Computers turned everyone into an accidental secretary, and artificial intelligence is set to turn everyone into an accidental manager. Rowland Manthorpe, a Sky News technology correspondent, makes that case in an essay about office work.
Before the personal computer, an office ran through a person. A manager dictated a memo. A secretary typed it, made copies, filed one and carried the others to whoever needed to act. Every request in the building passed through the secretary’s desk and stalled there too.
In 1984, about 18 million people in the United States worked as secretaries or clerks, according to figures Manthorpe gathered. That was close to one in five jobs in the country. Those jobs did not vanish when personal computers reached the desks. Demand for admin work rose for years afterward, because when a task gets easier, people take on more of it. Manthorpe calls this a Jevons paradox for bureaucracy.
The bigger disappearance came later and slower, as computers spread the secretary’s job across every desk. In 1980, writing a memo, booking travel, filing an expense report and running a calendar were each a separate person’s job. Those tasks have since moved into every employee’s own workday, unpaid and unlisted. No worker in 1995 voted on becoming a part-time typist, filer and travel agent. The change arrived through software updates, one release at a time.
That shift showed up as a keyboard on every desk. The AI shift shows up as a dashboard tracking the overnight work of AI agents, software that works through tasks on its own. Few employees get asked whether they want to manage those agents, yet most will wake up to what the agents produced. Some of it will be fine to file away, and some will need a person to step in and settle an exception. The job becomes watching the work, correcting it and deciding what to do with it.
Managing agents takes the discipline of delegation: deciding what to hand off, then checking the results. The consultant Ari Meisel wrote a book, The Replaceable Founder, about that problem years before AI. Founders succeed, he argues, once they learn to run a business without doing every task themselves. Anyone handed a team of agents is running a small operation, whatever the job title says.
