AI Can Write Your Code. Does It Understand Your Business?

Getting software written is becoming easier. Getting the right software built is still hard.

David Heinemeier Hansson (DHH), creator of Ruby on Rails and Omarchy, and co-owner of 37signals, the company behind Basecamp, recently argued that AI agents will weaken the position of programmers whose scarce technical knowledge has excused being difficult to work with. There’s a broader point here: clear communication is becoming more important, not less.

A client shouldn’t need to write a perfect specification. They should be able to explain their problem to someone who asks sensible questions, spots what’s missing and challenges assumptions before they become expensive mistakes.

Agents need that clarity too. They can produce convincing work from an incomplete brief. But convincing isn’t the same as correct, and a working demonstration isn’t proof of security or safety.

That’s where Datasouth comes in.

We communicate clearly with clients and agents, backed by the technical knowledge to question the implementation and verify the outcome. Understanding the business requirement and examining the code are parts of the same job.

Does it solve the actual problem? Does it handle failures properly? Who can access the data? What happens when someone uses it in a way nobody expected?

Those questions don’t disappear because an agent wrote the software. Neither does our responsibility for the answers.

AI can make development faster. It doesn’t make “that looks about right” an acceptable standard.

The code may be easier to produce. Knowing what to ask for — and whether it’s good enough to trust — is where the value lies.