Please Stop Putting AI in Everything (From a Pro-AI Operator)

From someone who builds AI into operations for a living.

Let me start where it counts: I love AI. I build it into how companies operate, I've staked part of my career on it, and I think the teams that use it well will pull away from the teams that don't. So take what follows as a believer's complaint, not a skeptic's.

Please stop putting AI in everything.

There are two very different jobs people ask AI to do, and we've stopped distinguishing between them. The first job is the good one: taking the judgment-free toil off someone's plate — the reporting, the status-chasing, the copy-paste coordination — so they can finally spend their time on the work that actually needed a human. That's AI in service of the work. I'm all the way in on that.

The second job is the one quietly eating everyone's goodwill: bolting AI onto a task that already works. And it's everywhere right now, because “we added AI” has become a thing teams say to prove they're modern, whether or not the task ever asked for it.

Here's the shape of it. Something took thirty seconds. It worked. People had it down cold. Then AI gets layered on top, and now the same thing takes twenty-nine seconds — except you also inherit latency, a new default nobody requested, and the mental tax of relearning a process that was already muscle memory. Net-net, you didn't save anyone time. You spent the team's attention to shave a second off something that was never the bottleneck.

That's not efficiency. That's motion dressed up as progress.

I've watched teams build genuinely impressive AI that was simply redundant to the task it was meant to improve — so instead of removing work, it added a layer. The tool wasn't bad. The judgment about where to point it was.

This is the part leaders miss: forced process doesn't create efficiency. It creates workarounds. The moment you make a tool mandatory for work that didn't need it, people don't get more efficient — they get creative about avoiding it, and then you've spent political capital and gotten resistance in return. Mandating adoption and calling the pushback “change management” is one of the most expensive mistakes I see.

Because AI adoption, at the individual level, is a personal choice — and it should be. How you organize your own desk, your own flow, what earns a place in your day: that's yours. The best rollouts I've been part of treated AI like any other tool. You make it available, you make the case, and you let it earn its place by being obviously better. When it's genuinely better, adoption isn't a fight. People aren't resistant to things that make their lives easier. They're resistant to things imposed on a system that already worked.

So here's my actual operating rule, and I'd offer it to any leader mid-AI-rollout:

Point AI at what's broken. Leave what works alone.

If a process is slow, error-prone, or soul-crushing, that's your target — automate it, augment it, hand the toil to the machine. But if something takes thirty seconds, runs clean, and the person doing it likes their setup? You don't need AI to fix that. You need the discipline to leave it alone.

The goal was never “put AI in everything.” The goal was always to make the work better. Sometimes the most operationally sophisticated thing you can do is decide that a tool, however impressive, isn't needed here — and move on.