I can produce in an hour what used to take me weeks. I choose this every day, and I’ll keep choosing it.
But something has been nagging at me for months, and I don’t think we’re being honest enough about it.
Work used to grow slowly. A small idea would turn over in my mind, resist, push back, gradually become something I could call mine. Now it materialises in minutes. The output is often better than what I’d have produced alone. That’s not the issue.
The issue is that I’m no longer inside the thinking. I’ve become a validator of generated thoughts, a supervisor of someone else’s reasoning. The process that used to be the site of engagement, of pride, of the slow satisfaction of an idea becoming mine, has been replaced by something faster and, on some days, strangely hollow.
I’m not a sceptic writing from the outside. I spend my days as a domain architect in Enterprise AI at Novo Nordisk, designing the strategy, the foundations, the systems meant to unlock this transformation across one of the world’s most complex organisations. I’ve argued publicly and internally that enterprise AI, done well, is one of the most important shifts of our time.
I still believe that. Which is exactly why I can’t stop thinking about this.
Friction is not inefficiency
Anthropologist Anna Lowenhaupt Tsing argued that friction, the awkward, uneven, contested nature of encounter, is not an obstacle to connection. It’s the mechanism of it. Victor Turner, studying ritual and community, found the same thing: the most durable bonds between people form not in comfort, but through shared difficulty. Friction creates relations. Struggle creates meaning.
This is not a design flaw in human experience. The effort we invest in something is inseparable from the meaning we get out of it. That’s why handmade gifts carry more weight than purchased ones. Why a meal cooked for someone says something a restaurant booking cannot. The things we’ve had to fight for stay with us long after the easy things have gone.
Mihaly Csikszentmihalyi spent decades studying what makes people feel most alive. His answer wasn’t comfort or convenience. It was flow, the state where challenge and capability are precisely matched. Too easy and we drift. Too hard and we freeze. Right in the middle, where the work actually resists us, is where we come fully awake.
Richard Sennett made a related argument in The Craftsman: that making things well, slowly, through resistance and failure, is one of the deepest forms of human meaning-making. Not because the object matters, but because the process shapes the person doing the making.
AI removes that friction. And we’re calling it progress.
The depersonalization mechanism
When I write with AI, I’m no longer the author of the thought. I’m the editor of generated options. When I design with AI, I’m no longer the architect of the idea. I’m the person who approves the proposal.
The work still gets done. The output is often impressive. But the cognitive struggle, the resistance that used to be the place where my thinking developed, where I got better, where the work became mine, has been outsourced.
Psychologists use the term depersonalisation for this: the experience of watching yourself from the outside, feeling detached from your own thoughts and actions. In occupational research it shows up as a core dimension of burnout, not the exhaustion, but the disconnection, the sense of going through the motions. What I’m describing isn’t really about being overworked, though. It’s more specific than that.
It’s a direct consequence of frictionlessness. When you take away the struggle, you also take away the mechanism by which the work becomes yours. The frictionless path is faster, yes. It’s also the path that leads quietly away from yourself.
The evolutionary trap
Human cognitive capacity didn’t arrive ready-made. It was built across hundreds of thousands of years of struggle, with the environment, with each other, with ideas. Our ability to reason, create, and find meaning developed through resistance. Difficulty was the forcing function.
We are now building systems that improve their cognitive capabilities in months.
I keep coming back to species that have gone extinct, not through disaster, but through competitive displacement. When two species occupy the same niche, the one that uses it more efficiently wins, not through evolution speed alone, but by occupying the space the other depended on, more and more completely, until there is no room left. It’s a slow, structural process, and it’s irreversible.
What’s different about what we’re doing is this: those species didn’t choose their displacement. They didn’t build the thing that displaced them, and they didn’t hand their niche over voluntarily.
We’re doing all three.
We’re not being replaced by AI. We’re hollowing ourselves out in favour of it. With every frictionless choice, in our work, in our thinking, in our creative lives, we quietly outsource the cognitive struggle that made us what we are across millennia. Evolution gave us the capacity to handle hard problems. We’re now systematically choosing not to use it, consciously, collectively, and at a pace evolution could never correct for.
The species that went extinct had no choice. We do.
So here’s the question I keep sitting with: if we stop struggling with ideas, really struggling, sitting with the uncertainty, letting something resist us long enough to become ours, what exactly are we preserving when we preserve ourselves?
I don’t think the answer is to use AI less. I think the answer is to use it differently.
There’s a version of this technology that removes friction and a version that creates it. The first generates the thought so you can approve it. The second challenges the thought you already had, pushes back on your reasoning, forces you to go deeper than you would have alone. One replaces the cognitive struggle. The other amplifies it.
This article came out of the second kind of use. Not AI writing my thinking, but AI interrogating it, asking the questions that made me uncomfortable, refusing to let me stay on the surface. That process had friction. It was slower than just generating a polished draft. And it felt like mine in a way that content approval never does.
The pitfall isn’t AI. It’s the frictionless path, the temptation to let the tool do the cognitive heavy lifting because it can, and to call that productivity. What we actually risk losing in that bargain isn’t time. It’s the thinking that makes us worth anything in the first place.