Most AI projects are paving the cow path
Last week I spoke with someone who is building one of the large AI programs here in Abu Dhabi. He told me about all the use cases that have been discovered already, and that now someone needs to bring them to life.
An impressive list, honestly. But something about it has been bothering me since, and it took me a while to frame what exactly.
It’s this: everyone is asking how to use AI. Almost nobody is asking whether the problem underneath is real.
What happens is that a team takes an existing process and asks, how do we add AI to this? They automate it, it gets faster, everyone claps. I’ve done this myself. It feels like progress. But nobody in the room has questioned the process itself.
The cow path
There’s an old phrase for this that I only learned recently: paving the cow path. Cows wander and wear a crooked trail into the ground. Then people come along and, instead of questioning the route, pave over it. Now you have a smooth road that still goes the wrong way.
I think most AI projects right now are exactly this.
The hard part was never the technology. The hard part is asking: should this step even exist? What are we actually trying to achieve here? And if this problem landed on us today, with AI already in hand, how would we design the whole thing?
AI scales whatever thinking you already had. If the thinking was lazy, congratulations — you automated the laziness.
Three things I keep reminding myself
- Before asking where AI fits into a process, ask whether the process should exist at all.
- Automating an unquestioned process gives you the wrong thing, faster.
- Start from the outcome and design backwards, with AI in hand from day one.
This is the same fracture I keep running into when adoption stalls: the system works, but it never becomes how we work. A paved cow path is a well-built answer to a question nobody checked.
So the question worth asking isn’t where AI fits. It’s which process everyone has been optimising and nobody has thought to question.