In 1967, Barclays Bank installed the world’s first automated teller machine (ATM) in London. The prediction was immediate: machines would eliminate jobs in bank branches. More than half a century later, history had a sense of irony. Banking didn’t shrink. It evolved into a different kind of banking.
Now the same question is back, this time with artificial intelligence. Will AI replace workers in banking and insurance? Perhaps the best answer isn’t found in the future, but in a much older scene: a machine dispensing cash in a British bank branch while an entire industry believed it was witnessing the end of human work. It wasn’t. It was witnessing its transformation.
The “Computers” Nobody Remembers
Before electronic computers existed, many companies—including banks and insurance firms—employed people with that exact job title: computers. They were mostly women who spent entire workdays performing mathematical calculations by hand or using mechanical adding machines known as comptometers.
When electronic computers arrived in the 1950s and 1960s, those jobs disappeared. They weren’t rebranded. They weren’t repackaged. They disappeared. But something even more important happened: demand for financial services grew. Automation lowered costs, expanded capabilities, and enabled the industry to do more, for more people, at a scale that had previously been unimaginable.
The ATM Paradox
ATMs reduced the operating cost of bank branches. Because fewer employees were needed for transactional tasks, banks were able to open more branches. And more branches ultimately meant more employees overall.
Economist James Bessen studied this phenomenon and showed that between 1970 and 2010, while the number of ATMs in the United States grew from zero to hundreds of thousands, the number of human bank tellers also increased.
The machine didn’t eliminate the human worker. It removed part of the job. And in doing so, it forced the role to be redefined. Bank tellers stopped being primarily cash dispensers and became relationship managers, product advisors, and interpreters of customer needs.
Why AI Resembles the ATM—Even Though It Isn’t the Same
Large language models don’t just automate repetitive tasks. They can draft insurance policies, analyze claims, answer complex inquiries, generate risk reports, and hold sophisticated conversations with customers. This is a much bigger leap: we’re no longer talking only about replacing manual work, but about automating cognitive tasks.
Yet the question remains the same one history has taught us to ask more thoughtfully. It’s not enough to ask how many jobs a technology will eliminate. We also need to ask what new demand it will create.
What Really Changes: The Speed
The ATM took decades to become widespread. Generative AI is being adopted in a matter of months. That acceleration changes the conversation.
Teams that currently process routine claims or draft repetitive communications don’t have a decade to adapt. They need new skills now. Not necessarily to build AI models, but to work alongside them: supervising outputs, identifying mistakes, interpreting exceptions, and providing context.
This is where institutions that simply adopt technology diverge from those that truly transform. The former buy tools. The latter redesign capabilities. AI doesn’t replace culture. It exposes it.
As Julián Colombo, Founder and CEO of N5, often says, “The future of work is not defined by the technology that emerges, but by organizations’ ability to redesign what people do with it.”
That statement captures the central idea: AI should not be viewed merely as an efficiency tool, but as an opportunity to move teams toward work that requires greater judgment, greater responsibility, and greater impact.
The Lesson History Keeps Repeating
In 1589, Reverend William Lee invented a knitting machine. Queen Elizabeth I refused to grant him a patent because she feared it would destroy the jobs of English knitters. Four centuries later, the textile industry employs millions of people around the world.
History does not tell us that automation is painless. It isn’t. And it almost always hurts those who have the least power to influence the pace of change.
For banks and insurance companies, AI does not represent the end of human work. It represents the end of a particular kind of human work: the most routine, the most predictable, and the most repetitive.
What remains—and will likely continue to grow—is a different kind of work: work that interprets context, manages trust, and assumes responsibility.
The ATM taught us that a machine can transform an industry without necessarily eliminating the people who work in it. Artificial intelligence is teaching us the same lesson—only much faster.
And this time, the question is not whether technology will transform work. The question is who will be ready when it does.
