The CEO whose company everyone says will automate jobs is now saying the companies actually using his tools are hiring, not firing.
The Summary
- Sam Altman claims companies adopting AI most aggressively are hiring more, not less, while firms citing AI for layoffs are laggards in adoption
- He admits underestimating how "jagged" AI capabilities are—stellar at specific tasks, terrible at long-term complex work
- 50% of Americans are more concerned than excited about AI in daily life, per March Pew poll, as tech leaders warn of job displacement
The Signal
Altman's argument flips the narrative, but he's also covering his ass. Companies like Block, Cisco, Coinbase, Snap, and Salesforce have all explicitly cited AI while cutting headcount. His counter: those aren't the real AI adopters, they're just using AI as cover for cuts they wanted to make anyway. That's probably true for some. But it also conveniently positions OpenAI on the side of job creation while its models get better at replacing knowledge work.
The more interesting admission is about "jaggedness." Altman said the models do some things incredibly well but fail at long-term, complex task supervision. This maps to what we're seeing in the wild: AI crushes narrow, repetitive tasks but collapses when you need judgment across weeks or months. The companies hiring are the ones who figured this out first.
"The companies adopting AI the most are hiring the most because they're expanding what's possible, not replacing what exists."
They're not using AI to fire their customer service team. They're using it to handle 10x the customer volume with the same team, then hiring more people to build new products now that support costs dropped. Or they're using Codex to let engineers ship faster, then hiring more engineers because they can suddenly tackle projects that were too slow before. The productivity gain creates new work. For now.
Key pattern emerging:
- Early AI adopters are in expansion mode, not contraction mode
- Laggards use "AI" as a socially acceptable layoff excuse
- The actual automation wave hasn't hit yet because models are still too unreliable for autonomous work
But here's the tension: Altman's optimism relies on AI staying jagged. The moment GPT-7 or Claude Opus 5 can handle "long-term, complex task supervision," his entire argument inverts. Then the companies that adopted early and learned how to orchestrate AI workflows will be the ones that can cut deepest. They'll have the institutional knowledge to trust the tools. The laggards will still be trying to figure out prompts.
Public sentiment is tracking this. Half of Americans are more concerned than excited about AI, and the gap between concern and excitement is 5:1. That's not irrational fear. People see tech CEOs like Mustafa Suleyman and Dario Amodei warning about mass displacement while Sam Altman breaks ground on a 1 gigawatt data center in Michigan. They're connecting the dots: more compute means more capable models means fewer jobs that require humans.
The Implication
If you're a company leader, the move is clear: adopt AI aggressively now while it's still jagged enough to require human oversight. Learn the workflows. Build the guardrails. Hire into the productivity gains. Because when the models smooth out, the companies that didn't learn how to orchestrate them won't suddenly catch up. They'll just execute the layoffs they should have done two years earlier.
If you're a worker, the signal is the same. The safe jobs aren't the ones AI can't do yet. They're the ones that sit between AI capabilities: reviewing AI output, directing AI workflows, handling escalations AI can't resolve. That middle layer is hiring. But it's a temporary equilibrium. When task supervision gets solved, that layer compresses fast.