67% of CEOs Say AI Is Creating Entry-Level Jobs. 66% of Enterprises Say It's Killing Them. Follow the Money.
Two surveys, conducted in the same quarter, asked nearly identical questions about AI's impact on junior hiring. They got opposite answers. The contradiction maps perfectly to who funded the research and who profits from the conclusion.
Two numbers, released within weeks of each other in early 2026, should have caused a small riot in business journalism. Neither did, because nobody put them next to each other.
Number one: An IDC survey found that 66% of enterprises are reducing entry-level hiring due to AI, with 91% reporting that AI had changed or eliminated jobs. Number two: A Teneo survey of 350+ public-company CEOs found that 67% of global CEOs say AI is increasing entry-level headcount. Not preserving it. Increasing it.
Same economy. Same quarter. Opposite conclusions. Nobody asked the obvious question: who paid for each study, and what were they selling?
The Fear Industry Has a Business Model
IDC is a technology market research firm. Its customers are enterprises buying AI tools and the vendors selling them. When IDC publishes a survey showing that 66% of enterprises are cutting junior roles because of AI, two things happen. First, AI vendors gain ammunition for their sales decks: "Your competitors are already automating entry-level work." Second, enterprises feel validated in their own cost-cutting decisions. IDC's revenue comes from subscriptions bought by both groups.
Gartner operates an identical model. In February 2026, Gartner published that 55% of supply chain leaders expect agentic AI to reduce entry-level hiring needs. Gartner's annual revenue exceeds $6 billion, overwhelmingly from enterprises and IT vendors paying for advice on technology adoption. Its surveys consistently find that enterprises need to adopt more technology. This is not corruption. It is incentive alignment so clean it barely registers as bias.
Cognizant provides the starkest case study. In January 2026, Cognizant released its "New Work, New World 2026" report finding that AI can handle $4.5 trillion in U.S. work tasks and affects 93% of jobs. Legal work exposure jumped from 9% to 63%. Average AI exposure sits at 39%, already 30% higher than the firm's own 2024 forecast for 2032. Its headline message: AI is eating work faster than anyone predicted.
That same quarter, Cognizant CEO Ravi Kumar told Fortune that AI is "an amplifier of human potential" and "not a displacement strategy." The company announced it was expanding entry-level recruitment, specifically targeting liberal arts and non-STEM graduates. Same company. Same CEO. Two audiences. Its report sells consulting engagements to enterprises panicking about AI readiness. Its hiring announcement attracts talent to Cognizant itself.
What the Hiring Data Actually Shows
When you shift from surveys about what companies think will happen to data about what companies are doing, the fear narrative weakens.
IBM announced plans to triple U.S. entry-level hiring in 2026. At a Charter AI Summit in New York, IBM Chief Human Resources Officer Nickle LaMoreaux explained the logic: "If we don't continue to invest in entry-level hires, what happens in 3 to 5 years? There's no pipeline; the well simply dries up." IBM is hiring across all business units, from developers to quantum data scientists to social media marketers.
McKinsey is planning a 12% increase in North American headcount for 2026 and now uses a gamified assessment tool called Solve to screen applicants on critical thinking and systems thinking rather than prior business knowledge. Heather Stefanski, McKinsey's Chief Learning and Development Officer, described the approach: "We are doubling down on what makes you uniquely human, and inserting more tech."
Citadel Securities reports that software engineering job postings are up 11% year over year. The Yale Budget Lab recently concluded that concerns about AI displacing today's workforce "remain largely speculative." The Brookings Institution found in 2025 that, in general, AI adoption has led to employment growth and firm growth, not widespread job loss.
| Source | Finding | Business Model |
|---|---|---|
| IDC (2025) | 66% of enterprises cutting entry-level roles | Sells research to AI vendors & buyers |
| Gartner (Feb 2026) | 55% expect agentic AI to reduce junior hiring | Sells research to AI vendors & buyers |
| ISE UK (Oct 2025) | Tech grad roles down 46%, projected 53% more | Education sector advocacy |
| Cognizant report (Jan 2026) | 93% of jobs affected, $4.5T in automatable tasks | Sells AI consulting services |
| Teneo CEO survey (2026) | 67% of CEOs say AI increases entry-level hiring | Advisory firm, no AI product sales |
| IBM (2026) | Tripling U.S. entry-level hiring | Major employer (300K+ workers) |
| McKinsey (2026) | +12% North American hiring planned | Major employer (45K+ workers) |
| Yale Budget Lab (2025) | AI displacement concerns "largely speculative" | Academic research, no commercial interest |
The pattern is not subtle. Organizations whose revenue depends on enterprises buying AI solutions or AI-readiness consulting consistently produce research showing that AI is destroying entry-level work. Organizations that actually run large workforces are expanding junior hiring. Academic institutions with no commercial stake in either outcome call the evidence inconclusive.
But the Stanford Data Is Real
Before this becomes a story about manufactured panic, one dataset deserves serious engagement. Stanford's Digital Economy Lab published "Canaries in the Coal Mine" using payroll data from one of the largest payroll software providers. They found a 13% relative decline in employment for early-career workers (ages 22 to 25) in occupations with high generative AI exposure. This is not a survey about intentions. It is payroll data measuring actual employment.
This finding is important. It is also narrower than the headlines suggested. That 13% decline is relative, comparing high-AI-exposure occupations to low-AI-exposure ones. It does not mean 13% of young workers lost their jobs. It means that in occupations where generative AI can do more of the work, young workers' share of employment dropped by 13 percentage points more than in occupations where AI has less impact. Brynjolfsson and co-authors themselves titled the paper as a question, not a declaration, and published a follow-up examining whether interest rates and post-pandemic hiring corrections explained part of the effect.
Absolute numbers tell a calmer story. As of September 2025, the U.S. unemployment rate for recent college graduates (Bachelor's, ages 20 to 24) stood at 9.5%, compared to 4.3% for the overall labor force. That gap is real and painful for graduates living it. It is also not historically unprecedented. Youth unemployment ran higher during the 2008 financial crisis recovery and was elevated through 2014. As Forbes contributor Hessie Jones noted that the current entry-level hiring slump started with post-pandemic economic shifts and monetary tightening, not AI adoption timelines.
The Cognizant Math, Deconstructed
Cognizant's report claims AI can now handle $4.5 trillion in U.S. work tasks. That number sounds apocalyptic until you read the methodology. It is based on reassessing 18,000 tasks across 1,000 jobs in the O*NET labor database, scoring each task's "exposure" to AI assistance or automation.
"Exposure" does heavy lifting as a word. An AI exposure score of 39% does not mean 39% of work will be automated. It means 39% of work tasks could be assisted or automated by AI at current capability levels. Between "could" and "will" is where entire industries live. Legal work exposure jumped from 9% to 63%, but U.S. law firm headcount has grown every year since 2020, including 2025. "Exposure" measures technical capability. Hiring data measures what humans actually decide to do.
Consider the inputs. Cognizant's average exposure score is 39% today, versus their 2024 forecast of 27% for 2032. That is a dramatic acceleration. But the report also notes that AI "is unable to automate upwards of 40% of management, business/financial operations, and administrative tasks." In other words, even by their own most aggressive modeling, AI cannot do most of the work that actually runs companies. That $4.5 trillion figure is a ceiling, not a forecast.
The Strongest Case Against This Analysis
An obvious counterargument is that IBM and McKinsey are outliers. Big consultancies and tech giants can afford to hire speculatively. Real damage is happening at mid-market companies nobody writes press releases about. A small accounting firm that replaces three junior associates with AI tax software does not hold a press conference. It just does not post the job listing.
This is fair. ISE data from the UK is granular and real: tech graduate hiring fell 46% in a single year, and the Institute surveyed actual employers, not CEOs at Davos. Stephen Isherwood, ISE's joint chief executive, told the Financial Times that AI was "already displacing young professionals." Mid-market displacement could be widespread and invisible in aggregate CEO surveys.
There is also a timing problem. IBM's tripling of entry-level hiring is a plan, not a headcount report. McKinsey's 12% increase is a projection. Plans change. If the economy softens, junior hires are the first to get cut. The fear narrative could be early and right while the hiring narrative is late and wrong.
Limitations
This analysis relies on publicly announced hiring plans and published surveys. IBM has not disclosed actual entry-level hiring numbers for 2026 yet, only plans. The Teneo survey sampled 350+ CEOs, a meaningful but not enormous sample, and CEO optimism about their own hiring is a known bias. The Stanford payroll data ends in mid-2025 and may not capture the most recent quarter's trends. U.S.-centric data dominates; the UK ISE data suggests the pattern may differ across labor markets with different employment protections. The IDC "66%" figure conflates intent (planning to reduce) with action (having reduced), and the survey methodology is behind a paywall.
What You Can Do
If you are a recent graduate or about to be one: Look at what companies are doing, not what research firms are publishing. IBM, McKinsey, and Cognizant itself are all expanding junior hiring. Roles are changing: systems thinking, AI-output validation, and client-facing judgment are replacing routine coding and data entry. McKinsey now screens for critical thinking via gamified assessments, not prior business knowledge. Learn to work alongside AI tools rather than competing with them on tasks AI does better.
If you are a hiring manager: IBM's CHRO made the strategic case plainly: skip entry-level hiring now and you have no pipeline in three to five years. Companies cutting junior roles are optimizing for this quarter's margin. Companies expanding junior hiring are investing in the workforce they will need when today's senior staff retire.
If you are reading a headline about AI killing jobs: Check who funded the research. If the publisher's revenue comes from selling AI tools, AI consulting, or AI-readiness assessments to enterprises, the finding that "AI is transforming work and you need to act now" is not wrong, exactly. It can be legitimate research and a sales tool at the same time. Knowing who paid for it tells you which function is primary.
The Bottom Line
The entry-level job crisis is real for the people living it. A 9.5% unemployment rate for recent graduates means roughly one in ten people with new bachelor's degrees cannot find work. That is not speculative. But the narrative that AI is the primary driver, and that the displacement is accelerating beyond control, tracks almost perfectly to organizations with a commercial interest in that conclusion. The companies actually writing paychecks are telling a different story. In economics, there is a name for this: revealed preferences. What people do is more reliable than what people say. Right now, CEOs are saying one thing to conference audiences and doing another with their hiring budgets. For once, what they are doing is less alarming than what they are saying.