Research Brief · AI Strategy

If AI Makes Employees More Valuable, Why Are Companies Laying Them Off?

What the reconciliation implies

The counter-evidence

The companion piece to this brief argues that AI raises the quality and judgment-intensity of work without shrinking hours — a "perception vs. reality" story where the perceived threat (AI shrinks my job) doesn't match the measured reality (AI expands what one person can do). That argument has to survive an obvious objection: if that's true, why is 2026 producing the largest wave of AI-attributed layoffs on record?

The layoff data is real, well-documented, and large enough that it can't be waved away.

The scale. Outplacement firm Challenger, Gray & Christmas found that AI has been the leading stated reason for U.S. job cuts for four consecutive months in 2026 — a streak with no precedent in the firm's data, which has tracked AI as a distinct cut reason since 2023.^[1] Tech employers announced 139,156 job cuts in the first half of 2026, an 83% surge over the same period in 2025, and AI was explicitly cited in 101,743 of all U.S. layoff announcements economy-wide — roughly 23% of every job cut Challenger tracked.^[1]

The company list. Snap cut about 1,000 employees (16% of its workforce) with CEO Evan Spiegel citing AI advancements in an SEC filing.^[3] Amazon cut 16,000 corporate jobs following 14,000 cuts in October 2025, with CEO Andy Jassy having told staff that "as we roll out more generative AI and agents... we will need fewer people doing some of the jobs that are being done today."^[3] Block's Jack Dorsey cut headcount nearly in half — from 10,000 to fewer than 6,000 — attributing the move directly to AI.^[10] Accenture cut roughly 11,000 roles, with CEO Julie Sweet stating that employees who "cannot [be] reskill[ed] will be exited."^[10]

The tell: several of these companies were not struggling. Cloudflare cut 20% of its workforce (1,100 people) the same period it reported quarterly revenue up 34% year-over-year — its highest single quarter ever.^[3] Cisco cut 4,000 jobs the day it reported record quarterly revenue of $15.8 billion, with its CFO stating the move was "not savings-driven."^[3] Meta cut roughly 8,000 roles in May 2026 while reporting Q1 revenue up a third year-over-year and guiding 2026 AI infrastructure capex to $115–145 billion.^[2]

That pattern — cutting staff at the moment of strongest financial performance, explicitly to fund AI infrastructure spend — is the detail that starts to separate two different stories that are currently being told as one.

The measurement that resolves the contradiction

The single most important piece of evidence here is not a headline count of layoffs — it's a study that checked those layoffs against actual return on investment.

Gartner surveyed 350 global business executives at companies with at least $1 billion in annual revenue, all of them already piloting or deploying AI agents, intelligent automation, or autonomous technologies.^[6] About 80% reported workforce reductions tied to their AI initiatives. But those reductions showed no correlation with improved ROI — workforce-reduction rates were nearly identical between companies reporting strong returns and companies reporting modest or negative returns from their AI investments.^[6,7] Gartner's lead analyst, Helen Poitevin, put it directly: "Workforce reductions may create budget room, but they do not create return. Organizations that improve ROI are not those that eliminate the need for people, but those that amplify them."^[6]

This is the fact that reconciles the two paradigms. The productivity research (Jagged Frontier, Cybernetic Teammate, Generative AI at Work — see the companion brief) measures what happens inside a task when a person uses AI to do it. The Gartner data measures what happens when a company's leadership decides to cut headcount and attributes the decision to AI. These are not the same event, and conflating them is exactly what's producing the apparent contradiction. Gartner's data says plainly: the companies eliminating the most people are not the companies generating the most productivity or profit.^[5]

Two genuinely different phenomena hiding under one headline

1. A real, narrow substitution effect at the entry level

Some of the labor-market shift is a legitimate, direct consequence of the same mechanism the productivity studies document — just applied to hiring rather than to an existing employee's daily task list. Stanford HAI's 2026 AI Index found that employment for workers ages 22–25 in AI-exposed occupations has declined roughly 13–20% since late 2022, while employment for older workers in the same occupations held steady or grew.^[2,9] The mechanism: AI code-generation and testing tools let senior staff absorb the boilerplate work that junior roles existed to do, so companies quietly stop hiring juniors rather than firing incumbents. This is consistent with — not contradictory to — the "one person with AI matches a two-person team" finding from the Cybernetic Teammate study; it's the same acumen-concentration effect showing up in a hiring line instead of a task log.

It's also consistent with the "jagged frontier" concept itself: within any workflow, some tasks fall inside AI's reliable capability zone and get automated outright (entry-level drafting, routine customer support, first-pass code generation), while others fall outside it and still require a human's judgment. Layoffs concentrated in roles built almost entirely around inside-the-frontier tasks — Chegg cutting 45% of its workforce as students shifted to generative AI instead of homework-help platforms is a clean example — are a real, direct AI effect, not an accounting story.^[10]

2. A capital-allocation and narrative effect that borrows the AI label

The larger and more contested share of 2026's layoffs looks like something else: a convenient, market-legible explanation for decisions that were going to happen anyway. Multiple economists studying the same layoff data are explicit about this. Ben May of Oxford Economics told CBS News: "We suspect some firms are trying to dress up layoffs as a good news story rather than a bad one — for example, by pointing to technological change instead of past overhiring."^[1] Lisa Simon of Revelio Labs called AI "a little bit of a front and an excuse" for cuts companies wanted to make for other reasons.^[9] Nearly six in ten companies surveyed have admitted to framing layoffs or hiring slowdowns as AI-driven when the real reason was financial.^[9]

Even a company with a direct financial stake in the AI narrative pushed back on it: Nvidia CEO Jensen Huang called CEOs who blame AI for layoffs "lazy," saying it doesn't make business sense that companies are cutting people to the extent being reported while genuinely relying on AI to backfill the work.^[4]

There's also a structural reason this is hard to verify either way: the federal WARN Act requires 60 days' notice before mass layoffs but does not require a company to disclose why it's cutting staff, so Challenger's AI-attribution numbers rest entirely on voluntary, unverified statements in memos, SEC filings, and earnings calls.^[1] A pending bill, the AI Workforce PREPARE Act, would require companies to name the specific AI systems used and estimate what percentage of a cut is actually attributable to them — but it has not passed, and California's parallel SB 951 is stalled.^[1]

What the reconciliation implies

The comparison above is the whole argument in one glance. Gartner's own forecast underlines which mechanism it thinks is durable: the firm expects that by 2027, half of companies that attributed headcount reductions to AI will end up rehiring for similar functions under different job titles, and it forecasts autonomous business becoming a net job creator by 2028–2029 as new categories of AI-adjacent work emerge that current systems can't fully absorb on their own.^[9,8]

The reconciliation, stated plainly: the productivity research and the layoff headlines are not measuring the same claim, so they can't actually contradict each other. "AI raises quality-per-hour for the people using it" is well-evidenced. "Companies are firing large numbers of people and calling it AI" is also well-evidenced — but the same body of research that would need to be true for that to reflect real AI-driven redundancy (the productivity studies) is the evidence showing that headcount-first strategies are the ones failing to generate returns. The organizations actually capturing AI's economic value are, per Gartner, the ones amplifying their people — not the ones whose press release uses AI as the explanation for a smaller headcount.

References