ServiceNow CEO Says AI Could Push Grad Unemployment Past 30%
ServiceNow CEO Bill McDermott told CNBC that AI agents could drive college graduate unemployment into the mid-30% range, while his own company has automated 90% of customer service use cases.
5 min read
ServiceNow CEO Bill McDermott told CNBC's "Squawk on the Street" on March 13 that unemployment for new college graduates "could easily go into the mid-30s in the next couple of years" as AI agents take over entry-level knowledge work. The statement came from a CEO whose company has already automated 90% of the customer service use cases that previously required humans - and whose stock price depends on selling that automation to other enterprises.
TL;DR
- Claim: College grad unemployment could hit mid-30% range in the next couple of years due to AI agents
- Our take: The direction is right but the number is roughly 6x higher than any credible forecast
- Current college grad unemployment (ages 22-27) sits at 5.8%, per the New York Fed
- Senator Mark Warner's warning of 25% was already considered extreme; McDermott's mid-30% goes further with no data
- ServiceNow's own AI (Now Assist) has passed $600M in annual contract value - McDermott profits from the displacement he's warning about
What They Showed
McDermott's argument rests on the speed of AI agent deployment in enterprise software. ServiceNow's own tools handle 90% of employee IT requests autonomously, he said. The "soul-crushing" work - support tickets, HR inquiries, routine IT troubleshooting - is being done by agents that "work 24/7, never eat, and never need benefits."
He framed this as a warning, not a celebration. If companies broadly adopt agentic AI at the pace ServiceNow's customers are adopting it, the entry-level jobs that absorb new graduates dry up before those graduates finish their degrees.
ServiceNow's numbers back the adoption claim. Now Assist, the company's AI agent platform, has passed $600 million in annual contract value, doubling year-over-year. The company projects $15.5 billion in subscription revenue for 2026, up 20% from 2025. All 28,000 of ServiceNow's own employees have undergone AI skill assessments. McDermott says none were laid off - they were reskilled instead. But reskilling 28,000 people at a $200+ billion market cap company is a different proposition than reskilling a fresh graduate who hasn't been hired yet.
What We Tried
The claim is specific enough to check against existing data. Current unemployment for recent college graduates (ages 22-27) stands at 5.8%, per the New York Federal Reserve's College Labor Market tracker. The underemployment rate - graduates working in jobs that don't require a degree - hit 42.5% in Q4 2025, its highest since 2020.
| Metric | Current | McDermott's Prediction |
|---|---|---|
| College grad unemployment (22-27) | 5.8% | "Mid-30s" (~35%) |
| Implied increase | - | ~6x current rate |
| Timeline | Now | "Next couple of years" |
| Comparable historical peak | 8.9% (2010 post-recession) | ~4x the worst recorded |
| Senator Warner's estimate | 25% (over 3-5 years) | 10 points higher |
| Amodei's estimate (Anthropic) | 10-20% general unemployment | Narrower scope (all workers) |
For context: U.S. college graduate unemployment hasn't exceeded 9% in modern history, even during the 2008 financial crisis. McDermott's mid-30% figure would be roughly 4x the worst recorded rate for this demographic. The Great Depression peaked at 25% unemployment across all workers.
How Others See It
McDermott isn't the only one raising the alarm, but he's the most aggressive with numbers.
Senator Mark Warner called AI job displacement a "code red" in November 2025, warning that recent graduate unemployment could hit 25% over three to five years if unaddressed. That prediction was already considered extreme.
Anthropic CEO Dario Amodei said in May 2025 that AI could remove 50% of entry-level white-collar roles within five years, with overall unemployment spiking to 10-20%. Amodei explicitly called on AI companies to "stop sugar-coating" the threat.
At Davos in January 2026, other CEOs pushed back on Amodei's timeline, arguing that AI adoption in enterprises is slower than Silicon Valley assumes.
The Cleveland Fed published research in 2025 showing that young college graduates are "losing their edge" in the labor market, with the wage premium for degrees narrowing as AI handles more analytical tasks. But narrowing wage premiums and 35% unemployment are different orders of magnitude.
The Gap
McDermott's warning has two problems. The first is that no economic model, government projection, or academic forecast supports 35% unemployment for any demographic in any timeline discussed. The number appears to be a gut estimate from a CEO, not a data-driven prediction.
The second problem is the conflict of interest. ServiceNow's entire growth story depends on enterprises replacing human labor with AI agents. McDermott's prediction is also a sales pitch: the future is agentic AI, companies that adopt it early will thrive, and ServiceNow sells the tools to adopt it. Now Assist's $600 million ACV and 20% revenue growth targets make the incentive structure clear.
McDermott tried to thread the needle by saying AI "need not mean massive cuts to corporate headcounts" - that workers can be reskilled rather than replaced. But if every company reskills rather than lays off, the unemployment rate doesn't reach the mid-30s. The prediction contradicts his own proposed solution.
Verdict
The direction of McDermott's concern is correct. Entry-level knowledge work is the most exposed category to AI displacement, and the data is already moving in the wrong direction - 5.8% unemployment and 42.5% underemployment for recent grads, both trending upward. The jobs that traditionally absorbed new graduates (analyst roles, junior support, entry-level IT) overlap almost perfectly with what agentic AI does well.
But "mid-30s" is a number without a methodology, delivered on a financial news network by a CEO whose company earns more revenue the faster that displacement happens. The most useful version of this warning came from Amodei, who gave a range (10-20%), a mechanism (automation outpacing augmentation), and a call to measure the problem systematically through Anthropic's Economic Index. McDermott gave a headline.
Sources:
- AI agents could send college grad unemployment over 30% - CNBC
- College Labor Market tracker - New York Federal Reserve
- Young college grads losing their edge - Cleveland Fed
- Anthropic CEO warns 50% of entry-level roles at risk - Marketing AI Institute
- College grads face new obstacle in finding jobs: AI - CBS News
- Senator Warner warns of 25% grad unemployment - CNBC
- ServiceNow projects 20% subscription revenue growth for 2026 - Seeking Alpha
