It started with a problem.
Barclays Bank installed what is widely recognized as the first cash dispensing machine at its Enfield, London branch on June 27, 1967. The machine, developed by John Shepherd-Barron, dispensed a fixed amount of cash when a customer inserted a special paper token matched to a personal identification number. Don Wetzel, working for Docutel in the United States, independently developed a network-capable ATM machine deployed by Chemical Bank in New York in 1969. Multiple inventors in several countries were working on the concept simultaneously. The Computer History Museum's Revolution exhibition documents the competing claims to ATM invention and the rapid spread of the technology through the 1970s.
Historical sources below ↓Then something changed.
The ATM expanded the hours and locations where banking services were available. A branch open five days a week for eight hours per day served a fraction of the hours customers might need cash. A machine available twenty-four hours changed the assumption that cash access required a bank's operating schedule. As ATM networks grew — first within individual banks, then shared networks, then internationally — the machine decoupled cash access from bank geography. Customers could access their funds in cities they had never visited. This geographic and temporal extension of a service, without the cost of staffing every location, was the model that later shaped how digital financial services thought about availability.
What a builder can learn.
Our interpretation: the ATM solved the access problem, not the service problem. It could dispense cash but not answer a question. This is a common pattern in automation: the routine task becomes available everywhere and at all hours; the complex or personalized task remains dependent on a person. For a business builder, automation works best when it extends access to well-defined tasks. Trying to automate tasks that require judgment or conversation typically reduces quality rather than extending access.
A door still open.
A hypothesis worth testing: if a small business identified its most routine customer request — the equivalent of cash withdrawal — and made it available outside business hours through a simple automated channel, would the volume handled by staff decrease enough to free time for the more complex requests?
What is the most routine request your customers make that currently requires a person to handle — and would those customers be well served by an automated option?
Follow the evidence.
History is sourced below. Business lessons are editorial interpretations; future possibilities are questions to test, not predictions.
Prepared from primary-source histories. First edition: 25 Sept 2026.