Block cut 4,000 employees — 40% of its workforce — on February 27, 2026, and the stock jumped nearly 20% the same day. Wall Street rewarded a bet on AI efficiency that Block hasn’t proven yet. And neither has anyone else.
The February 2026 wave of AI-justified layoffs reveals a pattern: companies are cutting workers for productivity gains they can’t measure, while the data on retraining, rehiring, or job creation remains nonexistent.
The layoffs are real. The productivity gains are mostly theoretical.
Block’s cut wasn’t an outlier. Amazon eliminated 30,000 jobs in January 2026, focusing on management layers and AI infrastructure. CEO Jack Dorsey told investors the company could now operate with “smaller, more streamlined teams” thanks to new intelligence tools.
Goldman Sachs estimates generative AI exposes 300 million jobs worldwide to automation, but the gap between exposure and actual displacement is where the story gets murky.
Block joins a growing list of AI-justified workforce reductions that accelerated in early 2026. WiseTech cut 2,000 jobs (30% of staff) on February 25, citing AI-driven efficiency gains. Autodesk eliminated roughly 1,000 positions in January. Angi cut 350 jobs the same month, explicitly crediting “AI-driven efficiency improvements.”
But here’s the problem: Harvard Business Review found companies are laying off workers because of AI’s potential, not its performance. Only 6% of organizations report meaningful bottom-line impact from AI initiatives, according to McKinsey — despite $200 billion in global investment through 2025. Meanwhile, 30% of AI-adopting companies report quantifiable financial or productivity gains as of Q4 2025, up from 16% a year earlier.
That means 70% still can’t prove it works. Yet the layoffs accelerated anyway.
Nobody’s tracking what happens to the workers
This is where the story breaks down entirely. There’s no public data on retraining success rates for workers displaced by AI in 2024-2025. No documented cases of companies rehiring employees after AI implementations failed. No studies tracking placement timelines for “new-collar” roles like prompt engineering or AI training specialists.
The pattern echoes findings from earlier this year: companies have fired workers for AI that doesn’t work yet, but none have published data on what happened to those employees six months later.
And AI implementations fail constantly. Recent studies confirm what implementation teams already know: AI fails at real work more often than vendor demos suggest. Carnegie Mellon and Anthropic research found AI agents make “too many errors for unsupervised deployment in high-stakes business processes.” 51% of B2B go-to-market teams report AI implementation failures in 2026.
Despite these failure rates, not a single company has announced rehiring workers after an AI project collapsed. The narrative is unidirectional: layoff announcements with explicit AI justification, widespread implementation dysfunction, organizational chaos — but zero corresponding data on worker transitions or hiring recovery.
Federal Reserve research shows employment has declined 1% since late 2022 in the 10% of sectors most exposed to AI, with young workers hit hardest. But there’s no tracking of where those workers went, whether they found comparable roles, or how long transitions took.
The honest trade-off: betting on a future nobody can measure
The 30% gains figure is real progress — it doubled from 16% a year earlier. Some companies are extracting value from AI tools. But the central problem remains: companies are making irreversible workforce decisions based on projections, not outcomes.
The absence of retraining data, rehiring cases, or job creation metrics means this is a one-way bet. If AI implementations fail at the current 51% rate for some functions, there’s no public record of companies reversing course. Workers are the variable being optimized out while the optimization itself remains unproven at scale.
Block’s stock surged 20% on the promise of AI efficiency. McKinsey’s data shows only 6% of companies can prove bottom-line impact. The market has already decided which number matters more.
