Where AI Already Works Today — the Spec+DoD Formula

If you want to understand where AI works and where it doesn't, two questions are essential: Can you precisely describe what you want beforehand? And can you verify whether the result is correct afterwards?

AI agents work in loops — try, check, correct, repeat. Like a sat-nav constantly recalculating the route. But without a destination, the sat-nav drives in circles. And without a "destination reached" signal, it keeps driving — even when you're already there.

In software development, good engineers can produce both — a clear specification and automated tests. That's why AI works so well there. But the principle extends beyond software. Data reconciliation, document review, structured analysis — wherever tasks follow clear rules and the outcome is verifiable, AI agents already deliver strong results.

The catch: getting such agents to run reliably is technically demanding. Many underestimate this — whether IT tries to set it up "on the side" or a business unit cobbles something together with ChatGPT. Between an impressive demo and a reliable process lies a considerable effort.

For corporate strategy, creative work, or complex advisory, both prerequisites are still largely missing. There, AI is a useful support today, but not an autopilot. Whether this fundamentally changes with current models is an open question — it would likely require different architectures. All the more reason: those who start with clearly defined tasks today gain the experience that makes the difference later.