If AI replaces developers, who will decide what it builds?

I regularly hear and read that “AI will replace everyone”. And, in a slightly more targeted version, that “AI will replace developers”.

On the first claim, I do not know where to start. Replacing plumbers, nurses, teachers, firefighters, mechanics, craftspeople: at that point it is no longer foresight, it is hallucination.

On developers, I find it almost funny. Because if AI replaces developers, who will decide what it should build? The customer? In that case every salesperson will build their own Salesforce, every HR manager their own HRIS, every finance officer their own ERP, and every department will reinvent, alone, the tools it depends on.

Above all, we forget the essential: a developer’s job is not typing code. Code is only one part of the work. Understanding a vague need. Identifying the constraints. Making architecture choices. Weighing the risks. Choosing between several solutions. Understanding the consequences of a technical decision. Questioning a request that makes no sense. Owning a result.

An AI can propose options. It can generate code. It can speed up some tasks enormously. But proposing choices is not taking a decision, and a probabilistic model, however powerful, does not magically become responsible for the relevance of what it produces.

So yes, AI is profoundly transforming the developer’s job, and some tasks are already fully automated.

But replacing the developer because code can now be generated is confusing the profession with its tool

Somewhere between delusion and humour, I am starting to wonder whether the people most at risk of being replaced are not those who spend their time talking nonsense about what AI will replace, when they are sometimes among the people who influence the decisions.

Because the real question is not “can AI write the code?”. It can, that much is certain.

The real question is: “will you be able to tell when the code it generated is wrong?”