> **TL;DR.** In June 2026, economists at Ramp and Revelio Labs published the first study that measures AI adoption not through surveys or theoretical "occupational exposure" scores, but through actual company payments to AI vendors — across 21,559 U.S. firms. The result breaks the dominant narrative: companies that adopted AI seriously **grew headcount by roughly 10% over two years**, and their entry-level hiring rose 12%. Companies that just "bought a subscription and moved on" got nothing — no growth, no layoffs either. Here is what exactly was measured, where the limits of the conclusions are, and what it means for a business deciding whether to adopt now or wait.

## What happened: numbers instead of scare stories

The public narrative of the past few years goes like this: AI will take jobs, starting with junior staff. CEOs of large companies routinely blame layoffs on "AI adoption," while AI developers themselves oscillate between promises of unprecedented wealth and warnings of mass unemployment.

The study [«A New Look at AI's Impact on Jobs» (Kharazian, Simon, Stevens, June 2026)](https://ramp.com/data/ai-jobs-impact) is the first to test that narrative against what companies actually **do**, not what they say. The headline results:

- **Total headcount: +10.2% over 24 months** after adoption — for companies with high AI usage intensity. Light users show no statistically significant change.
- **Entry-level roles: +12%.** Junior headcount at intensive AI users grew even faster than total headcount — contrary to predictions that AI would displace juniors first.
- **The growth is broad, not just in the engineering department:** engineers, sales, administrative roles, customer support, finance — nearly every function expanded.
- **The effect builds gradually:** the first signs of growth appear 6–12 months after adoption, and the curve keeps rising. It looks like a learning curve: companies first search for use cases and rebuild processes — and only then capture the payoff.

## Why this data deserves more trust than surveys

Until now, nearly all research on "AI and employment" relied either on surveys ("do you use AI?") or on theoretical occupational-exposure indices — which tasks AI *could* perform. The two methodologies regularly contradict each other.

The measurement here is different. Ramp is a corporate card and bill-pay platform: the authors observe every real payment a company makes to AI vendors — models, APIs, coding agents, GPU clouds. Revelio Labs tracks monthly employment histories for the same companies. Linking the two produced a panel of 21,559 U.S. firms from January 2021 through February 2026.

"Adoption" is not a one-off experiment but sustained spending: at least $100 per month paid to AI vendors for three consecutive months. Intensity is AI spend per employee: light users average $2.78 per employee per month, heavy users $33.67. And the entire effect sits in the second group.

## Honest caveats: what the study does not prove

Good reasons not to turn the result into a slogan of "AI creates jobs, case closed":

1. **Adopters are not random companies.** Firms that adopt AI are already larger, more technical, faster-growing, and more likely to be venture-backed *before* adoption. The authors state this openly — which is why they compare early adopters not with never-adopters, but with similar companies that adopt AI later and simply haven't yet.
2. **The sector breakdown is uneven.** Statistically significant growth so far shows up only in the Information sector — software, internet, media (+13.4%). In other industries the effect is either small or not yet visible: AI tools there haven't reached product maturity.
3. **This is about firms, not the economy as a whole.** The study says: adopting firms grow. It does not overturn other evidence — for example, declining hiring of young workers in the most AI-exposed occupations economy-wide. Growth at adopters may partly come at the expense of competitors who didn't adopt.

## The key practical takeaway: an intensity threshold

The most underrated part of the study is not the "+10%" but the gap between the groups. Companies that bought a couple of chat subscriptions ($2–3 per employee per month) got **no** measurable effect at all. All of the growth belongs to those who invested seriously: coding agents, APIs, multiple models and vendors, redesigned workflows.

| | Light adoption | Intensive adoption |
|---|---|---|
| AI spend per employee per month | $2.78 | $33.67 |
| Total headcount over 24 months | no significant change | **+10.2%** |
| Entry-level roles | no significant change | **+12.0%** |
| Typical toolkit | chat subscription | agents, APIs, multiple vendors |

In other words: a chatbot subscription is not AI adoption. The effect appears when AI is genuinely embedded in daily work and takes over whole tasks, while the freed-up capacity goes into growth — new products, sales, customer support.

## What this means for sellers and small businesses

The study is built on U.S. data, but the logic transfers directly:

- **"Will AI replace my employees?" is the wrong question.** The data points to a different fork: companies that use AI intensively grow faster than those that wait. The risk is not that AI takes your team's jobs — it's that a competitor who adopted AI earlier takes them.
- **A pilot ≠ results.** Three months of experimenting with a chatbot changes nothing — that is literally visible in the data. The payoff comes when routine work (reports, reviews, ad bids, restocking, documents) is reliably done by AI, and people focus on what grows the business.
- **The effect takes time.** 6–12 months before the first measurable results is a normal horizon, not a sign that "it doesn't work."

At [Samreshuuu](/) we build exactly the usage model that produces the effect in this study: not "yet another chat," but an agent that takes over a seller's tasks end-to-end — through the marketplaces' official APIs, following rules you set in plain language. That is precisely the move from the left column of the table to the right one — without hiring a dedicated AI department.

## FAQ

**So AI doesn't eliminate any jobs at all?**
No — that conclusion does not follow from the study. It shows that at the level of individual companies, AI adoption is associated with headcount growth, not layoffs. Economy-wide dynamics — including the reallocation of hiring across occupations and firms — are a separate question, and the evidence there is more mixed.

**Why are juniors growing? Everyone said they'd be replaced first.**
The authors see the opposite in the data: at intensive AI users, entry-level headcount grew 12% — faster than the total. One explanation: AI lowers the cost of training a newcomer and makes them useful from day one, so hiring juniors becomes more attractive, not less.

**My business isn't in tech — is this even relevant to me?**
So far the measurable growth is concentrated in the Information sector — where AI tools matured first. But the authors say it plainly: this is an early point on the adoption curve, not a ceiling. For other industries the question is not "does AI work" but "does a tool for my routine exist yet." For marketplace selling — it already does.

**What does "intensive adoption" mean in practice?**
Not the number of subscriptions, but the share of real work AI performs. The study measures intensity in dollars per employee, but behind the dollars is behavior: agents and APIs instead of occasional chat, multiple tools, deep embedding in daily processes.

---

*Source: Kharazian, A., Simon, L., & Stevens, R. (2026). A New Look at AI's Impact on Jobs: Firm-Level AI Spending and Workforce Adjustment. Ramp Economics Lab. [ramp.com/data/ai-jobs-impact](https://ramp.com/data/ai-jobs-impact)*