The narrative surrounding artificial intelligence and its imminent threat to jobs is being met with growing skepticism from experts who have seen this story before. According to Peter Cappelli, the George W. Taylor Professor of Management at the Wharton School, the current AI frenzy feels more like a replay of past technological hypes than a genuine revolution.
Learning from the Driverless Truck Debacle
Cappelli traces a direct line from today's AI exuberance to the mid-2010s, when major consultancies and the World Economic Forum predicted the near-total elimination of truck drivers. The forecasts were dramatic but the reality was far less accommodating. Cappelli points out that practical hurdles, like refuelling, deliveries, and the need for human oversight, quickly made the idea of completely driverless trucks impractical. "If they have to have an employee sitting with them, of course, it defeats the purpose," he told Fortune.
He argues that this gap between technological possibility and operational reality is where most futuristic labour forecasts fall apart. The current AI hype, Cappelli suggests, is largely generated by the sellers of the technology, whose sensational predictions about mass job losses often overshadow the complex realities of implementation.
The High Cost and Slow Pace of Real-World AI
Cappelli's skepticism found concrete support in late 2025, following an influential MIT study that found 95% of generative AI pilots failed to deliver any meaningful return. This data disrupted the popular narrative of AI as a quick and cheap efficiency booster.
His preferred counter-narrative comes from a Harvard Business Review case study on Ricoh, an insurance claims processor. This is exactly the kind of repetitive administrative work often cited as ripe for AI automation. The assumption was that AI would handle it effortlessly. The execution told a different story.
Ricoh spent a full year with a six-person implementation team, including three outside consultants, and paid roughly $500,000 in consulting fees just to get the system running. The first discovery was stark: large language models could do the work, but at three times the cost of human employees. After extensive optimisation, the company was still spending about $200,000 a month on AI fees, which was more than its total previous payroll for the task.
Modest Job Impact and the Enduring Human Role
The headcount reduction at Ricoh was far from the mass displacement often forecast. The team shrank only modestly, from 44 employees to 39. Cappelli explains that technology rarely removes the need for people; it reshapes their roles. "The reason they still need employees is that lots of problems have to be chased down, and they're harder to chase down if they come off of AI," he said.
Ultimately, the Ricoh division did become three times more productive. However, the journey exposed an uncomfortable truth: progress was slow, expensive, and did not eliminate human oversight. "So that's the payoff, but it's not cheap [and] it took a hell of a long time to do," Cappelli concluded.
At a time when AI is often framed as an unstoppable force destined to erase work, Cappelli's research offers a grounded, data-driven perspective. AI will likely reshape jobs, but not on the rapid timelines or at the catastrophic scale its most vocal advocates promise. True understanding comes not from embracing hype, but from asking whether the promised future can actually function in the complex real world.
