When the Ground Moves

There is a silence I have started to hear in one-on-ones. You are talking to someone good, who spent years getting good, and somewhere in it they go quiet. They have just watched a model do in a few seconds what used to take them a week, the thing they were proud of, and the question they are too polite to ask is sitting right there: what is the point of getting better at this? That silence is the most dangerous thing in your company right now, and it isn't the AI. It's skill fatalism. The private belief that the ground you built a career on is coming apart, and that the only thing left to do is wait for your turn. You don't usually hear it directly. It's the senior who has stopped mentoring, because somewhere along the way they stopped being sure their craft was worth passing on. A designer who checks the model before they trust their own eye. The junior who has decided the fundamentals are a waste of time now the machine has them. Left long enough it becomes an exit interview, and it turns out it had been there for months, dressed up as ordinary busyness. The fatalism gets one thing badly wrong, though. The ground is moving, but not evenly. When it shifts, the loose stuff on top goes first, and what stays is whatever was solid underneath. That's what AI is doing to knowledge work. It takes the routine layer first, the boilerplate and the first drafts and the lookups, the busywork that was never really the point. What it can't easily move is the bedrock underneath. Knowing which problem is worth solving in the first place. Being able to tell a good answer from a merely plausible one. The stubbornness to keep pushing for a better question, and the trust that holds a team together when nothing is certain. These aren't really skills, not the way we normally use the word. No course will give them to you. They're closer to traits, the kind of thing that sets slowly, under years of pressure and feedback and getting plenty wrong. Which is why they matter now. You can't fake them for long, and the machine can't hand them to you. And here is the part that keeps me up at night. Traits set underfoot, in the years of ordinary, unglamorous work that nobody flags as important while it's happening. A senior has that layer under them already, so when the ground moves they hold, and they're worth more than ever. A junior has nothing set. The ground starts moving before anything has hardened, and there's no footing anywhere to find. Everyone calls this top-loading and assumes it's a story about hiring senior people. It's worse than that. We are handing juniors the reach of a senior with none of the ground underneath it, none of the decade of standing on your own work that taught the senior what the reach was even for. Let it run and the space between your seniors and juniors stops being a gap you can coach across. It becomes a cliff. The road that used to carry someone from the one to the other, those slow, boring years where the judgment set, is being dug up while they are still walking on it. The mental health of a team goes the same way, on the same timeline. Fatalism and burnout run on each other. Nobody says it directly, but people absorb the idea that they now have to keep pace with something that gets better every single week, and that the week they fall behind is the week they are finished. It's a treadmill with the speed dial hidden. There's no way to tell how fast you are supposed to be going, only that you are behind, and that grinds people down in a way that has nothing to do with hours worked. You can't upskill someone who is scared and running on empty, whatever you have set aside for training. Get that part wrong and none of the rest of it lands. Something else is shifting underneath all this. When the doing gets cheap, the rare and valuable thing becomes deciding what is worth doing at all. That used to belong to the product people and the founders; now it belongs to everyone, whether they were trained for it or not. A person who can only pick up a clean ticket and close it is the most exposed of all. The ones who can take something vague and messy and work out what actually matters in it are the ones you can't easily replace. Most people were never taught that, because nobody thought they would need to be. This is also, oddly, where AI stops being a threat and starts being useful, though only for a certain kind of person. They become a conduit for it rather than its competition. Not because they have memorised the current tool, which will be a different tool by next quarter, but because they bring the judgment the machine hasn't got and know where to point it. None of this is really about technical skill. It's about whether there is something worth saying behind the prompt, or just a faster way to say nothing. So how do you actually lead through this, instead of standing still and hoping the ground settles? Standing still is easy to spot. You wait until the fatalism has set in, then send round a training platform and a note about embracing the future. The alternative is slower and less comfortable. A handful of things have actually made a difference for me: - Hire for the traits, not the toolkit. A toolkit is the loose stuff on top, and most of what impresses you in an interview today will be gone in a year. What you are really hiring is curiosity, and the willingness to be wrong in front of other people. The instinct to ask whether a thing is even worth doing before rushing to do it. None of that shows up on a CV, so you have to go looking for it on purpose. - Put the reps back by hand. The everyday work that used to grow a person's judgment is exactly what AI is taking, so you have to recreate it deliberately. Make juniors argue with the model's output instead of pasting it in, and make them tell you what the actual problem is before they are allowed anywhere near a solution. Put them next to your best people for the thinking, not just the hand-off. - Give people ownership before they feel ready for it. Nobody's judgment ever grew out of being handed safe, well-defined work for a decade. Let them own an outcome, make some real calls, and get a few of them wrong. It's the only way the weight-bearing stuff forms at all. - Take morale as seriously as the roadmap. Say the word "fatalism" to your team, because it loses most of its power the moment someone names it. Give them a reason to be here that doesn't depend on out-running the models, because that is a race none of you win. And learn to spot burnout for what it is when it arrives dressed up as someone being extremely diligent. - Use the thing yourself, in the open, badly at first. Nothing kills fatalism faster than watching the person in charge fumble with AI and then fix it with judgment the model plainly doesn't have. None of this guarantees everyone comes through it intact. Some skills really are going, and it would be a lie to pretend otherwise. A few of your people probably should feel the fatalism. If someone built their whole identity on the part that was always going to erode, reassuring them is not a kindness. The kinder thing, and the harder one, is to tell them straight and early, while there is still time to build something else to stand on. For most people, though, that isn't the situation. Fatalism is a forecast they have started treating as a fact. It looks at the moving ground, decides there is nothing solid left anywhere, and never checks whether it is already standing on rock. The models are going to keep getting better; that was always coming. What actually matters is whether the people around you have found their feet by the time it counts, and whether you have found yours. That's the job now. If I'm honest, it always was.