ServiceNow's CEO Just Told Every New Grad Their Career Is Already in Trouble

March 15, 2026

Bill McDermott went on CNBC last Friday and said something that nobody in a C-suite is supposed to say out loud. Unemployment for new college graduates, he told Squawk on the Street, "could easily go into the mid-30s in the next couple of years." He said it plainly. No qualifiers, no pivot to a silver lining about upskilling. Just: here is what's coming, and it's aimed at you specifically if you just finished a four-year degree.

I think he's right. And I think the reason it landed so strangely - the uncomfortable pause, the way some anchors sort of laughed it off - is because we are still in the part of the AI story where the people doing the disrupting are supposed to pretend they're actually creating opportunity. McDermott just skipped that part.

What He Actually Said, and Why It Matters More Than One Headline

His specific point was that for non-differentiating roles, "so much of the work is going to be done by agents" - and that "it's going to be challenging for young people to differentiate themselves in the corporate environment." That's not a vague warning. That's a description of what entry-level work has always been: the non-differentiating stuff. The work that teaches you. The work you do while you're building toward something.

Entry-level jobs are not just jobs. They are the mechanism by which people learn how companies actually function. You spend two years doing things that feel trivial - logging calls, drafting briefs, updating trackers, supporting more senior people - and in doing so you absorb how an organization moves. You learn the informal rules. You figure out who actually makes decisions versus who gets credit for them. Nobody teaches you that stuff. You acquire it by showing up and doing the boring work.

AI agents are eating that work. Not the strategic work - the boring work. Which sounds fine until you realize that's the exact work that prepares you for the strategic work. Agents now handle routine tasks end-to-end - tickets, drafts, QA, reporting - leaving less space for "learning by doing." That's not a productivity problem. That's a developmental crisis wearing productivity's clothes.

I mentioned this to Tory on Tuesday. He said something about how every generation has faced disruption and come out stronger, which I think was meant to be reassuring. He said it very cheerfully. He's been sleeping in his office for three weeks since the separation from Melinda, which I did not bring up.

Baroque oil painting of a young figure in academic robes standing at the base of a grand stone staircase, looking up toward a heavy door at the top that is nearly closed, with a narrow beam of warm amber light cutting through deep dramatic shadows — Wanted to make something that felt like the whole article in one image - a person at the bottom of a staircase looking up at a door that is almost shut. Derek said it looked like a painting from a church and asked if that was intentional, which it was not, but I told him yes.

The Numbers Before the Prediction Even Arrives

Here's the thing about McDermott's mid-30s projection: it hasn't happened yet, and the current numbers are already bad. The Federal Reserve Bank of New York put the unemployment rate for recent college graduates at the end of 2025 at about 5.7%, and the underemployment rate of 42.5% was the highest level since 2020. That underemployment number is the one people are glossing over. Nearly half of recent graduates are working jobs that don't use their degrees. Before AI agents fully land. Before this accelerates.

That 42.5% figure deserves more attention than it's getting. It means the crisis McDermott is predicting is not a future scenario - it's already the present scenario with better optics. The degree gets you hired somewhere. Just not somewhere relevant to the degree. And if agents now handle the entry points into relevant work, the path from "somewhere" to "somewhere meaningful" gets dramatically harder to navigate.

Compared to previous technological revolutions, experts say AI is chipping away at many white-collar jobs, including coding and marketing roles, and allowing companies to reduce hiring and improve productivity with fewer workers. Previous automation waves hit manufacturing. Factory jobs. Things that were easy to point to. This one is hitting the knowledge worker pipeline at the exact moment people enter it. That's a different kind of problem.

McDermott Is Selling Something - That Doesn't Make Him Wrong

The obvious critique of everything McDermott said is that he is the CEO of a company that sells the exact technology he's warning about. ServiceNow's tools will help businesses slash hiring costs, and the firm has already taken out 90% of the use cases that previously relied on humans in customer service. He is not a neutral observer. He has strong financial incentives to describe the world in a way that makes his product sound inevitable and necessary.

And yet. The conflict of interest doesn't make the prediction wrong. Sometimes the person most motivated to describe a trend accurately is the person profiting from it. McDermott knows exactly how fast enterprise adoption is moving because he's watching it happen across the Fortune 500 in real time. Almost nine in ten companies in the Fortune 500 use ServiceNow's cloud-based platform to automate their workflows. That's not a vendor claiming reach. That's a vendor with actual visibility into what is being automated, at what speed, inside the companies that employ the most people.

When he says "I do think it's coming quicker than people anticipate," he's not speculating. He's reporting.

Palantir CEO Alex Karp has said he wants to grow revenue by 10 times while reducing headcount. Amazon CEO Andy Jassy has said the company will also shrink its corporate workforce with AI tools. Atlassian, whose stock has dropped 54% this year amid AI disruption concerns, said this week it would lay off about 10% of employees to support AI investments. These are not fringe predictions from futurists. These are sitting CEOs announcing operational plans on earnings calls and business television.

My travel agent for the Americas - the one I use for anything outside Europe - mentioned something about her niece graduating in May. Communications major. I didn't know what to say, which is unusual for me. I just told her the graduation trip should be somewhere meaningful. That felt like the wrong answer but it was the only honest one I had.

What This Actually Means for People Running Businesses

I want to be direct about this, because I think the coverage of McDermott's comments has gone in two directions - either it's treated as a sky-is-falling hot take, or it's reframed as an upskilling opportunity - and neither of those responses is useful if you're actually running a business and thinking about hiring.

Here is the real implication: the talent pipeline is about to get structurally disrupted in a way that makes hiring harder, not easier, even as headcount shrinks. If agents are taking the entry-level work, and entry-level work is how people develop into mid-level capability, then in four or five years you will have a shortage of mid-level talent even as the total number of people with degrees keeps climbing. You won't be able to hire your way to experience. The people who would have been your competent 30-year-olds in 2030 are currently graduating into a market that won't give them the formation they need.

That's not a problem AI solves. That's a problem AI creates and then leaves on your desk.

The businesses that are quietly thinking about this - that are actively figuring out how to use AI tools to expand what their existing people can do rather than just reduce who they need - are going to be in a very different position in five years. At ServiceNow itself, AI agents are already doing what McDermott calls "90% of the soul-crushing work" in support, IT, and HR - but the company hasn't laid those people off. Instead, each of its 28,000 employees has undergone an AI skill assessment and begun learning what they'll need to be competitive in an agentic workforce. That's a specific choice. Not every company is making it.

We had a whole conversation in the office about this. Derek said it wouldn't affect people in creative fields, which I thought was interesting because he said it very confidently right before showing me an AI-generated Star Wars image he'd been working with for twenty minutes. Chris just listened and nodded and looked thoughtful. He always does that. It's very calming and contributes nothing.

The Part Nobody Wants to Acknowledge

There is a version of this conversation that always gets steered back to "new jobs will be created." And maybe that's true. I don't have a strong opinion on the 30-year arc of labor economics. What I do have a strong opinion on is that the transition is not going to be smooth, and the people absorbing the cost of the transition are going to be the ones entering the workforce in the next five years - not the executives announcing the efficiencies.

McDermott himself explained it bluntly: "I can literally have the same headcount going out of this year as I came to this year with, expand free cash flow margin, grow my revenues by an even greater amount" - and assumed that other CEOs watching were asking themselves why they weren't doing the same. That's not the language of someone describing a regrettable side effect. That's a sales pitch. A genuinely effective one. One that a lot of CFOs are hearing right now and finding very persuasive.

The reality of AI integration for most businesses isn't the elegant workflow transformation McDermott describes on television. It's messier, slower, and more expensive than the pitch suggests. But the direction is not ambiguous. The entry-level work is going. The question is whether the people who used to do that work are going with it, or whether they're being brought forward into something more valuable. Most companies are not thinking hard enough about that question. They're thinking about the line item.

Linda stopped by while I was writing this. She said Gerald's nephew is graduating this spring and couldn't find an internship anywhere - applied to something like forty places. She seemed genuinely rattled. Linda has been married 31 years and raised three kids and is not easily rattled. I think that detail landed differently for me than the CNBC segment did.

What McDermott Got Wrong - and Why It Doesn't Actually Help

The one thing I'll push back on is the framing of this as primarily a graduate problem. Unlike previous technological shifts, AI is now targeting white-collar roles such as coding and marketing, allowing firms to do more with fewer workers. That's not just entry-level. The compression is happening across the career ladder. Graduates are the most visible and most vulnerable, but the mid-career professional whose value was built on doing a lot of the things AI now handles faster is in a position that's more precarious than it looks right now.

The disruption McDermott is describing starts at the bottom because that's where the non-differentiating work is concentrated. But the logic doesn't stop there. Differentiation now depends on context, judgment, and owning outcomes - not task completion. That standard moves up the ladder too. It will keep moving. The people who are safe are the ones whose value was never about task volume in the first place.

For anyone looking at how their sales or outreach operations are running, or thinking about what's worth automating versus what's worth protecting, the thinking happening around sales engagement platforms right now reflects a lot of the same tensions - what do you hand to the machine, and what do you keep human, and what does that choice cost you in ways that don't show up on a quarterly report.

McDermott said what a lot of people in his position know but don't say. That's worth something, even if he's also the one selling the shovel. The graduates walking into this market deserve to understand what they're walking into. The businesses hiring them - or deciding not to - deserve to understand what they're actually optimizing for when they make that call.

I don't think the answer is panic. I think the answer is honesty. And on that front, McDermott's Friday morning segment was more useful than a year's worth of reassuring conference keynotes about AI empowering the workforce. He just said the thing. That's rare enough to be worth noting.