Will AI Replace Your Employees? Ask the Bank Tellers.

There are about as many bank tellers in America today as there were in 1970.

I had to look that stat up three times because it broke my brain. How is that possible? We have ATMs everywhere. Mobile deposits. Zelle. Venmo. The entire financial system runs on software now.

And yet: same number of tellers.

This feels important for anyone freaking out about AI and jobs, but I'm not sure anyone's actually learning the right lesson from it.

The thing about ATMs everyone forgets

When Bank of America started rolling out ATMs in the 80s, the prediction was pretty straightforward. Machines count money faster than humans. Machines don't take lunch breaks or call in sick. Machines don't need health insurance. Therefore: bye bye tellers.

Except that's not what happened, and the reason why is kind of annoying if you were hoping for a clean narrative.

Banks didn't keep tellers because they're nice or because of unions or because of some HR policy about "people first." They kept tellers because it turned out that opening a checking account or explaining why someone's mortgage application got denied or talking someone off a ledge when their card got skimmed requires, you know, a human.

But here's the weird part. The ATM didn't save teller jobs. It changed what teller jobs were. And a bunch of tellers couldn't or didn't want to make that shift, so they left anyway. The ones who stayed had to learn a completely different job that just happened to have the same title.

I don't think we talk about that part enough.

Why I'm thinking about this now

I was in a meeting two months ago with a VP of Operations at a company that makes industrial equipment. Third hour of the day, everyone's a little fried, and he just blurts out: "If we're being honest, AI is going to replace like 40% of what my team does, and I have no idea what I'm supposed to tell them."

At least he said it out loud. Most executives are thinking it and dancing around it with words like "augmentation" and "enhancement."

His team does a lot of supplier evaluation. Reading technical specs, comparing vendors, flagging issues in contracts. Tedious, detail-oriented work that LLMs are genuinely pretty good at now. So yeah, he's right. AI can do a lot of that.

But then I asked him what his team complains about most, and he didn't even hesitate: "They hate supplier evaluation. They want to be in the field working with engineers on new product designs, but they're stuck at their desks reading PDFs."

Oh.

So the work AI can replace is the work nobody wants to do anyway. And the work people actually want to do is the work AI can't touch. That should be a huge opportunity. But this guy's stuck because he hasn't figured out what his team becomes when the boring part goes away.

And honestly? That's on him. You can't just automate half of someone's job and hope they figure out the rest. That's how you get mass resignation.

The plan that worked (kind of)

Different client, different industry. Manufacturing. They brought in computer vision AI for quality inspection. Cameras scanning parts, flagging defects, way faster and more consistent than humans.

The QA team lost their minds. Not in a good way.

Plant manager tried the usual reassurance talk. "Your jobs are safe, you'll just be working alongside the AI." Which is executive-speak for "I haven't thought this through yet, please don't quit."

Predictably, two of their best inspectors started interviewing elsewhere within a week.

So they stopped bullshitting and actually mapped out what the new job would look like. AI scans parts. Flags anomalies. QA team investigates why those anomalies happened. They work with engineering to fix upstream processes. They train the AI on new edge cases. Basically they become process improvement specialists who happen to use AI as a tool.

Here's the part that surprised me: the inspectors who stayed weren't the ones who were "good with technology." They were the ones who'd been complaining for years that they never had time to dig into root causes because they were too busy staring at parts all day.

The AI gave them back the interesting part of their job. But it took the plant manager three months to figure that out and another two months to retrain everyone. And even then, one guy left because he actually liked the old job better. Preferred the simplicity of pass/fail visual inspection. Didn't want to deal with engineers or process optimization.

Which is fine? Not everyone wants to upskill. That's a thing we should be allowed to say.

What your employees are actually worried about

I've been in maybe a dozen "AI townhalls" at this point, and the vibe is always the same. Leadership talks about exciting new capabilities and increased productivity. Employees sit there doing mental math about job security.

But here's what I've noticed: when you talk to people one-on-one after these meetings, almost nobody says "I'm worried AI will do my job better than me."

What they say is: "I'm worried I won't learn this fast enough and I'll get left behind."

Status anxiety. Relative position. The fear isn't that AI replaces you. It's that AI makes you irrelevant while your coworkers become AI power users and you become deadweight.

That's a very different problem. And it means the solution isn't "don't worry, your job is safe." The solution is "here's exactly how we're going to make sure you don't get left behind, with dates and milestones and specific training."

Most companies aren't doing that. They're announcing AI pilots and crossing their fingers that everyone figures it out.

The thing nobody wants to admit

Look, if your strategy is "implement AI to reduce headcount," just say that. Be honest about it. It sucks, but at least people can make decisions with real information.

What's worse is this vague "transformation" talk where leadership hasn't actually decided whether they're keeping people or cutting people. So employees are stuck in limbo, best performers start job hunting just in case, and by the time you figure out your strategy, your A-players are gone.

I see this constantly. CEO announces big AI investment. Six months later they're confused why their best people left for competitors. Well, you spent six months talking about efficiency gains and automation. What did you think they'd do, wait around to see if they made the cut?

The ATM thing is instructive here too. Some banks absolutely did lay off tellers after installing ATMs. Those banks saved money in year one. Then they discovered that branches without human staff had terrible customer satisfaction scores and started losing deposits to competitors. Turns out people don't love banking at a kiosk.

The banks that won weren't the ones that kept everyone out of kindness. They were the ones who were smart enough to realize that customer relationships were worth more than transaction efficiency. So they kept people and changed what those people did.

Pure business decision. But it required actually thinking about what branches would become, not just what ATMs could do.

What I think is actually happening

Here's my read: executives are so focused on what AI can automate that they're missing what AI can unlock.

When you automate grunt work, you don't get headcount savings (well, you can, but that's the boring play). You get capacity. Your team suddenly has 10 hours a week they didn't have before.

Question is: what are you going to do with it?

Because if the answer is "nothing specific," those hours evaporate into Slack and status meetings and general corporate entropy. Or your people use those hours to work on their resumes.

But if you actually assign that capacity to something that matters (new product development, customer experience projects, strategic analysis, whatever), then you're not just implementing AI. You're upgrading your entire operation.

The companies that figure this out are going to run over the ones that don't. Not because they have better AI. Because they know what to do with the time AI gives them back.

My actual advice (if you want it)

Stop doing AI announcements that are all about the technology. Start doing AI announcements that are about what people's jobs become.

Not "we're implementing AI-powered analytics." Instead: "Starting Q2, the system generates the weekly variance reports you currently spend 6 hours building. Your job becomes analyzing those variances and presenting strategic recommendations to department heads. We're bringing in a training partner to teach executive communication and data storytelling. First session is March 15."

See the difference? One is a threat. The other is a roadmap.

Also, be real about the fact that some people won't want to make the transition. The bank tellers who just wanted to count money and go home? They left. That's okay. Not everyone wants their job to change. But if you're upfront about what's changing and you invest in helping people get there, most will surprise you.

And for the love of god, track where the capacity is going. If AI saves your team 1000 hours a month, where are those hours being redeployed? If you don't know, you're wasting them.

Why this matters right now

The next two years are going to determine which companies come out of this ahead and which ones get smoked. Everyone has access to the same AI tools. The differentiation isn't the technology. It's whether you know what to do with it.

The bank teller thing isn't a feel-good story about how automation doesn't kill jobs. It's a case study in how the companies that win are the ones who think about transformation, not just efficiency.

Your employees are waiting to see which one you are. What are you going to show them?