Ramp's economics team spent the last several months doing something almost nobody bothers to do with AI claims: they measured it. Instead of surveying executives about their intentions, Ramp pulled actual corporate-card and bill-pay data — real dollars flowing to OpenAI, Anthropic, and every other AI vendor — for more than 21,000 U.S. companies, then linked it to Revelio Labs' workforce records to see what happened to headcount after the spending started. The finding: companies with sustained, high-intensity AI spend per employee grew headcount 10% over the following two years. Entry-level headcount grew 12%. Low-intensity adopters — the companies dabbling — saw no statistically significant change either way.
"Our research shows that firms that invest more in AI also hire more following adoption, including in entry-level roles."
— Ara Kharazian, Lead Economist, Ramp (PR Newswire)
Two weeks after that study circulated, Sprout Social started notifying 260 people (20% of its staff) that their jobs were gone. CEO Ryan Barretto framed it as a move "from a position of strength," not distress: the company raised guidance to the high end of its prior outlook, the stock popped 7% on the news, and the stated purpose of the cut was to fund "ongoing investments in AI-powered social intelligence." Sprout joins Oracle (21,000 over the past year, disclosed in a June filing), Cisco, and Meta on the growing list of companies pairing a real layoff with the word "AI" in the same sentence as the justification.
Read those two paragraphs back to back and you'll notice they're using identical language to describe opposite motions. Ramp's data says real AI investment — the sustained, per-employee kind — correlates with more people on payroll, not fewer. Sprout Social's announcement says an AI investment just cost 260 people their jobs. One of these is describing what building actually looks like.
The other is describing a budget reallocation wearing a hiring headline.
That's not a gotcha against Sprout specifically — a services company facing margin pressure has every right to restructure, and "invest in AI" is a defensible strategic direction even attached to a cut. The point is narrower and more useful: Ramp just handed every CTO and hiring manager a way to tell the two stories apart that doesn't depend on trusting either company's press release. Real AI adoption, measured in dollars sustained over months, shows up as growth, most visibly at the entry level, where the people running and validating model output actually get hired. A layoff that cites AI as the reason for the cut, with no rehire commitment attached, isn't the same motion. It's cost-cutting borrowing the vocabulary of a growth story, and after two years of headlines equating "AI investment" with "fewer humans," that vocabulary is now cheap enough that almost anyone can spend it.
The entry-level number is the one worth sitting with longest. For two years the working assumption in every planning meeting has been that AI kills the junior req first. Why hire someone to do what a model does for free? Ramp's data says the opposite is happening at the companies actually running the AI motion at scale: junior headcount is growing faster than headcount overall. The scarce skill isn't writing code — it's judgment about model output, and judgment is trainable in a junior hire faster than it's definable in a job posting. Companies that are actually building are discovering they need more people to do that, not fewer.
So before you take any company's "we're investing in AI" line at face value — including when it's attached to your own headcount plan — ask the Ramp question instead of the press-release question: is headcount going up in the two years after this money starts moving, especially at entry level? If the answer is no, you're not looking at an AI investment. You're looking at a budget line that needed better cover.
VC5 Consulting builds hiring pipelines for companies running the real AI motion — including the entry-level and junior roles that keep getting cut on paper and rehired in practice once the model-output judgment gap shows up. If your AI investment is supposed to grow headcount and you don't have a pipeline ready for it, let's talk.