itself.

That has benefits. The OECD points out that AI at work can contribute to higher productivity, better job quality and stronger working conditions. The scientific picture is also less simple than many headlines suggest. In a longitudinal study among German employees, Omar Giuntella and colleagues found no evidence yet of a broad negative effect of AI exposure on mental wellbeing. On some points they even saw slight health gains. If you want to be honest, you therefore have to say two things at once: AI demonstrably helps and speeds things up, but the impact on employees depends heavily on how organisations introduce it and on what is asked of people in the meantime.

At the same time, the WHO has long warned about psychosocial risks at work such as excessive workload or pace, lack of control, unclear roles, limited support, long working hours and job insecurity. Exactly those factors can be amplified by AI acceleration. And according to recent research into generative AI, something else comes on top: new forms of technostress, such as compliance uncertainty, privacy concerns, dependence on AI, doubts about the reliability of output, and the risk that craftsmanship erodes when people mainly monitor and correct.

In practice, we see several groups of employees emerging. These are not fixed personality types, but positions people can end up in as soon as AI not only supports the work, but also starts to set the rhythm of work itself.

1. The front-runners: people who can keep up at full speed

These are the people who work daily with tools such as Claude Code, ChatGPT Codex and similar, build their own automations, refine prompts, keep track of new tooling and are often the first to see what is possible. They usually benefit fastest from the productivity gains. But this group is also at risk. Whoever can handle a lot often gets even more on their plate. Work becomes faster, more diffuse and less bounded. Research by Högemann and colleagues shows that, alongside efficiency gains, generative AI can also trigger specific technostress. The practical metaphor of the “AI vampire”, described by Eric Porres in March 2026, is therefore apt: AI gives energy and output, but also drains energy. It can drain sleep, recovery and attention at the same time, precisely in the people who work with it most effectively. New tools, new features and new possibilities every day demand continuous adjustment.

2. The followers: people who keep up for a while, but eventually burn out

This is often the most vulnerable group. These employees do their best, attend training, test tools, use Copilot, ChatGPT or other instruments and, on paper, simply keep delivering. But underneath that performance a second layer of load can develop: constantly having to learn, checking output, interpreting new working agreements and at the same time proving they are not falling behind. The WHO risks become tangible here: a high pace of work, role ambiguity, too little control and insufficient recovery. What still looks like adaptability from the outside may already be moving towards exhaustion on the inside.

3. The holdouts: people who don’t go along, don’t want to, or come too late

You should not be too quick to dismiss this group as old-fashioned or unmotivated. Behind not moving along there may be reluctance, but also uncertainty, shame, loss of control, a lack of psychological safety or simply a mismatch in training and guidance. And more importantly: not taking part is no longer a neutral position. When AI changes the norm, the workflow and the pressure to perform, employees who hardly use AI themselves are affected too. After all, the system around them does change. That is exactly why AI is not only an innovation issue, but also an issue of inclusion, sustainable employability and mental safety.

The biggest mistake organisations can make now is to treat this development as an individual adjustment problem. Then group 1 gets applause, group 2 gets a course in AI or time management, and group 3 gets blamed for not wanting to move forward. But mental load caused by AI is usually not an individual defect. It is an organisational dynamic. As Bart van der Sloot describes in 2026, AI ultimately is not only about efficiency, but also about how digital systems amplify old tensions around autonomy, being human and ambiguity. So the question is not only what AI can do, but also what it does to people when work changes faster than they can process.

For employers, this means AI does not only belong with IT, innovation or procurement. It also belongs in the conversation about mental wellbeing and occupational healthcare. Not because AI makes people ill by definition, but because its introduction can amplify the known risk factors for mental overload. The relevant question is therefore not only how much faster the work goes, but also what that does to autonomy, recovery, clarity, social support and sustainable employability.

At Healthguard, we help organisations ask that question in time. Every day we see the effects of overload and (work) stress in the consulting room. We make AI acceleration open to discussion as part of psychosocial workload. We help employers see where the real pressure arises: in pace of work, availability, rising standards, learning load, role ambiguity or loss of control. We support managers in making clear agreements about expectations, responsibilities and boundaries. And we guide employees who get stuck early and in a targeted way, through preventive support, coaching and mental health support.

We explicitly do not do this only at the individual level. Because if you take AI seriously, you also have to take the system seriously. Sometimes the problem lies with the employee who “struggles to keep up”, and sometimes with an organisation that wants to move too fast, where frameworks and guidance still fall short. That is exactly where good occupational healthcare adds value: not stepping in only when someone drops out, but spotting earlier where work, pace and people get out of balance.

So AI does indeed help and speed things up. And with good early detection and guidance of employees, that return only becomes truly optimal.