Revelio Labs CEO Ben Zweig: 9% of firms are seriously adopting AI, freelancers are hardest hit, CS enrollment down 28%

Revelio Labs: What workforce data actually shows about AI adoption

Ben Zweig built Revelio Labs around a straightforward premise: instead of surveys or payroll feeds, scrape and synthesize everything public — LinkedIn profiles, job postings, Glassdoor reviews, layoff notices, immigration filings, freelance platforms — then classify it by occupation, skill, seniority, and work activity. Hedge funds are the primary customer, using the data to assess workforce dynamics at companies they don't operate. HR departments use it for benchmarking; academics use it for research.

The 9% figure

The cleanest measure Zweig has for serious AI adoption is whether a firm is actively hiring AI integration teams. By that threshold, 9% of firms qualify. It's conservative by design, but Zweig argues it correlates well with other signals and is a reasonable proxy for companies embedding AI into core business processes rather than just paying for subscriptions.

Revelio also has a data-sharing arrangement with Ramp, using spend data on AI token purchases as a parallel adoption signal. The approach works conceptually but has friction — Ramp's data is anonymized, skews toward tech companies, and requires a matching process between Zweig's labor data and Ramp's spend data to produce anything usable.

“By that metric we see about 9% of firms getting very serious about AI — it's a very conservative way to measure AI adoption but seems to be pretty good, correlated with all sorts of other things. Freelancing is hit pretty hard — that's really an environment where people transact tasks. Computer science enrollment is down 28% from its 2022 peak, which I think is an overreaction.”

Where job losses are actually appearing

Broad AI adoption is generally correlated with firm-level growth, not headcount cuts. But there are pockets of real displacement. Freelancers are the clearest casualty — Zweig says freelancing is down across the board — because the freelance economy runs on tasks, and AI is good at tasks. Stock photography, b-roll video, podcast intro music, basic copywriting, one-off website builds, logo mockups: all of this was previously transacted on platforms like Upwork and 99designs, and AI has absorbed a significant share of that demand.

The harder-to-quantify version of this is that many of these weren't formal jobs. They were micro-transactions. So the displacement may be real and material without showing up cleanly in employment statistics.

CS enrollment down 28%

Computer science enrollment is down 28% from its 2022 peak. Zweig reads this as an overreaction. The firms most aggressively adopting AI are not shedding engineers — if anything, engineering work is shifting toward DevOps and orchestration rather than disappearing. Agentic tools favor people with engineering backgrounds, not replace them. His analogy is radiology: for years, AI researchers predicted radiologists would be automated away, and instead there's now a shortage with elevated wages. He suspects a similar dynamic is possible in software engineering.

The hiring signal problem

The hires-to-postings ratio is down 38.6% since late 2022, and Zweig is more worried about this than about job losses. The mechanism is a matching breakdown on both sides. Candidates are using AI-assisted auto-apply tools — some job boards, including Indeed, are actively enabling mass applications — so a new posting can receive a thousand applications within minutes. Most are AI-generated and hard to verify. Employers, signal-jammed and unable to effectively screen at scale, are retreating to professional networks.

There's a regulatory asymmetry making this worse. Using AI to screen candidates became legally restricted a few years ago over concerns about algorithmic bias, creating a situation where candidates can flood the market with AI-generated applications but employers face liability for using AI to filter them. Zweig doesn't know how actively the rules are enforced, but the asymmetry is real and is contributing to a low-hire, low-fire economy with less labor market mobility overall.

The obvious fix — algorithmic filtering that surfaces the top ten candidates the way social feeds filter out low-quality content — exists as a technical possibility but remains largely unavailable to employers, partly because of that regulatory exposure.