AI-Drunk Under The Streetlamp

Let's say you want to use AI to solve a problem. The problem choice is negotiable, but it has to be solved with AI, because your boss, board of directors, shareholders or inner voice say so. Being smart about it, you know that just because you have a hammer, not everything is a nail. In other words, you know you have to look for problems where using AI makes sense.

But that can easily lead to a different fallacy, exemplified in the story of the drunk who looks for his car keys under a street lamp, not because that's where he lost them, but because that's where the light is good.

In that sense, if you've decided you'll use AI no matter what, you'll be drawn to problems where AI is an easy and obvious fit, not stopping to consider whether the problem is worth solving at all, which brings us all the way back to identifying the true bottlenecks.

We see that in today's landscape of agentic use cases: Hook up this service to that so when an email comes in, it automagically updates this other app and sends a message to this other one.

All neat and cool but does it let you produce more value, faster, at higher quality? Does it move the needle? If not, it's a distraction. If it's not helping with the one true bottleneck in the value-creation chain, it's not helping at all.