The swap that doesn't add up

General Motors cut 600 IT workers while simultaneously posting roles for AI-native developers and prompt engineers. Not a reduction, a swap. The old team gets severance; the new team gets job listings before the desks are cold.

GM isn't alone in the musical chairs. GitLab announced a 7% workforce reduction and reorganised for the 'agentic era', framing the whole thing as transformation rather than cost-cutting.

And SAP launched its Autonomous Enterprise vision at Sapphire 2026 with 200+ AI agents rolling out across finance, HR, procurement, and supply chain. The message from every stage and earnings call is consistent: the future belongs to organisations that replace human workflows with agentic ones.

Here's the problem with the swap. A landmark Gartner study found that firms cutting staff were no more likely to see returns than those that didn't. The highest-performing companies weren't replacing workers with agents. They were using AI to amplify the people they already had.

That finding should stop every executive mid-restructuring memo. The playbook GM, GitLab, and others are running has no empirical basis for generating returns. As Gartner's analysts put it: "Workforce reductions may create budget room, but they do not create return."

The restructuring fallacy

This pattern has a history. The corporate reengineering craze of the 1990s promised that redesigning business processes around new technology would unlock massive efficiency gains. Firms restructured aggressively. A few years later, studies found that most reengineering projects failed to deliver their projected returns, not because the technology didn't work, but because organisations treated restructuring as the strategy rather than a consequence of having one.

The same logic applies. SAP deploying 200+ agents across enterprise workflows is a product bet, not proof that replacing human judgement with agentic workflows produces better outcomes. GitLab reorganising for the 'agentic era' might work brilliantly or might just be a different shape of org chart painted with AI branding. GM swapping 600 IT workers for AI-skilled replacements assumes the bottleneck was the people, not the processes, priorities, or institutional knowledge those people carried.

The way I see it, the companies getting real returns from AI aren't the ones making the loudest structural changes. They're the ones doing the quiet work of figuring out where AI actually helps their existing teams make better decisions or handle complexity that was previously unmanageable. People amplification over replacement.

That distinction matters for anyone building products or running teams right now. If your AI strategy starts with a headcount target, you're optimising for the wrong metric. The Gartner data suggests the question isn't "how many people can we replace with agents?" but "where does an agent make a good person better?"

Every restructuring announcement this week used the language of inevitability. The evidence points somewhere more uncomfortable: the swap might feel like progress while delivering nothing but churn.

---