
Did the automated teller machine already prove why premature redundancies fail?
Whenever technological automation slashes transactional overheads, executive suites invariably rush to prune headcounts. History suggests that making routine tasks cheaper invariably expands the market for human judgement.
When Barclays introduced the first automated cash dispenser in Enfield in 1967, followed by the rapid spread of networked automated teller machines in the 1970s and 1980s, the eventual demise of the bank branch seemed mathematically assured. If a mechanical safe could count banknotes reliably at midnight, why pay counter clerks to stand behind timber counters?
The elasticity of human advice
What actually unfolded remains one of modern industrial economics' most instructive lessons. Because automated machines lowered the operational cost of running an individual branch, banks did not close down high-street presences en masse; they opened far more of them. Freed from the physical grind of handling paper notes, branch employees were retrained as relationship managers, mortgage advisers, and commercial underwriters. The volume of customer contact multiplied.
That cycle is repeating itself almost verbatim. A striking Gartner prediction warns that by 2029, roughly 30 per cent of employees dismissed in haste under the guise of AI replacement will need to be rehired. As their analysts note, the early error was assuming work automation was the objective, when workforce amplification was the true commercial opening.
The compounding burden of oversight
Automating execution does not eliminate responsibility; it multiplies the need for supervision. In a pioneering academic paper on human oversight, researchers found that developers using autonomous agents do not simply kick back; they take on four intense, preventative oversight duties: upfront constraint setting, co-planning, real-time monitoring, and retrospective auditing.
“When business and IT executives look back on the early AI-era, they will realize their greatest mistake was believing that work automation was the point.”
Just as the teller machine transformed banknote counters into financial consultants, agentic systems are converting procedural knowledge workers into active system overseers. The work never vanished—it simply demanded better judgement.
When your organisation calculates the savings of automation, is anyone pricing the cost of the institutional wisdom you might throw away?