Why AI Forces Us to Finally Speak Plainly

You see it everywhere right now: managers looking forward with excitement to a world where AI writes the code, plans the systems, and deploys everything at the press of a button. The vision is maximum speed with minimal manual effort.

But a crucial point is being overlooked — one that will determine whether projects succeed or end in a technological disaster:

Logical understanding of the problem domain.

The real problem was never the typing

The actual challenge in software development has rarely been writing the code. The problem has always been: what exactly do we want to build?

Experienced developers can compensate for vague instructions ("we need something flexible for customer retention") through experience and logical thinking. They quietly fill the logic gaps left by management and specialist departments. They sense where things are heading. They structure development so that the initial work creates a foundation that can be extended later — rather than thrown away and rebuilt from scratch.

AI does not do this. It is a perfect instruction-follower. If the input is illogical, contradictory or non-committal, the output will be exactly that: garbage in, garbage out — but gilded in super fancy code.

Fred Brooks saw this coming in 1987

In his famous essay No Silver Bullet — Essence and Accidents of Software Engineering, Fred Brooks described how the real difficulty of software development lies in its essential difficulties — understanding and structuring the problem itself. AI only solves the accidental difficulties — the production of the code.

When you tell an AI model what you want and it generates everything you need, that act of "just telling it" is precisely the process known as requirements engineering. Bertrand Meyer describes this entertainingly in AI for software engineering: from probable to provable.

The end of deliberate vagueness

Staying non-committal has long been used as a survival strategy. Keep things vague and you can't be held responsible when things go wrong — but you can still take credit when they go right.

Aside from the fact that this removes any chance of doing better next time, AI punishes every vagueness immediately. When woolly formulations replace clear decisions, AI automation becomes an accelerant for bad development.

My conclusion

The most important skill in the age of AI is probably not programming itself. It is the ability to think logically, understand context, and articulate clearly. Anyone who cannot say precisely what they want — and equally, what they do not want — will simply fail faster with AI than they did before.