UBER's AI Ice bucket problem

In a recent Yahoo! Finance article, Uber reportedly ranked its engineers on leaderboards according to how much they used AI tools. Leadership declared that the results proved “adoption.” In other words, they declared “mission accomplished” on their AI adoption program.

The CTO, on the other hand, said “back to the drawing board.” That’s because there we were in April, and Uber had already blown through its AI budget for the year, “suggesting AI may be as much a cost driver as a productivity lever,” as this article writes.

I’ve seen this movie before.

In fact, a buddy I started my first company with actually lived it. It was New Year’s Eve. A group of guys in their early 20s. Beach town, off-season, my buddy and his friends had one plan: fill the bathtub with beer and ice.

The ice machine on their floor was broken, so my friend called the front desk. They said come on down.

My friend showed up at the front desk with two buckets and serious intent. The man behind the desk was wearing a bathrobe. He had one ice tray. He looked at my friend, looked at the buckets, and with great ceremony transferred approximately nine cubes into my buddy’s hands.

That was the motel’s Ice Program.

Nobody upstream had asked whether the motel could actually support a holiday crowd. They just opened the doors, told people there was ice, and measured success by how many guests showed up at the desk.

Uber’s AI story is the same type of failure at a much larger scale. The directive came from the top: Use the tools. Then use them more. Here’s a leaderboard. Nobody stress-tested what “everyone goes all-in” actually costs until the budget was already gone. Uber is the motel management. They didn’t count the ice cubes before the holiday weekend.

The lesson here is this: if your modus operandi is “mandate AI enthusiasm and track AI usage,” you’ll still end up holding two empty ice buckets. A firm’s AI usage should start with a strategy. Then you’ll know how much ice you need.

The question worth asking your team isn’t “is everybody using AI?” It’s “how much ice are we going to need if our plan works?”