The Matchstick and the Body
By Jonas Astrup Beck · April 2026
In my first article, I argued that AI applied to work you haven’t studied just does the wrong things faster — and that the real opportunity only begins when the people who understand the problem can use AI to build solutions. In the second, I went deeper into the system conditions that determine whether AI helps or harms: demand nobody has studied, targets that drive the wrong behaviour, work broken across silos, and decision authority removed from the people closest to the problem. I introduced a way of thinking about where AI actually belongs — in the value work, not as a layer of control. In the third, I asked you to go and look at your own system — and showed what happens when someone actually does.
This article is about the people inside all of that. The ones expected to use AI. The ones burning out from it. And the ones making decisions about how it gets used — who are at risk of the same, whether they realise it or not.
The matchstick
I want to start with an image.
A matchstick lights up fast. It burns bright. And then it goes out. You can light it again — but it needs help to ignite, and it won’t burn quite like it did the first time.
That image keeps coming back to me when I read about what is happening to software developers right now. A recent study found that 67% of developers are spending more time debugging AI-generated code than they would have spent writing it themselves (arXiv, 2025). 68% are spending more time fixing security issues created by AI. Harvard Business Review published a piece in March 2026 called “When Using AI Leads to Brain Fry” (HBR). And 77% of AI users — not just developers — report increased workload after adopting AI tools (Upwork Research Institute, 2024).
The matchstick lights up. The matchstick burns out.
This is the story the industry tells about coders: the tool produces faster than the human can think alongside it. The pace is the problem. The human becomes a bottleneck in a process that was supposed to remove bottlenecks.
But I think the story is bigger than coders. And I think the burnout looks different depending on where you sit in the organisational system — whether you are on the frontline, in sales, in legal, in underwriting, or in the leadership team making the decisions.
Two kinds of burning
I am not a developer. I am a management consultant. I work with organisations — in the public and private sector — helping them understand their systems from the customer’s perspective and redesign the work against purpose. I have been doing this for twenty years, drawing on the work of John Seddon, W. Edwards Deming, Russell Ackoff, and others.
A year ago, I was deeply sceptical about AI — especially about how it was being implemented and for what purpose. Today, I still am — but I also see far more clearly what it can do when used well. I use it every day now. And I can feel it. The same brain fry the coders are reporting. Just from a different angle.
What I feel is not just the pace — though that is real too. What hits hardest is the possibility space exploding. Tasks that were always too time-consuming to attempt — assembling twenty years of case knowledge into a connected library, testing ideas in hours instead of months — are suddenly possible. The friction on improving my ability to do the work — the work on the work — has almost disappeared. And that is genuinely powerful. It has made me better at the actual consulting work with clients.
But the brain has limits. I can physically feel it using capacity. Not only because the work is faster, but because the number of things I could meaningfully do has multiplied. Every conversation with AI opens more doors. Every prototype suggests more experiments. The possibility space doesn’t close — it expands. And the brain tries to process all of it.
Both pace and possibility are burning people out — and I suspect most people are experiencing both at once. The coder who is debugging faster than they can think is also seeing new possibilities they never had before. The knowledge worker whose possibility space has exploded is also being pulled into a faster rhythm of delivery. Both are real. Both are the matchstick.
And here is what I think gets missed: this is not only a worker problem. It is a leader and manager problem too.
Leaders face the matchstick burn from multiple directions at once. They are expected to keep pace with what AI can do. They are expected to navigate the exploding possibility space — which tools, which processes, which strategy. And they are expected to decide where AI goes and where it doesn’t — while everyone around them is either telling them to move faster or warning them about going anywhere at all.
Three cognitive loads stacked on top of each other: pace, possibility, and the pressure to decide. If you are a leader reading this, you are not outside this story. You are in it.
Even solving the right problem doesn’t protect you
Here is where I want to be careful, because this article could easily become a “just use AI properly and you’ll be fine” piece. It isn’t.
Even when the purpose is clear — even when you are using AI to understand demand, to strengthen your capability to do the work, to think more deeply about the right problems — the matchstick can still burn out. The possibility space explodes regardless of whether what you are doing is valuable.
The athlete who trains for a worthy goal can still overtrain. The surgeon who saves lives can still burn out. Having the right purpose is necessary. It is not sufficient on its own. The body still needs recovery.
What happens when nobody notices
When organisations don’t notice the burn, they tend to do one of two things. They push harder. Or they remove people entirely.
Klarna did both. In 2024, the CEO claimed their AI was doing the equivalent of 700 customer service employees’ work. They froze hiring, let roles disappear through attrition, and watched the workforce shrink from 4,500 to about 3,500 (Fortune, 2024). The market applauded. Then customer satisfaction dropped on complex interactions, the savings didn’t materialise, and by 2026 they were quietly rehiring. The CEO admitted: “Cost unfortunately seems to have been a too predominant evaluation factor… what you end up having is lower quality” (MLQ, 2026). They are not alone — two out of three companies that cut roles due to AI are already rehiring, often within six months (Inc.), and fifty-five percent of employers regret AI-related layoffs (Forrester via The Register). There is a deeper systemic logic here — about what happens when the consequence of improving the system is losing your job — but that deserves its own article.
But Klarna is the dramatic version of a failure that happens more quietly every day. The manager who sees the signs of burnout in the team but has no language or permission to raise it. The consultant whose possibility space has tripled but whose calendar hasn’t changed. The leader who is supposed to have an AI strategy by next quarter but hasn’t been given the space to understand the system well enough to know where AI belongs.
Nobody asked the manager: “what are you seeing in your team?” Nobody asked the consultant: “what’s sustainable for you right now?” Nobody asked the leader: “what do you actually understand about how the work has changed?”
Designed to notice
This is the question I keep coming back to: what would it look like if your organisation were actually designed to notice what AI is doing to its people?
Not designed to measure output. Not designed to track adoption rates. Designed to notice — to be curious about how the whole system is functioning, including the humans in it.
Most organisations are not designed this way. They are designed to promote busyness. Productivity means visible output. And AI fits perfectly into that logic — it makes people look more productive, right up until they can’t sustain it anymore.
But what if the system conditions were different?
Here is a concrete question: if your best coder discovers that the cognitive strain of working with AI means they can only do deep, AI-assisted work for two hours a day — is your organisation ready for that conversation? Not ready with a policy. Ready with genuine curiosity. What does the rest of their day look like? Do they go home? Do they pick up different work? Do they mentor others? Do they help the team understand the demand they’re working against?
I don’t know the answer. And that is exactly the point.
The answer is contextual. It depends on the work, the person, the demand, the purpose of the organisation. For a leader, it might be: stop attending the AI strategy meetings and go spend time understanding the work. For some people, the answer might genuinely be: go home and spend time with your family. For others in the same organisation, it might be: pick up the phone in customer service and learn what customers actually need.
You cannot answer a contextual question with a general answer. This is one of the oldest traps in management — believing that what works in one context can be prescribed across all contexts. It is the thinking that gave us standardised targets, best-practice frameworks, and service factories designed for efficiency rather than purpose. AI should not become the next thing we standardise across contexts that are fundamentally different.
But if your organisation is curious about how the whole is functioning — if it has the habit of studying work, of understanding demand, of listening to the people in the system — then when signs of burnout appear, you already have the capacity to understand the causes. AI might be one cause. But measures, roles, targets, silos, and the design of the work itself can be others. The causes will point you toward answers that fit your context, not someone else’s.
The organisation that notices is not the one with the best burnout policy. It is the one that understands its system well enough to see what is changing, ask why — and prevent it from happening.
The body, not the matchstick
An organisation is not a machine you can optimise by removing parts or running faster. It is more like a body — complex, adaptive, and dependent on every part functioning in relation to the whole.
AI is a new capability introduced into that body. The question is not what it can do for any single part. The question is what it does to the body as a whole — and whether the body can sustain it.
The athlete who understands recovery performs for decades. The athlete who does not burns bright and breaks down. The matchstick.
The question for every leader reading this is not “how do we use AI to do more?” It is: “are we designing a system that can sustain what AI makes possible — and are we curious enough to notice when it can’t?”
In my first article, I said: understand the problem before you automate. In my second: the system conditions determine everything — and AI amplifies whatever you already are. In my third: go and look at the work, and see for yourself. In this one, I am saying: the people in the system matter — all of them, from frontline to boardroom. And the organisation’s ability to notice what is happening to those people is not a nice-to-have. It is what determines whether AI improves the system or burns it out.
That leaves one question. If you have understood the problem, seen the system conditions, looked at the work, and recognised the human cost — what do you do now? That is where the next article begins. It starts with you.
If this resonates — if you are a leader trying to make sense of AI without losing sight of the humans in the system — I would love to hear from you. You can reach me at jonas@vanguard-consult.dk or connect with me on LinkedIn.
This article was written with AI assistance — which, given the subject matter, felt like the right thing to acknowledge.
Sources
- “From Gains to Strains: Modeling Developer Burnout with GenAI Adoption,” arXiv, 2025 — Link
- “When Using AI Leads to Brain Fry,” Harvard Business Review, March 2026 — Link
- “Upwork Study Finds Employee Workloads Rising Despite Increased C-Suite Investment in Artificial Intelligence,” Upwork Research Institute / Workplace Intelligence, 2024 — Link
- “Klarna stopped all hiring a year ago to replace workers with AI,” Fortune, 2024 — Link
- “New Report Says AI Layoffs Are Backfiring,” Inc. — Link
- “Forrester: AI Rehiring,” The Register — Link
- “Klarna CEO Admits Aggressive AI Job Cuts Went Too Far,” MLQ, 2026 — Link