Will AI Replace Your People? A Cash Machine Already Answered That

Part 1 of a 5-part series on AI myths and misconceptions.

In 1969, one of the first cash machines in the country went into a bank on Long Island. It did something that was supposed to end an entire job. It handed a customer cash with no teller involved.

The prediction wrote itself. Machines that dispense money would spread, and bank tellers would disappear.

They didn't.

Here is what actually happened. As cash machines rolled out, the number of tellers a single branch needed did drop, from about 20 down to about 13 (Bessen, IMF, 2015). But a branch that needs fewer people is cheaper to run. So banks opened more branches. Total teller jobs did not fall through those years. They grew. And the work changed. Tellers stopped spending the day counting cash, because the machine did that now, and started spending it on the things the machine couldn't touch. Solving a problem. Answering the question that didn't fit the form. Helping a nervous customer make a decision.

Keep that story close, because you are being sold the same prediction about AI that people were sold about the cash machine. The machine does the work, so you won't need the people. It sounded obvious in 1969. It was wrong in 1969.

What the machine actually takes

A cash machine is very good at exactly one thing. It hands out money, the same way, all day, without getting tired or annoyed. It is useless at everything else. It can't read a worried face. It can't catch the transaction that looks off. It can't decide anything.

AI in your business works the same way. It is good at routine, repetitive volume. The same five questions asked fifty times a day. The call that comes in after you've closed. The follow-up that would have slipped through by Friday because everyone was busy. It is not good at judgment, or patience, or the messy situation that doesn't fit the script. That part still needs a person. It always did.

So the honest question was never “machine or human.” It was “which parts is each one actually good at.”

What that means for your team

Picture your front desk on a normal Tuesday. Someone's on hold. Two calls already went to voicemail. The same question about your hours and your pricing has come in six times before lunch. Your person is busy. They are just busy doing things a machine could have answered in a second.

Now take that routine pile off their plate. Every simple question answered the moment it's asked, day or night. Every missed call caught and followed up instead of lost. What's left for your person is the customer whose situation is genuinely complicated, the one who needs a real answer from someone who now has the time to give it.

That is not a smaller team. It is the same team, freed up to do the part of the job that needed a human in the first place. Better work, not less of it.

The pitch that gets this backward

Here is the part worth being straight about. The replacement version is easier to sell. Salary against subscription. One number against another. Cut the cost. The math even looks clean on a page.

It just answers the wrong question. The question is not what a role costs you. It is what that person does that a machine can't, and whether you're giving them room to do more of it or less. Every time someone frames AI as a way to remove a person, ask where that person's judgment is supposed to go. If the answer is “the machine handles it,” the bank teller story is fifty years of evidence that this is the part the pitch gets wrong.

The version everyone hates

You already know the bad way to do this, because you've been stuck in it. You call about a charge on your account. You get the robot. It doesn't understand you. It loops you back to the same three options. You press zero. You say “representative” four times through your teeth. That is AI used as a wall, dropped between you and a human on purpose, because a wall is cheaper than a person.

That is the real fear hiding under “AI is taking over.” It isn't only about jobs. It's that the moment you actually need help, there's nobody left to reach.

Done right, it works the other way. AI clears the simple stuff fast, the balance, the due date, the password reset, so that when something is genuinely wrong, a real person is free to pick up, because they aren't buried under the same fifty easy questions. The machine's job is to move the routine out of the way and get the hard problem to a human faster, not to keep the human away from you.

Your customers already grade you on this, whether you notice it or not. When something goes wrong, can they reach a person quickly? If yes, you built a bridge. If no, you built a wall. And a wall quietly costs you the customers you most wanted to keep.

Frequently Asked Questions

Doesn't this just mean fewer jobs eventually, even if not right away?

The historical record says otherwise, and not just for bank tellers. Every time technology took over a repetitive task, total employment in that field grew, not shrank, because the work left over expanded to fill the space. That's not a guarantee for every job everywhere, but it's the opposite of what the “AI is coming for your job” headlines assume.

What if my team is worried AI is coming for their position?

Show them exactly what's changing, not just tell them. If the plan is genuinely to take routine volume off their plate so they have room for harder, more interesting work, walk them through which specific tasks move and which don't. Vague reassurance breeds more worry than a real answer does.

How do I actually tell if a system is the augmentation kind or the replacement kind before I buy it?

Ask what happens to the person's job the day after it's installed. If the honest answer is “they have less to do and nothing to fill it,” that's replacement, whatever it's marketed as. If the answer is “they have less busywork and more room for the calls that actually need them,” that's augmentation.

A simple next step

You don't need to change anything today to see this for yourself.

Pick one task. The one that eats your team's time and needs the least judgment. The repeated question, the after-hours call, the follow-up nobody gets to. Just that one.

If it would help, I'll look at that single task with you and show you plainly what handing it to a system would free your people up to do instead. No contract, no obligation, and you walk away with a clear picture either way, whether you ever work with me or not.

Continue to the next piece in this series: The Cheap Way to Automate Has a Hidden Bill.

Sources & Further Reading

Toil and Technology, James Bessen, IMF Finance & Development, March 2015