The Companies Spending Most on AI Are Hiring, Not Firing

The layoff headlines say one thing. The spending data says another. Through May 2026, companies blamed close to 90,000 job cuts on AI, and some forecasts say AI could wipe out 15% of U.S. jobs in five years. That's the story your team is scared of. But the numbers underneath tell a stranger story. Let me catch you up.

The spenders are the hirers

Here's the finding that should reset your planning. Companies that spend heavily on AI are growing headcount faster, not slower. A report from Ramp and Revelio Labs tracked AI spend and workforce records across nearly 22,000 companies. The "high-intensity adopters," firms spending about $30 per employee per month on AI in their first three months, saw headcount rise 10.2%.

And it wasn't just engineers. Growth showed up across sales, admin, customer service, finance, marketing, and science roles. Even the jobs everyone assumes are doomed held up. In tech-forward firms, entry-level headcount actually rose 12%. So the pattern isn't AI replacing people. It's AI making core work cheaper, which makes it worth building a bigger firm around that work.

But the buffet only feeds the rich

Before you relax, read the fine print. The report skews hard toward tech-forward, knowledge-work firms, the kind with VC money that were growing fast anyway. The authors are honest about it: the paper "does not show that AI universally creates jobs," only that broad job loss isn't the whole story.

And buying subscriptions is not the same as winning. Companies that ran pilots but never made sustained investments saw no headcount gains at all. The firms pulling ahead already had the capital, technical staff, and management bandwidth to turn adoption into real business gains. The rest are stuck experimenting. If your budget is a few seats and a pilot, you're in the group that falls behind.

America builds it, then leaves it in the box

The gap between owning AI and using it shows up at the country level too, and it's wild. The 2026 AI Index found the country that leads AI development is not the one that leads adoption. The UAE tops usage at 64%. The U.S. ranks 24th, at 28.3%, despite pouring $285.9 billion into AI in 2025, more than the rest of the world combined.

Access isn't the problem. Americans get the same tools, same day, same price. What's different is trust and mood. Only 33% of Americans expect AI to make their jobs better. That fear is the drag. Your team feels it too, and it's the thing slowing your rollout more than any tooling gap.

The kids already decided

While leaders debate policy, the pipeline is moving. A study of 31,000 college syllabi found faculty shifting from blanket bans to task-based rules, allowing AI for some work and blocking it for others. Business courses moved fastest, with 27% now assigning AI-based tasks. Humanities stayed most restrictive.

The graduates you hire next year will arrive knowing how to work with these tools, at least in some fields. But the researchers flag a real risk: if AI does the skill-building tasks, students may graduate weak in exactly the areas AI is strong. You'll get people fluent at prompting who never built the underlying judgment. Plan your onboarding for that gap.

The deep cut

The fear that AI will shrink your team is the thing most likely to shrink your team. The spend data shows the firms that commit are the ones that grow. The country data shows the U.S. is 13 points below its expected adoption, dragged down by workers who expect AI to hurt them. Both point the same way: hesitation is the cost, not the safety move.

So don't frame your next AI decision as "how many roles can we cut." Frame it as "what can we now afford to build." And watch the trap Alison Gopnik names in her Berkeley talk: these models predict patterns from old data, but they don't do the open-ended, curious exploration a four-year-old does. That's still your team's job. Buy the tools to make more of that work possible, not to replace the people doing it.

Three questions for your team

1. Are we a real adopter or a pilot tourist? If we're paying for seats but haven't committed to sustained investment, we're in the group the data says gets no gains. What would committing actually look like this quarter?

2. What are we now cheap enough to build? If AI lowers the cost of docs, code, and internal tools, name one thing we couldn't staff before that we can staff now.

3. How do we onboard grads who can prompt but haven't built judgment? What do we teach in the first 90 days to close the skill gap the tools skipped?