AI can build. It won't (and shouldn't) choose what.

A few days ago we were interviewing a candidate and they turned the question back on us: 'How much are you actually using AI, and what do you think about it as a product and tech team?' The room went quiet. Not the awkward kind. The kind where everyone realises they have an opinion they haven't fully tested yet. It made me reflect more than I expected, and left me feeling strangely optimistic — even if that optimism comes with the admission that parts of my job, the job I was doing two or three years ago, are becoming redundant. The role is changing. I think for the better.

My view is optimistic, but not in the lazy way. I don't think AI is going to hand us four-day weeks and universal creativity. I think it is going to make a lot of current tasks irrelevant, and in doing so it will force designers, developers, and product people to become something more like architects.

"The machine can write the code. It cannot read the room."

What this means for developers

There is no polite way to say it: AI is already better than me at generating lines of code. It is faster, it has read more repositories than I ever will, and it can produce a working version of almost anything I can describe. The catch is that 'working' is not the same as 'right for us.'

An AI can see the file structure and the existing components. It can even be told about the current architecture. What it cannot reliably hold is the future. It does not know the feature that is being discussed in the next roadmap meeting. It does not know the business constraint that hasn't been written down yet. It does not know the performance budget that exists because of a decision made two years ago by someone who has since left.

So it ships shortcuts. It imports a library the team already decided to avoid. It adds state in a place that will make the next refactor painful. It solves the problem in front of it, not the problem three problems ahead. That is not failure. That is just the limit of context.

What this means for designers

The same thing, dressed differently. AI can produce a hundred layouts in the time it takes me to make one frame. It can match a style, extend a colour system, generate placeholder copy that is almost good enough. What it cannot do is know what should not exist.

Design is already half editing. The next version will be more editing and less drawing. The value shifts from 'can you make it look good?' to 'do you know what good means here?' — which is a completely different job, and a much harder one to fake.

Where the machine wins

Where the human still matters

The thinking behind the work. Sometimes that is research, sometimes it is taste, sometimes it is just the weird brilliance of a human mind making a connection that no training set would predict. Product work is full of decisions that look irrational on paper and turn out to be right because someone understood the people involved.

What to ship. How to ship it. What not to ship, even though it is possible. Why one feature creates clarity and another creates noise. These are not optimisation problems. They are judgment problems, and judgment needs context that lives outside the codebase or the design file.

This is not the first time

When I started out, I called myself a graphic designer and a game designer. Those jobs did not really exist thirty years before. The personal computer did not exist. The internet, as we know it, did not exist. My entire career has been made possible by technology that made older roles redundant and invented new ones.

AI is doing the same thing, just faster and louder. It will delete tasks first, then roles, then whole categories of work that we currently think of as specialist. The people who survive are not the ones who resist it. They are the ones who move up the ladder of judgment before the ladder of execution collapses underneath them.

The honest question about juniors

Here is the part that worries me. If AI does so much of the execution, how does a junior developer or designer build the intuition that makes a senior person valuable? A lot of craft knowledge comes from doing the work the slow way, failing, and remembering why. If the machine skips the slow way, where does the intuition come from?

I don't know the answer. But every generation looks at the one before it and thinks they had it easier, then invents a better way to solve the problem than anyone older could have imagined. They will be smarter, younger, and full of energy. They will probably laugh at how much we worried.

"The job is not going away. It is becoming the part of the job we used to pretend was optional."

For now, the practical move is the same for designers and developers: let AI handle the parts that do not require your judgment, and spend your time on the parts that do. Understand the architecture. Understand the user. Understand the business enough to know why a technically correct solution might still be the wrong one.

The future is not human versus machine. It is people who can think with the machine versus people who only know how to ask it for things. I would rather be in the first group.