Klarna cut deeply into customer service with AI, then had to bring humans back. That was not a messaging issue. It was an early sign of the rehiring crisis that follows when companies move faster than they can adapt.
What this shows is not that AI failed. It shows that the organization changed faster than it could absorb the change. Service quality dropped. Trust thinned. The work did not disappear because the headcount did.
In periods of fast change, headcount is a fake scoreboard. The only one that matters is epimetabolic rate: how quickly a company can turn new knowledge into new roles, norms, and decisions without hollowing out judgment or customer experience.
If an AI plan outruns a company’s epimetabolic rate, the savings are temporary. The costs return as rework, degraded service, and expensive rehiring. The real question is not how many people AI can replace. It is how fast an organization can metabolize what AI disrupts.