AI makes roles wider, not obsolete

The debate about AI and jobs has two camps. One says: the jobs are going away. The other says: AI can't do that, just look at the nonsense ChatGPT writes. I think both are wrong, and for the same reason. Most people only know AI as a chatbot. They never see what it can actually do today. So they miss what is really changing: roles aren't disappearing, they're getting wider.

Why do the two camps talk past each other?

If you're reading this, you're probably fairly deep into code and AI. But for most people, AI is a chat window. You ask a question, get an answer, paste it somewhere. Few have heard of Claude Code, MCP means nothing to them, and the command line is a foreign word.

That is a huge difference. A chatbot only sees what you paste into it. An AI agent with access to files, tools and the command line can read a whole project, run commands, look at the result and keep fixing it until it fits. If you only know the chat window, you judge AI by its weakest form.

There's also something human going on. Many people feel their pride is hurt. They spent years learning their craft and don't want to hear that a machine does part of it in minutes. I understand that. But resistance doesn't change what the technology can do.

Is "AI can't do that" often a usage problem?

Take web design. Tell an AI "design me a website" and you usually get something generic. Stock layout, default colours, nothing a client would pay for. People who try it that way rightly conclude that AI is bad at design.

But that isn't what AI can do. That's what AI can do without guidance.

Give it a clear brief, the right skills (written-down knowledge of what good design looks like) and an iteration loop in which it renders its own result, looks at it, critiques it and improves it, and the picture changes completely. In very little time, it can in many cases produce a better website than many web designers would. That's not a promise about the future. That's where things are today.

So the difference isn't the AI, it's how you use it. And most people don't see that, because they never got past the chat window.

What does this mean for the next three years?

If that's where things are today, don't assume they'll stay there. What AI could do three years ago looks almost silly now. In three years, what impresses us today will look the same way.

That's why I think it's risky to rest on "AI can't do that". Even when it's true for a given task today, it's a snapshot. You have to face reality: a lot of what is manual work today, AI will take over.

That doesn't mean the people disappear. It means what they're needed for shifts.

Will junior and senior roles stay?

Yes, I think so. The distinction stays because it was never only about tasks. Senior means more experience, more depth, a better sense of what will go wrong in a year. With AI you need that more, not less, because someone has to judge whether what the AI delivers actually holds up.

What changes is the shape of the junior role. A junior used to take on a small, clearly bounded piece of work. Those bounded pieces are exactly what AI keeps getting better at. So the junior role has to widen.

What does a wider role look like?

Two examples.

Web design becomes brand work. Instead of only building a website, you take on the whole corporate identity: logo, colours, typography, tone of voice, social media templates, presentations. That used to be a project for several people. With AI, one person can set the frame and let the AI do the detailed work, provided that person knows what a coherent brand looks like.

Data analyst becomes end-to-end ownership. Instead of only analysing data, you look after the whole pipeline: where the data comes from, how it's transformed and processed, how it lands in the dashboard and which alerts fire when something is off. That used to be data engineering, analysis and reporting, often split across different people or teams. With AI, one person can keep the whole path in view.

That's the crux: you need a broad range. You don't have to go as deep as a specialist everywhere, because AI helps out in many places. But you need to understand enough to know what to ask and to spot when the answer is wrong. The future means covering a much larger area than today.

What follows from this?

For me, three things.

1. Go beyond the chat window. If you want to judge AI, use it the way it can be used today: with agents, tools and iteration loops. Only then do you see what it really does.
2. Extend your own area to the left and right. If you build websites, understand brands. If you analyse data, know how it gets into the system and how you'd notice it's wrong.
3. Keep depth where it counts. Breadth only works with a solid core. Without it you can't steer the AI, only hope.

Jobs aren't simply vanishing, and it isn't true that AI can do nothing. The truth is in between: roles are getting wider, and those who see that early will be more efficient than everyone still arguing about it.

Why AI takes over junior tasks but not junior people is in AI replaces junior tasks, not junior people. And why AI is strongest wherever the output is code, in AI speaks code, not clicks.