Everyone tells the ATM story wrong.
Everyone tells the ATM story wrong.
They treat labor substitution like a task list: automate the cash withdrawal, the teller disappears.
But labor substitution is comparative advantage, not automation.
ATMs did remove a set of transactions from the counter. What they mostly changed was the unit economics of opening branches. Cheaper branches meant more branches. More branches meant more customer touchpoints. Tellers shifted from counting cash to selling products, solving exceptions, and handling trust-heavy moments.
The real break came later.
Smartphones did not just automate a task. They moved the entire interface of the bank into the customer's pocket. That is a paradigm shift: the default channel changed, and so did the customer's tolerance for friction. The demand for "someone at a counter" collapsed because the habit loop moved.
This is the part operators miss when they plan "AI automation" programs.
If your plan is a queue of tasks to remove, you will get local efficiency and global confusion. You will also keep the same org chart incentives, the same handoffs, the same throughput constraints, just with fewer people absorbing the slack.
The question is not "What can we automate?"
It is "Where is the interface moving, and what roles lose their comparative advantage when it does?"