The number that lit up the AI press this week is 99. Per Mercer’s Global Talent Trends 2026 report, 99% of the C-suite executives in a nearly-12,000-person global survey now expect AI to cause at least some headcount reduction inside their organizations within the next two years. The coverage went viral the week of May 24, 2026 — Tom’s Hardware, Gizmodo, TechSpot, and Inc. all ran some version of “CEOs admit out loud what employees already feared.”

That headline number is real. It’s also incomplete on its own.

The harder picture takes three data points side by side. Mercer’s survey says CEOs are restructuring around an AI productivity gain. Q1 2026 tech layoff data says the restructuring is already happening — 78,557 tech workers lost their jobs in Q1, the largest quarterly total in two years. And OpenAI’s B2B Signals report says the productivity gain those layoffs are supposed to fund is real, but it’s concentrated in a small slice of firms that use AI very differently than the median company.

Three numbers. One uncomfortable conclusion. Most CEOs are cutting before the ROI lands.

Quick Summary: The Displacement-ROI Paradox at a Glance

Detail	Info
Headline survey	Mercer Global Talent Trends 2026
Sample size	~12,000 executives, HR leaders, investors, employees
C-suite expecting AI layoffs in 2 years	99%
Q1 2026 tech layoffs	78,557 workers (source)
Q1 cuts attributed to AI/automation	47.9% (Nikkei Asia)
March alone	Challenger, Gray & Christmas tagged 15,341 cuts as AI-driven
Employee thriving sentiment	66% in 2024 → 44% in 2026 (Mercer)
Frontier firms vs. typical firms	3.5x more AI intelligence per worker, 16x more Codex usage (OpenAI B2B Signals)
AI initiatives meeting ROI expectations	~28% per Gartner (Tech Startups); 25% per IBM CEO study

Bottom line: The labor restructuring is moving faster than the productivity proof. For AI pros, this is the defining tension of 2026 — the people most able to use AI well are also the people most exposed when the cuts get sloppy.

---

What the Survey Actually Says

Read past the 99% headline and Mercer’s Global Talent Trends 2026 report is more interesting than the press cycle made it look.

The 99% figure covers C-suite leaders who expect “at least some” headcount reduction over the next two years. That phrasing matters. A leader planning to cut two roles and a leader planning to cut two thousand both count inside the same number. The 99% is the floor of restructuring expectation, not the ceiling of mass layoffs.

What sits underneath is closer to the real story. Per Mercer, 65% of executives surveyed expect to redeploy or reskill between 11% and 30% of staff. That’s a restructuring plan, not a slash-and-burn one. The same report shows 63% see AI-enabled work redesign as the strongest ROI driver — but only 32% believe their current workforce can deliver that redesign. There’s the gap. The strategy is “rebuild the workforce around AI.” The honest assessment is “we don’t think the current workforce can do it.”

The employee data is the part that doesn’t lie. Mercer’s thriving-at-work metric collapsed from 66% in 2024 to 44% in 2026. Employee concern about AI-driven job loss climbed from 28% to 40% in the same window. Per the IBTimes UK coverage, the anxiety story is global and consistent across markets. People can read a restructuring memo.

So the C-suite is restructuring confidently. The workforce knows what restructuring means. And nobody in the survey has clean data on whether the restructuring will actually pay off.

What Already Happened in Q1

Mercer’s survey is a forecast. The Q1 2026 tech layoff data is a record.

Per Metaintro’s analysis of Nikkei Asia data, 78,557 tech workers lost their jobs in Q1 2026, with 76.7% of cuts happening in the US and 47.9% attributed directly to AI and automation. That’s the largest quarterly tech layoff total in two years. Oracle, Amazon, Microsoft, and Block led the wave — Oracle alone shed over 25,000 positions.

The AI-attribution number depends on who you ask. Challenger, Gray & Christmas tagged 15,341 of March’s total job cuts as AI-driven — about 25% of the month’s count. Nikkei’s 47.9% sits at the high end. The methodology gap is real: “AI-driven” gets defined differently depending on whether you count restructuring-around-AI as AI-attributed or only direct replacement.

Three things hold regardless of which methodology you trust.

The cuts are concentrated in tech. Per Andreas Timm’s analysis, aggregate Q1 layoffs hit a four-year low — except in technology, where the share went up 40%. The recession is sector-specific. It’s sitting on the industry that builds the AI.

The cuts are concentrated in mid-career and junior roles. Per Tom’s Hardware, the companies cutting hardest are explicitly racing to replace junior workers with AI tooling. That’s the workforce category Mercer flagged with the largest anxiety jump. The data on the ground matches the survey forecast.

The cuts are happening at companies that still have an AI ROI question. Oracle, Amazon, and Microsoft aren’t under existential financial pressure. The decisions aren’t “cut to survive.” They’re “cut to free up capital for AI infrastructure spend that hasn’t paid off yet.” Tech Journal estimated $725B in AI spending in 2026 — much of it funded by labor reduction in the same companies doing the spending.

That’s the part Mercer’s survey makes legible. The restructuring is partly a workforce thesis and partly a budget transfer.

The ROI Gap Nobody on the Earnings Call Talks About

The Mercer data and the layoff data only become coherent when you read them next to the AI ROI data.

Per Tech Startups’ coverage of Gartner data, about 28% of AI initiatives meet ROI expectations, with one in five outright failures. A separate Gartner study covered by Fortune found that workforce reductions to fund AI have not consistently translated into better ROI. IBM’s parallel CEO study puts the figure even lower — 25% of initiatives delivering expected returns, with 56% of CEOs reporting zero significant financial benefit so far.

Now stack the data points:

• 99% of CEOs plan to reduce headcount around AI within two years.
• Only ~28% of AI initiatives actually deliver expected ROI today.
• 78,557 tech workers already lost their jobs in Q1 against this ROI backdrop.

The labor restructuring is running roughly three times faster than the productivity proof.

There’s a name for that pattern. It’s called pricing in expected gains before they arrive. Public companies do it all the time during cost-cutting cycles — announce the productivity story to the market, take the labor expense down today, hope the productivity catches up before the next earnings call. The risk is that the productivity gain shows up smaller than the labor cut. The remaining workforce gets stuck doing more work with worse tooling than the CEO promised. Service quality dips. Customer churn ticks up. The cost savings get absorbed by revenue erosion within four quarters.

That’s the scenario the Mercer report flags most directly — and the part of the survey that almost none of the viral coverage focused on. CEOs who restructure aggressively without rebuilding the human-machine capability inside the firm tend to underperform peers who pace the restructuring against demonstrated capability.

What OpenAI’s B2B Signals Tells Us About Who Wins

Here’s where the third data point earns its place.

Per OpenAI’s B2B Signals report published earlier this month, frontier firms — the 95th-percentile users of AI inside their industries — use 3.5x as much AI intelligence per worker as typical firms. That gap was 2x a year ago. It’s widening, not closing.

The gap widens hardest in agentic tooling. Per Quantum Zeitgeist’s coverage of the report, frontier firms send 16x as many Codex messages per worker as typical firms. And the message-count gap only explains about 36% of the actual productivity gap. The rest of the gap comes from richer, more complex AI use — each interaction at a frontier firm is doing more of the actual work. Typical firms are using AI to answer questions. Frontier firms are using AI to execute multi-step work.

Lay that next to the Mercer survey and the picture sharpens.

The CEOs cutting hardest are cutting in anticipation of becoming a frontier firm. The ones who actually become frontier firms will see the productivity gain Mercer’s restructuring plan implies. The ones who don’t will end up with a smaller workforce, a typical-firm AI utilization rate, and an ROI report card their board won’t enjoy.

The dividing line between those two outcomes is not budget. It’s not vendor selection. It’s not even model choice. Per the B2B Signals data, it’s how the remaining workforce uses the tooling. Frontier firms invest in skills and new roles so humans guide AI through complex work. Typical firms ship the license and hope someone figures it out.

Which is exactly what Mercer’s 32%-of-workforce-can-deliver number was already telling us.

What AI Pros Should Actually Do This Quarter

Five concrete moves, in rough order of how much each pays back if you’re a working AI professional inside an organization doing this restructuring.

1. Become the person who can show the frontier-firm gap on a slide. The OpenAI B2B Signals report is doing more to reframe internal AI strategy conversations than any vendor pitch has all year. If you can present the 3.5x and 16x numbers to your leadership with a credible “here’s where our usage actually sits” overlay, you become the person whose recommendations get listened to.
2. Pick the workloads where complex AI use is verifiable. The frontier-firm advantage shows up in long, multi-step work — research-to-draft pipelines, Codex-style coding work, customer triage that resolves rather than routes, document analysis that produces a recommendation rather than a summary. These are the workloads where a unit of AI investment can be measured against a unit of labor saved. Generic “save time on emails” pilots will not survive a Q3 ROI review.
3. Push your org to invest in skills, not seats. Per the Mercer data, 63% of executives think AI-enabled work redesign is the ROI driver while only 32% think their workforce can deliver it. That 31-point gap is a budget line waiting to be claimed. Training, internal certifications, capability assessments — anything that closes that gap is the easiest pitch to make to a CFO this year.
4. Document your own productivity gain before someone else does it for you. Anyone working alongside AI tooling should have a concrete artifact showing what shifted — engagement length, error rate, throughput, escalation count. The cuts coming in Q2 and Q3 will preferentially hit people whose contribution is illegible to a manager running an AI-productivity review. Illegibility is a budget hazard in 2026.
5. Watch the enterprise AI distribution story, not just the model story. The vendors winning enterprise mind-share — Anthropic with KPMG, Microsoft with Agent 365, Google with Workspace-native Spark — are pricing and packaging for the productivity-gain narrative that’s driving the layoff cycle. The tooling your org buys next quarter will be selected against that narrative. Knowing the narrative gives you a seat in the procurement conversation.

The teams getting through this cycle in the best shape are not the ones with the most AI access. They are the ones whose AI access is the most measurably productive. That’s the part Mercer’s survey makes harder to see at first read.

Our Take

Three things stand out from the data this week.

First, the 99% number is being read wrong by most of the press. It’s not “99% of CEOs plan mass layoffs.” It’s “99% of CEOs expect at least some headcount reduction over two years.” Those are very different forecasts. The actual layoff intensity will sort along the lines Mercer flagged elsewhere in the report — how much each firm can realistically redesign work around AI capabilities its workforce can execute. The firms that try to redesign past their capability ceiling will overshoot the cuts and pay for it in the quarters that follow.

Second, the Q1 layoff data is the part the survey hasn’t fully caught up to. 78,557 tech workers in three months, with tech leading every other sector, is the leading indicator of what the next two years look like if the restructuring continues to outrun the ROI. That number going up in Q2 is the signal that the displacement-first thesis is in trouble. That number stabilizing or declining while productivity metrics improve is the signal that the restructuring is working as intended. Right now, the betting line is open.

Third, the OpenAI B2B Signals report is the most strategically useful document published in May. Most coverage of it treated it as marketing material — “look how much AI our frontier customers use.” Read against the Mercer survey, it’s something different. It’s an honest map of the workforce capability gap that will determine which firms get the productivity gain they’re pricing in and which firms end up with a thinned workforce and stalled ROI. The data on Codex usage in particular (16x more messages per worker at frontier firms) tells you precisely which capabilities the winning firms are developing inside their teams — and which the typical firms are still pretending will appear when the license arrives.

For AI professionals reading this in late May 2026, the practical work is straightforward. The labor restructuring is already underway. The ROI proof is not yet evenly distributed. The dividing line between the firms that win this cycle and the firms that lose it sits inside the workforce’s ability to use AI for complex, multi-step work — not inside the budget for licenses. If you can measure your own contribution against that line, you’re useful to the side that wins. If you can help your organization measure its position relative to the frontier-firm gap, you’re essential to it.

The number to track for the next 90 days is not the 99%. It’s the Q2 tech layoff count. If it cools materially below Q1’s 78,557, the restructuring is pacing against measured ROI. If it climbs, the displacement-first thesis is becoming the dominant operating model and the AI labor market enters a different phase.

We’ll know by August.

Frequently Asked Questions

Did 99% of CEOs really say AI will cause layoffs?

Per Mercer’s Global Talent Trends 2026 report, 99% of C-suite executives surveyed expect AI to cause “at least some” headcount reduction within two years. The 99% is the floor of restructuring expectation, not a forecast of mass layoffs at every firm. Underneath that figure, 65% of executives expect to redeploy or reskill between 11% and 30% of staff — restructuring more than wholesale cuts. The viral framing has flattened a more nuanced result.

How many tech workers lost jobs in Q1 2026?

Per analysis of Nikkei Asia data, 78,557 tech workers were laid off in Q1 2026, with 76.7% of cuts in the US and 47.9% attributed to AI and automation. Oracle, Amazon, Microsoft, and Block led the wave. Challenger, Gray & Christmas tagged 15,341 of March’s cuts alone as AI-driven, about 25% of that month’s total. The AI-attribution percentage varies by methodology — Nikkei sits at the high end, Challenger at the lower end.

What is the AI ROI gap and why does it matter?

Per Tech Startups’ coverage of Gartner data, only about 28% of enterprise AI initiatives meet ROI expectations. IBM’s parallel CEO study puts the figure at 25%, with 56% of CEOs reporting zero significant financial benefit. The gap matters because 99% of CEOs are restructuring labor around an AI productivity gain that, on current data, only roughly one in four AI initiatives actually delivers. The labor cuts are pricing in returns that haven’t fully arrived.

What is the OpenAI B2B Signals report?

Per OpenAI’s announcement, B2B Signals is the company’s published analysis of how different enterprise customers actually use AI. The headline finding: frontier firms — the 95th-percentile users in their industries — consume 3.5x as much AI intelligence per worker as typical firms, up from 2x a year ago. The gap is widest in agentic tooling. Per Quantum Zeitgeist, frontier firms send 16x as many Codex messages per worker as typical firms. The depth of AI use, not just the volume, separates the two groups.

Is AI actually replacing jobs or just shifting them?

Both, depending on the firm. Per Mercer’s data, 65% of executives plan to redeploy or reskill 11-30% of staff — that’s job shifting at scale. The Q1 layoff data shows roughly half of tech sector cuts attributed to AI/automation are pure replacement, with the other half tied to broader restructuring. The mix varies by role: junior knowledge work is being replaced fastest, senior work is being augmented, and net-new AI-native roles are being created at the top end. The shape of the labor market in 2027 is going to look meaningfully different from 2025.

Are CEOs cutting too fast?

The data suggests some are. Per Gartner findings via Fortune, about 80% of firms deploying AI have cut staff, but those cuts have not consistently translated into better ROI. Firms that gained the most invested in skills and new roles to guide AI rather than just shrinking headcount. The Mercer survey echoes the same finding — the 31-point gap between “AI is the ROI driver” (63%) and “our workforce can deliver it” (32%) is the early warning signal that some restructuring is running ahead of the underlying capability.

What should AI professionals do given this data?

Three practical moves: document your own AI-augmented productivity in measurable terms so your contribution is legible to managers running productivity reviews; push your org to invest in skills and capability-building rather than just licenses, since the frontier-firm gap is a workforce capability story not a budget story; and watch the enterprise AI distribution story — vendors like Anthropic, Microsoft, and Google are building the tooling the productivity-gain narrative depends on, and knowing that narrative buys you a voice in your org’s procurement conversations.

Where can I read the actual reports?

Primary sources: Mercer Global Talent Trends 2026, OpenAI B2B Signals, and Challenger, Gray & Christmas’ monthly layoff reports. For aggregated Q1 2026 layoff analysis, Metaintro’s breakdown and Andreas Timm’s sector analysis are the most useful secondary reads.

---

Last updated: May 27, 2026. Sources: Mercer Global Talent Trends 2026 report · Mercer talent insights hub · Tom’s Hardware · Gizmodo · TechSpot · Inc. · IBTimes UK · Metaintro Q1 2026 layoff analysis · Challenger, Gray & Christmas March report · Andreas Timm Q1 2026 analysis · OpenAI B2B Signals · Quantum Zeitgeist Codex analysis · Fortune AI ROI coverage · Tech Startups Gartner coverage · Tech Journal AI spending analysis · IBM IBV CEO Study.

Related reading: KPMG Deploys Claude to 276K Staff · Microsoft Agent 365 GA: What Enterprise Buyers Need to Know · Anthropic Overtakes OpenAI on Business Ramp Index · Gemini Spark Is Live · OpenAI Files for IPO · Anthropic Tops OpenAI’s Valuation at $900B