The essentials in 30 seconds:
→ The same AI can relieve or burden you — depending on two questions, not on the tool itself.
→ Question 1: Did I choose this myself, or was it decided for me?
→ Question 2: Does the AI take over the task, or do we work together?
→ Bakker et al. (2026) map this matrix onto the well-established Job Demands-Resources model — AI can raise or lower demands, and build up or deplete resources.
→ Those with few resources and little time easily fall into a "loss spiral": too little capacity to use AI meaningfully — and they fall further behind.
It frees up resources for other things and reduces planning stress — for instance, when coordinating appointments or simply organising sequence and timing. Self-determined, a positive human-technology combination.
The two questions that make the difference
That is exactly what Arnold Bakker and colleagues (2026) describe by mapping the Job Demands-Resources model onto a 2×2 matrix. Simplified: AI can mean relief or burden. It can raise or lower demands, build up or deplete resources. It comes down to two questions:
Did I choose this myself, or was it decided for me? And: Does the AI take over the task, or do we work together?
These two axes determine whether the same technology affects two teams in completely different ways — not the tool itself.
An example from practice
I notice this in my own work when I set up reporting and information systems with clients to improve communication across locations. Using structural templates, AI also helps those who cannot express themselves as eloquently or concisely, even though their information and decision basis is entirely sound. It makes things more readable and comparable for everyone involved.
This relieves site managers, for example, because information and reporting are not seen as core tasks in themselves. Once people notice that information gets refined and prepared quickly — and receive positive feedback for it — it helps several people coordinate and collaborate. Also self-determined, a human-technology combination.
When it doesn't work
It gets stressful when the use of AI doesn't pan out. An eloquent-sounding holiday packing list that you still have to correct yourself. Self-determined, automated — but annoying. Then again: maybe the prompt and the context were simply too vague. "Holiday by the sea" makes quite a difference between the North Sea and the Mediterranean.
The reason: AI doesn't know the nuances, the tone, the unsaid between the lines. It's the same at work. You end up with extra work instead of relief.
The loss spiral
The real point is this: you need time, you need to be able to take the time, to generate the positive effects of using AI at all. If you have no time, if you're under pressure, you probably won't experience the process as particularly worthwhile. It becomes a burden instead.
Bakker and colleagues call this a loss spiral: those with few resources also have little capacity to use AI meaningfully — and fall further behind as a result. This tends to hit exactly the teams that would need relief the most.
AI is a job design issue
AI can help a great deal. Used well, it can be a resource. Rely on it exclusively, or invest in the wrong place, and it becomes a demand — a burden. The difference is not decided by the tool, but by how much choice and how much time for reflection a team is given.
A question to think further about:
How much time do employees actually get to figure out for themselves what works for them and what doesn't? That is exactly what decides which side of the matrix a team ends up on.
What this means for Swiss organizations
In smaller organizations, AI adoption is often decided almost incidentally — a tool gets rolled out, and how it's used is left entirely up to the individual. According to the JD-R model, that's exactly the risky part: without choice and without time to experiment, a team is more likely to slide into the loss spiral than into relief.
What helps doesn't require a major initiative. It's enough to deliberately ask two questions before introducing a tool: Can the people affected help decide whether and how they use it? And does the AI take over a task completely, or does it remain collaboration? That clarity often determines success more than the choice of tool itself.
References
Bakker, A. B., Junker, T. L., Scharp, Y. S., & Černe, M. (2026). Artificial intelligence and job design: Insights from job demands–resources theory. Technology in Society, Article 103533.
How does your team handle AI?
30 minutes, no sales pitch. You tell me where AI relieves your team and where it tends to burden it. I listen, then straight talk.