Peter Cappelli

Organizations are deciding how many people they will need before they know what AI will do to the work. Peter Cappelli directs Wharton’s Center for Human Resources and has spent four decades studying how employers hire and manage people. Fortune calls him “Wharton’s great contrarian,” though he calls himself a skeptic. He checks big claims about work against what employers actually do. He takes AI seriously and hosted a season of Wharton’s podcast Where AI Works. He is a regular participant in the CEO Academy and McKinsey’s CEO events, and he won Wharton’s executive MBA teaching award six times.

His finding is that AI changes tasks long before it changes headcount. One reason is arithmetic: if AI takes over the report writing that fills 20 percent of a school principal’s time, he writes, the school still needs the whole principal. Another is that AI’s work still has to be checked, and someone has to handle the problems it cannot solve. The gains that could mean fewer workers come much later, he writes.

One operation he studied, in a case published in Harvard Business Review, put AI to work processing insurance claims. At first the AI cost three times as much as the employees doing the work, and getting it running took a six-person team a year. Headcount fell from 44 to 39, and the operation expects productivity to triple. Most of the staff stayed because, in his words, “lots of problems have to be chased down.” “It’s hugely expensive to do this,” he says. “And this was a success.” Breaking jobs into tasks, he says, has to be done with the employees, “because the employees really do know how their job is done.”

AI is the newest claim about work he has tested. Older claims about hiring, layoffs, the office, and generations still drive costly decisions. His book Our Least Important Asset shows why companies manage people so poorly: accounting counts them only as a cost, and practices such as layoffs are associated with worse financial results. Only about a third of U.S. companies say they check whether their hires work out, he reported in 2019. In Praise of the Office, his book with Ranya Nehmeh, sets rules that keep hybrid teams working. A 2020 National Academies report found generational labels unsupported by research.

WHY AUDIENCES WILL BE BETTER OFF

• Audiences leave able to sort their hiring, training, and staffing decisions into the ones to make this year and the ones that can wait.
• Audiences leave ready to sit down with the people who do a job, decide task by task what AI takes over, and plan for the money and time it requires.
• Audiences leave ready to measure what layoffs, hiring mistakes, turnover, and open roles cost the business, and to make the case for fixing each one.
• Audiences leave able to set clear rules that keep hybrid teams collaborating and keep new hires from falling through the cracks.
• Audiences leave able to drop workforce worries the evidence does not support, from generational differences to skills shortages, and focus on what they can control.