Every few weeks another headline runs the same experiment in reverse: AI is coming for your job, here's the layoff number, here's the doom quote from an economist. But it's worth sitting with a data set that actually says the opposite, especially because almost nobody outside finance Twitter noticed it. Ramp's Economics Lab joined its own firm-level spending data — which AI tools companies actually pay for, and how much — with workforce records from Revelio Labs, across more than 21,000 U.S. companies. The question was simple: do companies that spend the most on AI end up with fewer employees two years later, or more?
More. Firms in the top tier of AI adoption grew headcount 10.2% over the two years following adoption. Firms in the bottom tier — the ones dabbling, not investing — saw no statistically significant change at all. And the part that should stop you if you're running hiring plans off the "AI kills entry-level jobs" assumption: entry-level headcount at the heavy-adopter firms grew 12%, faster than headcount overall. The workforce share of entry-level roles went up, not down. Bernstein picked this up in an analyst note in early September, arguing it could be a tailwind for software companies rather than the margin-compressing threat everyone's been pricing in — comparing it to the PC and internet productivity booms, when companies mostly plowed efficiency gains into growth instead of headcount cuts.
"It's a promising (and intuitive) result for young people who have gotten seemingly exclusively negative news about their job prospects in recent years."
— Ara Kharazian, Lead Economist, Ramp, Ramp Economics Lab
Here's what I don't want you to do with this number: don't walk into your next planning meeting and announce that AI is now confirmed net-positive for jobs, so hire like it's 2021. That's the same lazy read in the opposite direction. The researchers are explicit that this is correlation, not causation — companies spending heavily on AI were already larger, faster-growing, and more technical than the ones that weren't. Correlation with a confound that big means the honest conclusion isn't "AI creates jobs." It's that the kind of company disciplined enough to operationalize AI at scale is also the kind of company growing for other reasons, and AI spend is a marker of that discipline, not necessarily its cause. Less exciting than the headline. Also the version that's actually useful to you.
Now here's where I disagree with how this finding has been summarized. Bernstein's note, as Investing.com reported it, grouped finance, scientist, and operations roles together as the functions with weaker or no statistically significant growth.
The Ramp paper itself doesn't support that grouping. Scientist headcount at heavy-adopter firms rose significantly, right alongside administrative roles. Finance grew, but only marginally. Operations is the one category where heavy adopters didn't grow headcount at all. So the pattern isn't "AI protects headcount" or "AI eliminates headcount," and it isn't a tidy story about knowledge work either.
My read, and it's my read rather than the researchers', is that operations is where AI output is hardest to trust without someone who owns the downside. An ops decision with real operational risk doesn't get a cheap second look from a non-expert. That's less a story about AI replacing ops people than about which functions have someone qualified sitting next to the model to catch what it gets wrong.
That gap is the staffing problem nobody's built a plan around.
If you're a CTO or VP Eng who read "heavy AI adopters grow headcount 10%" and concluded "hire more junior people, AI will handle the rest," you've got the causality backwards. The firms in this data grew because they had the organizational maturity to deploy AI well — clear ownership, review processes, people who could tell good output from confident-sounding garbage — and that maturity is exactly what most companies scaling up AI usage right now don't have. You don't get the 10% growth by buying the tools. You get it by filling the roles that make the tools safe to lean on: the senior engineer who reviews the AI-assisted PR, the ops lead who catches the automation's blind spot before it ships, the person whose whole job is knowing when the model's answer is wrong in a way that looks right.
I've placed enough technical and ops hires this year to tell you what's actually scarce right now, and it isn't people who can run a copilot. Every mid-level candidate can do that already. What's scarce is the senior layer above them: five-plus years in the domain, the pattern-matching to spot a plausible-but-wrong output before it becomes a decision. That's exactly the layer companies stopped hiring for during the "AI replaces senior headcount" scare of the last two years. If Ramp's data holds up, the companies winning the next round of growth are the folks who kept or rebuilt that layer instead of cutting it. Check your org chart before you become the next layoff headline.
Sources: Ramp Economics Lab — Companies hire more after AI adoption; Ramp Economics Lab — full paper; Bernstein note via Investing.com — AI hiring boom could turn into a surprise tailwind for software.