SHRM brought a number to its annual conference in Orlando that runs against the headlines. Senior labor economist Justin Ladner, working from a survey of more than 14,000 U.S. workers, put the share of jobs facing high near-term automation displacement risk at about 6%, roughly 9.2 million positions. The other 94% are being reshaped, not erased. Most workers report that automation and AI now handle 10% to 15% of their tasks, and about 60% of jobs carry at least one barrier that blocks full automation: a client who wants a human, a regulation, a cost that does not pencil out. The story this week is a measured counterweight to the displacement panic, and a reminder that “transformed” still means every role looks different on the other side.

Subscribe now

By the Numbers

• About 6% of U.S. jobs, roughly 9.2 million, face high near-term automation displacement risk (SHRM).

• Around 60% of U.S. jobs carry at least one nontechnical barrier to full automation, including client preference for human interaction, regulation, organizational constraints, and cost (SHRM).

• The most common level of automation and AI use reported by workers is 10% to 15% of job tasks (SHRM).

• Survey base: more than 14,000 U.S. workers (SHRM).

Reskilling and Education

SHRM puts a ceiling on the displacement story: 6% of jobs at high risk, the rest transformed

At SHRM26 in Orlando, senior labor economist Justin Ladner presented research drawn from more than 14,000 U.S. workers and landed on a figure smaller than the layoff coverage suggests. Roughly 6% of jobs, about 9.2 million, sit in the high near-term displacement zone, meaning they are both heavily automatable and short on the nontechnical features that keep humans in the loop. The remaining 94% have at least one of those features. Ladner found that about 60% of jobs include a barrier such as client preference for a person, legal or regulatory requirements, organizational constraints, or the plain cost of automating. Exposure is uneven. Technical white-collar roles, including web developers, software developers, computer programmers, and financial analysts, show the highest automation and AI adoption, while most workers report AI touching only 10% to 15% of their daily tasks. Ladner’s framing was deliberate: automation and the current AI wave have, so far, done more to transform work than to remove it.

Source: SHRM, June 24, 2026, “Why the AI Job Displacement News May Not Be What HR Leaders Think.” Read it here.

Why it matters: A 6% high-risk figure does not let leaders relax, because “transformed” reaches the other 94% and rewrites the task content of nearly every role. Plan for redesign at scale, not a small wave of cuts. The roles with the highest AI adoption are technical and well-paid, so the urgent work is reskilling people inside jobs that survive, not just protecting the few on the displacement edge.

What Workforce Leaders Are Watching

• If only 6% of roles face high displacement risk but 94% are changing, is your workforce plan built around a handful of eliminations or around redesigning the task content of nearly every job?

• Ladner’s barriers to automation, client preference, regulation, cost, are business choices as much as technical limits. Which of your roles stay human by design, and which are protected only until the cost of automating them drops?

• This claim notes the most AI-exposed roles are technical and senior, not the entry-level ones the most displacement narratives fixate on. Does your reskilling budget follow where adoption is actually highest?

• When 10% to 15% of tasks shift to AI across most of the workforce, the productivity gain is diffuse and easy to miss. How are you measuring what people do with the freed time, and who owns redirecting it to higher-value work?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

Satya Nadella spent the week telling his own industry that the jobs story it has been selling will not hold. In a Wall Street Journal interview, the Microsoft CEO named OpenAI and Anthropic and argued that an AI economy built on a handful of models capturing all the value, while executives forecast the end of white-collar work, has no political future. His phrase was “societal permission,” and he said the industry has not earned it. Separately, Kyndryl put a number on the gap Nadella is describing from the employer side: across 1,100 leaders, the share who say their workforce is ready for AI fell to 23%, down six points in a year, even as adoption accelerated. One executive is telling AI companies to slow their messaging. The data says companies are deploying faster than their own people can absorb.

Subscribe now

By the Numbers

• 23% of organizations say their workforce is fully ready for AI, a six-point drop from a year earlier (Kyndryl).

• 79% of leaders agree AI will move faster than their workforce, governance, and operating models can keep up (Kyndryl).

• 81% expect AI agents to make impactful decisions for their organization within a year; only 25% fully trust AI to operate without human oversight (Kyndryl).

• 61% of organizations have already redesigned roles around AI; 24% are creating new roles dedicated to managing it (Kyndryl).

• Agentic coding consumes roughly 1,000 times the tokens of a standard code-chat interaction, with subsidized cost near $2,000 per engineer per month (Microsoft Research, cited by Nadella).

Policy and Government

Nadella tells the AI industry it has no “societal permission” to hollow out jobs

Microsoft’s CEO used a Wall Street Journal interview to break with the displacement messaging coming from his closest AI partners. Nadella named OpenAI and Anthropic and argued that an AI future where value accrues to “a few models that eat everything they see” will not survive political scrutiny. His sharpest line was a limit on the industry’s own pitch: “You can’t say, hey, all white-collar jobs are gone and this could even be a weapon and we will use all the power to build data centers.” He drew the parallel to globalization, where GDP held up on paper while industrial communities were hollowed out and the displacement was real. The framing matters because it comes from the company selling the tools. Nadella is telling executives that forecasting mass job loss while demanding resources and regulatory room is a strategy that voters and governments will eventually reject, and that AI companies have to show the technology reorganizes work rather than only erasing it. He also pointed at the economics underneath the hype: Microsoft Research found agentic coding burns roughly 1,000 times the tokens of ordinary code chat, and the costs are real enough that the “replace everyone” math does not yet close.

Source: The Wall Street Journal, June 21, 2026. Read it here.

Why it matters: When the vendor with the most to gain says the job-elimination narrative is politically unsustainable, that reframes how leaders should talk about their own AI plans. Naming workers as line items to be removed invites the backlash Nadella is warning about. Frame AI decisions around the work people will do next, and be ready to show it.

Reskilling and Education

Workforce readiness for AI fell while deployment sped up

Kyndryl’s second annual People Readiness Report surveyed 1,100 senior business and technology leaders across eight countries and found readiness moving the wrong direction. Only 23% say their workforce is fully ready for AI, down six points from last year, in a period when adoption climbed. The leaders are candid about the mismatch: 79% agree AI will outpace their workforce, governance, and operating models, and 81% expect AI agents to make consequential decisions within a year while just 25% fully trust those systems to run without human oversight. The report identifies a “Pacesetters” group, about 9% of organizations, that redesign roles around AI, run real change management, and put guardrails in place before scaling. The rest are buying capability faster than they are preparing the people who have to use it. On the action side, 61% report redesigning roles already and 24% are creating jobs focused on managing AI, but a third still lack clear policy on which decisions AI is allowed to make.

Source: Kyndryl 2026 People Readiness Report, June 25, 2026. Read it here.

Why it matters: Readiness dropping while adoption rises means the burden is landing on workers who were not prepared for it. The 9% who redesign roles and set guardrails first are the control group worth copying. Treat workforce readiness as the gate on deployment speed, not a cleanup task that follows it.

What Workforce Leaders Are Watching

• If a vendor CEO is publicly warning that job-elimination messaging has no “societal permission,” how does your own internal language about AI and headcount read against that standard?

• Your AI deployment timeline and your workforce readiness curve are diverging. Which one are you willing to slow, and who in the room has the authority to make that call?

• Kyndryl’s Pacesetters redesign roles and set decision guardrails before scaling. What would it take to move your organization into that 9% this year rather than next?

• You expect AI agents to make impactful decisions within a year but mostly do not trust them unsupervised. Where, specifically, is the line between AI and human authority written down in your organization?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

TL;DR: Amazon, Google, and a growing list of Big Tech companies are widening spans of control and stripping out management layers, often citing AI. The first-order effect everyone names is manager overload. The second and third-order effects are the ones that decide whether this works: the manager’s old work doesn’t disappear when you raise the ratio, it gets redistributed, and the development pipeline thins out three years before anyone notices. Flatter can be the right move. It only pays off if you redesign what a manager does before you double the number of people reporting to them.

Subscribe now

Andy Jassy told Amazon in September 2024 to increase the ratio of individual contributors to managers by at least 15 percent by the end of the first quarter of 2025. The company hit the target without mass layoffs: it paused manager hiring, asked existing managers to take more direct reports, and moved some managers back into individual contributor roles. Jassy’s framing was about speed, not cost. He wants Amazon to “operate like the world’s largest startup,” fast and flat, with fewer layers between a decision and the person making it. He said the quiet part out loud: at Amazon, the path to promotion is no longer taking charge of a massive team.

Google did a sharper version of the same thing. The company cut roughly 35 percent of its small-team manager roles, with the heaviest impact on managers responsible for fewer than three people. Many of them moved back to individual contributor work. Sundar Pichai’s line to staff was a sentence every workforce planner should sit with: “We need to be more efficient as we grow, so we don’t just throw more people at every problem.”

Gartner projected that through 2026, one in five organizations would use AI to flatten their structure, eliminating more than half of current middle management positions. The direction is set. The interesting question is no longer whether companies will widen spans. It is whether they understand the knock-on effects of doing it.

The number the spreadsheet loves and the research doesn’t

Span of control is one of the few org-design levers that looks like an easy target on a spreadsheet. Raise the average number of reports per manager from eight to twelve, and you need a third fewer managers to run the same headcount. The savings (and hopefully productivity gains from repurposing managers to individual contributors) are immediate and tangible. That is exactly why this metric gets implemented first.

The research on what happens next is less convenient. Gallup found that manager engagement peaks at around eight to nine direct reports and declines past that point. Teams led by managers with more than ten reports show measurably lower engagement than teams of five to eight, and the quality of one-on-one relationships drops as the number climbs. Meanwhile the average manager already oversees 12.1 people, up nearly 50 percent since 2013. So the starting point for most “let’s widen spans” conversations is already past the line where the research says relationships fray.

I lead Tech Workforce Strategy and Architecture at Capital One, where I spend a lot of my time on exactly this question: what is the right mix and shape of teams inside an engineering organization. The temptation is to treat span of control as a single dial. It is not. A span of twelve junior engineers shipping similar work is a different animal from a span of twelve senior engineers each owning a distinct system. The first is supervisable. The second is a portfolio of twelve relationships, twelve career arcs, twelve sets of context, run by one person who also has a day job. The dial and the work behind it are not the same thing.

Subscribe now

Where the manager’s work actually goes

Here is the part the ratio math skips. When you remove a layer of managers, their work does not evaporate. It moves. The only question is where, and most companies never make an active decision about it.

A middle manager does a bundle of jobs that rarely appear in a single job description. They triage priorities and absorb ambiguity so the team can keep moving. Coaching is in there. So is running the performance and promotion process, translating strategy down, surfacing reality up, and noticing the quiet problem before it becomes a loud one. Strip the layer out and each of those jobs reappears somewhere: pushed up to an already-stretched senior leader, pushed down onto individual contributors who now coordinate themselves, or simply dropped.

Dropped is the dangerous one, because it is invisible for a while. Coaching is the first thing to go when a manager’s span doubles, and the cost shows up much later, in a promotion case that never got built and a mid-level engineer who quietly stopped growing. Reporting on the flattening trend keeps landing on the same finding: when layers come out, junior employees lose the people closest to them in the hierarchy, the ones who used to provide feedback and access. The org chart gets cleaner. The development system gets thinner.

AI is genuinely part of the answer here, which is why the flattening story and the AI story are tangled together. A manager who once spent a third of the week assembling status reports, reconciling dashboards, and chasing updates can hand a real share of that to tooling. That redesign is what makes a wider span survivable, because it frees the manager’s hours for the work only a human can do. The courage to assign more people was never the hard part. The companies treating AI as a reason to cut managers are doing the easy half. The companies using AI to change what a manager spends time on are doing the half that actually makes twelve reports work.

The third-order effect loops back to the pipeline

Push the timeline out and the consequence stops being about any individual manager. It becomes structural. If coaching erodes and the rungs between individual contributor and senior leader get pulled out, you have quietly stopped manufacturing your future leaders. The leadership pipeline gap is a named risk of the flattening trend, and it is the same shape as the entry-level collapse I wrote about in The Inverted Pyramid. An organization that stops developing people in the middle today is writing a senior-talent shortage for itself in 2030. The senior engineering leader of 2031 is the person getting coached, or not, in the wider span you build this year.

Flatter organizations also change how careers work, and most have not updated the story they tell their people. When the title ladder gets shorter, “get promoted into management” stops being the default path for a strong individual contributor. That can be healthy. Forcing every excellent engineer into management to earn more was always a way to lose a great engineer and gain a mediocre manager. A flatter structure only works if there is a real, well-paid technical track that runs parallel to the management track, with its own levels and its own status. Without it, flattening just means fewer places to go and more people competing for them.

There is a harder question hiding underneath the technical track, and it is about how you attract and retain managers. If you widen a manager’s span, hand them more performance conversations and more career arcs to tend, and pull them further from the technical work that built their reputation and career, you have made the job bigger and less rewarding at the same time. A strong engineer can now earn well without ever managing anyone, which removes the old, lazy incentive that pushed people into management to grow their compensation. Strip that out and the question of who volunteers for the harder, less technical job gets real. Companies that want a real management pipeline will have to pay for it directly, pricing the management premium against an IC track that has become genuinely lucrative, and treating the willingness to develop other people as a scarce skill worth compensating, not a rung everyone climbs by default. The alternative is a slow shortage of people willing to manage, discovered the year you need them most.

Subscribe now

What it takes to make wide spans actually work

Span of control is downstream of work design. You cannot decide the right number of reports until you have decided what the manager is for. A few things separate the organizations that make this work from the ones that just announce a ratio:

• Decide where the removed work lands before you remove the layer. Name it. Coaching to a technical lead, status reporting to tooling, prioritization to a clearer operating rhythm. Unassigned work is dropped work.

• Match span to work type, not a company-wide average. A team of senior engineers on distinct systems needs a smaller span than a team doing convergent work. One number across an org is the tell that nobody modeled it.

• Protect coaching explicitly, because it is the first casualty and the last to show damage. If the wider span eats one-on-ones, the cost is a development hole you will feel in three years.

• Build the technical track before you shorten the management ladder. Flattening without a parallel path for individual contributors is a retention problem dressed as an efficiency win.

• Use AI to change what a manager spends time on. Hand the status reports and reconciliation to tooling so the freed hours go to coaching and judgment. That redesign is what makes a span of twelve humane. Cutting managers and calling AI the reason, with the job unchanged, is how a span of twelve becomes a slow-motion attrition event.

Flatter and faster is a defensible bet. Capital One went all-in on the public cloud years ago and rebuilt itself as a technology company, a much larger structural move, and the lesson from that kind of decision is that structure works when it follows a clear-eyed view of the work. Widening spans can do the same thing for an engineering organization: push decisions closer to the people doing the work, strip out coordination that AI can now carry, and give strong individual contributors room to lead without a title change. The redesign has to come first for any of that to hold. When the ratio comes first and the redesign never arrives, the organization spends the savings now and pays for them later.

The companies announcing flatter structures this year are running an experiment whether they admit it or not. The ones who treated span of control as a math problem will spend 2028 wondering where their bench went. The ones who treated it as a work-design problem will have built something faster that still develops people. The arithmetic of flat is easy. The org design underneath it is the actual work.

Here’s How You Take Action

If you lead an engineering organization: Before you set a span-of-control target, write down where each piece of the removed manager’s work will go. If “coaching” and “performance development” don’t have a clear new owner, you have deferred a problem and called it a design.

If you set workforce strategy: Pull your actual span distribution, not the average. Find the managers already past twelve reports and look at their teams’ engagement and attrition. The damage from over-widening is usually already in your data before you formalize the policy.

If you manage a large team and feel it: Name the specific job that fell off when your span grew. Take it to your leadership as a design question, not a personal failing. “My span doubled and coaching is the thing I can no longer do well” is a sentence that changes how the org should be built.

If you’re an individual contributor watching the layers come out: Ask your leadership what the technical track looks like, with real levels and real pay. If the answer is vague, that is useful information about whether this flatter org was designed or just announced.

For everyone: In your next reorg conversation, ask the question that separates design from arithmetic: when we widen this span, what work are we deciding to stop doing? If no one can answer, the spreadsheet is driving, and the spreadsheet does not coach anyone.

Christina Lexa writes Workforce Rewired, on the intersection of workforce transformation, AI, and global talent.

The views expressed here are my own and do not represent the position of my employer or any organization I am affiliated with.

For two years, one number has anchored the argument that companies are slow to adopt AI: the Census Bureau’s firm survey put business adoption at 5% to 7%, while worker surveys showed 35% to 40% of people using AI on the job. Economists treated that gap as a real puzzle about diffusion. A St. Louis Fed analysis by Alexander Bick, Adam Blandin, David Deming, and two co-authors says the puzzle was mostly an artifact of the question. The old survey asked only whether firms used AI to produce goods and services, which misses marketing, finance, and administration, where most AI actually shows up. When the Census Bureau widened the question in November 2025, reported firm adoption nearly doubled overnight, from about 10% to 17%, with no change in actual behavior.

Subscribe now

By the Numbers

• Reported U.S. firm AI adoption nearly doubled, from about 10% to 17%, the moment the Census Bureau changed its survey question in November 2025, with no change in underlying behavior (St. Louis Fed).

• Worker surveys put on-the-job AI use at 35% to 40%, against the old firm-survey reading of 5% to 7%, a gap the analysis attributes mostly to measurement (St. Louis Fed).

• Among European firms that use AI, production accounts for only about 21% of use cases; marketing and sales (35%) and business-process organization (31%) lead (St. Louis Fed, citing EU-ICT-Firm data).

• Projected to the U.S. using the European pattern, true any-purpose firm adoption would sit near 34%, close to the worker-survey figure (St. Louis Fed).

Research and Data

The AI adoption gap that worried economists was mostly a survey-design problem

The St. Louis Fed’s “On the Economy” analysis takes apart a statistic that has shaped two years of AI-and-work commentary. The Census Bureau’s Business Trends and Outlook Survey, the main U.S. read on firm-level AI use, asked from 2023 through October 2025 whether a business used AI “in producing goods or services.” That phrasing quietly excludes the functions where AI lands most often. In European data, which asks about AI for any business purpose, production is only about a fifth of use cases; marketing, sales, and back-office work make up the rest. When the Census Bureau swapped “producing goods or services” for “any of its business functions” in November 2025, reported adoption jumped from roughly 10% to 17% in a single step. The authors are direct that the jump reflects firms finally being asked about marketing and finance use, not a surge in new behavior. Their projection from the European pattern puts genuine U.S. any-purpose adoption near 34%, which closes most of the famous gap with worker surveys. The worker side of this is the part HR leaders should sit with: employees were not quietly using tools their companies had not adopted. The companies had adopted, and the survey simply was not asking in a way that counted it. The measurement fix matters for anyone who built a workforce thesis on the idea that firm adoption was stuck in single digits.

Source: Federal Reserve Bank of St. Louis, “On the Economy,” June 1, 2026, “Measuring AI Adoption among Firms: How You Ask Matters.” Read it here.

Why it matters: The “firms are slow, workers are ahead” story was built on a number that the question itself depressed, and corrected adoption sits near 34%, not 7%. If you have been pacing your own AI rollout against benchmarks that said most companies had barely started, you have been racing a phantom. Check whether your internal adoption metrics measure use across all functions or only the narrow slice your survey happens to ask about.

What Workforce Leaders Are Watching

• Corrected firm adoption near 34% means AI is already embedded in marketing, finance, and administration at most companies. Which functions in your own organization are using AI that your formal tracking does not yet count?

• The worker-versus-firm gap was largely a measurement gap, not a shadow-IT problem. Are your AI usage numbers built on a question broad enough to capture every function, or are you underreporting your own adoption to leadership?

• Several published workforce theses leaned on the old 5% to 7% figure as evidence of slow diffusion. Which assumptions in your strategy deck rest on that number, and do they survive a 34% reading?

• The Census Bureau fixed its question and the picture changed immediately. When the Department of Labor’s AI workforce hub publishes its first private-sector data, how will you tell a real trend from a measurement artifact?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

Oracle put a number on something it had spent months declining to confirm. In its annual regulatory filing, the company disclosed that it shed about 21,000 jobs over the past year and named AI as a direct cause, writing that the deployment of AI across its operations “have resulted, and may continue to result, in reductions to our workforce.” Oracle’s headcount fell to 141,000 from roughly 162,000, close to a 13% cut, and restructuring costs jumped to $1.8 billion from $374 million a year earlier. The cuts arrived during a stretch of double-digit revenue growth and surging cloud demand, which is what makes the filing notable: a profitable company stating in a securities document that AI is replacing workers. The people on the other side of that disclosure have been pushing back since spring. More than 600 employees signed a letter asking for better severance, many lost unvested stock that was months from vesting, and several told reporters they were treated as line items rather than colleagues.

By the Numbers

• Oracle cut about 21,000 jobs over the past year, naming AI as a direct cause in its annual SEC filing (Bloomberg).

• Headcount fell to 141,000 as of May 31, 2026, down from roughly 162,000 a year earlier, close to a 13% reduction (Bloomberg).

• Restructuring costs reached $1.8 billion, up from $374 million the prior year (CNBC).

• More than 600 Oracle employees signed a letter to management requesting improved severance and extended healthcare; the company said it would address concerns individually (TIME).

Layoffs and Company Decisions

Oracle tells the SEC that AI is cutting its workforce, and 21,000 jobs are gone

Oracle’s annual 10-K filing, surfaced by Bloomberg on June 22, confirmed a workforce reduction the company had let trade outlets speculate about for weeks. Headcount dropped to 141,000 from about 162,000, a roughly 13% cut, and Oracle wrote plainly that adopting and deploying AI “have resulted, and may continue to result, in reductions to our workforce.” Restructuring costs hit $1.8 billion, nearly five times the prior year. The financial backdrop sharpens the point: Oracle is growing, with double-digit revenue gains and accelerating cloud and AI infrastructure demand, and it is cutting people anyway to redirect spending toward data centers. Workers carried the cost in ways the filing does not capture. Employees who spoke with TIME described losing hundreds of thousands of dollars in unvested restricted stock, with one long-serving worker forfeiting awards worth close to $1 million that were due to vest within months. Oracle declined to accelerate vesting. When more than 600 employees signed a letter in April asking the company to match the severance terms that Meta, Microsoft, and Cloudflare offered their laid-off staff, Oracle said it would handle concerns one at a time. Calling a million dollars in forfeited stock and a refusal to match peer severance a workforce “reduction” launders the human cost out of the sentence.

Source: Bloomberg, June 22, 2026, “Oracle Layoffs Fueled by AI, Reduces Workforce by 21,000.” Read it here. Worker impact: TIME; financial detail: CNBC.

Why it matters: A profitable, growing company has now told its investors, in writing, that AI is the reason it employs 21,000 fewer people, which moves AI-driven cuts from quarterly euphemism into the legal record. The detail that lands hardest is the forfeited stock and the refusal to match peer severance during a year of double-digit growth. Leaders watching this should set severance and vesting policy before the restructuring, because how you treat the people you let go is the message every remaining employee reads.

What Workforce Leaders Are Watching

• Oracle named AI as a workforce-reduction cause in an SEC filing while posting double-digit growth. If your company is profitable and still planning AI-related cuts, what justification goes in writing, and does it match what executives say out loud?

• Oracle forfeited unvested stock and declined to match peer severance. What is your written policy on vesting acceleration and severance floors before a restructuring starts, not after employees organize a letter?

• More than 600 workers signed a collective letter and went public. At what point does your own workforce move from private complaint to organized pushback, and who inside the company is listening for it?

• Investors rewarded Oracle for redirecting payroll toward data centers. How are you measuring whether AI-driven cuts produce the productivity gains that justified them, or whether you are trading institutional knowledge for a capex story?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

The graduating class of 2026 just told researchers something their employers do not believe. More than 80% of rising seniors use generative AI, three quarters call themselves skilled with it, and most still expect AI to shrink their job market. Hiring managers looking at the same technology expect it to add jobs. Separately, Deloitte makes the structural case for why entry-level work is changing: AI is stripping out the coordination and synthesis tasks that built management layers and trained junior staff, and that lets companies fuse roles and flatten hierarchies. BCG adds the warning underneath both: in a survey of 70 senior executives, half already see de-skilling inside their own organizations, and the skills they call most critical are the ones eroding fastest. Stanford’s new economic dashboard puts numbers to where this lands first, and the answer is workers aged 22 to 25 in the most exposed roles. The early-career pipeline and the existing workforce are under the same pressure.

Subscribe now

By the Numbers

• More than 80% of class of 2026 rising seniors have used generative AI tools, with over a third using them daily (Handshake).

• 24% of rising seniors think AI will create jobs, compared with more than 50% of hiring managers (Handshake).

• 61% of the class of 2026 feel pessimistic about their careers; nearly half of them tie that to AI (Handshake).

• In a redesigned monthly financial close, manager time spent on synthesis drops from about 2 days per cycle to under half a day (Deloitte).

• Half of 70 surveyed C-suite and senior executives already see de-skilling in their organizations; more than 60% expect it to be a material threat within three to five years (BCG).

• In the two occupation groups most exposed to AI, employment for workers aged 22 to 25 has declined since ChatGPT’s release, while less-exposed groups grew (Stanford Digital Economy Lab).

Layoffs and Company Decisions

Deloitte reframes the org chart as a “work chart” and flattens the middle

Deloitte argues that hierarchies grew to manage friction: information latency, handoffs, escalation chains, the human relay that moves work across silos. When AI absorbs that coordination and synthesis, the middle layers shrink, and roles fuse across functions with end-to-end ownership. Deloitte calls the result a work chart rather than an org chart, mapping tasks and processes instead of people. Managers move toward orchestration, coaching, and exception governance. Entry-level work moves past first-draft production into verification, monitoring, and evaluation. The firm walks through a monthly financial close where automated matching and AI drafting cut manager synthesis time from roughly two days a cycle to under half a day, with editorial review turning into auditable control checkpoints. Deloitte is direct about the risk it creates: the old apprenticeship scaffolding disappears, so a manager overseeing five people across a flattened workflow faces harder judgment calls, and companies that do not redesign how juniors learn will open a capability gap in a few years.

Source: Deloitte, “Org chart vs. work charts: Organizational delayering.” Read it here.

Why it matters: Delayering is being sold as cleaner decisions and lower coordination cost, not headcount cuts, but it removes the rungs junior workers used to climb. The apprenticeship redesign is the precondition, not the afterthought. Leaders who flatten first and figure out junior development later will train no one to replace the managers they are stretching.

Subscribe now

Reskilling and Education

The class of 2026 is fluent in AI and afraid of it at the same time

Handshake’s workforce outlook on this year’s graduating seniors found a cohort that has folded generative AI into how it studies and works. More than 80% have used the tools, and most treat them as a brainstorming partner or study aid rather than a way to skip the thinking. Three quarters describe themselves as reasonably or very skilled. The confidence stops at the office door. Among the 61% who feel pessimistic about their careers, nearly half point at least partly to AI. Only 24% of seniors expect AI to create jobs, while more than half of hiring managers do. Handshake’s own postings data complicates both readings: job descriptions mentioning generative AI have climbed about fivefold since 2023, and the platform finds no clear sign yet that roles considered more exposed to AI are seeing steeper hiring slowdowns.

Source: CNBC, June 2026, citing Handshake’s Class of 2026 workforce outlook. Read it here. Underlying report: Handshake.

Why it matters: The people you are recruiting and the people doing the recruiting hold opposite forecasts about the same technology, and the recruits are the ones who built the skill. Employers who dismiss new-graduate AI anxiety as naivety will misread their pipeline. Ask early-career hires what they see, then close the gap with concrete role design rather than reassurance.

BCG: half of executives already watch their organizations lose skill to AI

BCG surveyed 70 C-suite and senior executives and found de-skilling already underway: half report seeing it inside their own organizations, and more than 60% expect it to threaten performance within three to five years. The skills draining away are the ones leaders rank as most critical, including judgment and decision-making, problem framing, and creative thinking. Routine AI use quietly takes over the practice that keeps those capabilities sharp. BCG’s prescription is structural rather than motivational: rethink where AI is allowed to do the work, redesign workflows and performance systems so people still exercise judgment, and build skill replenishment into everyday tasks instead of treating it as separate training.

Source: BCG, June 10, 2026, “When Everyone Uses AI, Companies Risk Losing Critical Skills.” Read it here.

Why it matters: Adoption metrics reward usage, and usage is the mechanism eroding the judgment leaders say they need most. The fix lives in workflow and performance design, not a training catalog. Decide which decisions stay human before the capability to make them well disappears.

Stanford launches a live tracker, and it already points at the youngest workers

Erik Brynjolfsson’s Stanford Digital Economy Lab put out a free platform, the AI Economic Indicators, to track AI’s effect on work and productivity month by month. One piece, the Canaries Dashboard, draws on anonymized ADP payroll data and builds on the lab’s “Canaries in the Coal Mine” study. The early read is specific: for workers aged 22 to 25 in the two occupation groups most exposed to AI, employment has fallen since ChatGPT arrived, while the three less-exposed groups kept growing. Those declines flatten out and disappear for older workers in the same fields. The lab pairs the dashboard with a Takeoff Tracker that weighs the evidence for AI driving rapid economic growth, with Brynjolfsson framing the goal as timely, trusted data on where AI creates value and where it disrupts work.

Source: Stanford Digital Economy Lab, June 10, 2026, AI Economic Indicators launch and June 2026 research note. Read it here.

Why it matters: The entry-level compression that Handshake, Deloitte, and BCG describe now has a monthly payroll-based gauge attached to it. The age-22-to-25 gap in exposed roles is measurable and moving, not anecdotal. Workforce planners can watch this dashboard and ask whether their own early-career numbers track the same curve.

What Workforce Leaders Are Watching

• When your early-career candidates expect AI to shrink opportunity and your hiring managers expect it to grow, which forecast is shaping your 2027 entry-level headcount plan, and have you checked it against your own posting data?

• Career pessimism among new graduates is a retention signal before they are even hired. What in your early-career value proposition answers the fear directly?

• If you flatten a layer of management, where do your juniors now learn the judgment that layer used to teach them, and who owns building that path before the gap shows up?

• Your AI dashboards reward adoption. Which judgment-heavy decisions are you deliberately keeping in human hands so the skill BCG says is eroding stays exercised?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

TL;DR: I draft with AI and I am not embarrassed to say so. What I am embarrassed by is the first draft, which arrives reeking of robot: em dashes everywhere, every paragraph resolving into a tidy little bow, the phrase “it’s not just X, it’s Y” three times in nine hundred words. So I wrote my own anti-AI style guide, and the audit against it now runs as a step inside my writing workflow, not as me alone with a red pen. I direct it, I read what comes back, and I make the final calls. Here is the whole checklist, free to steal.

Subscribe now

The confession

I write a newsletter about AI. I use AI to help write it. The people who find this scandalous are usually the same people whose “human-written” LinkedIn posts open with “In a world where...” and close with “The future is now.” So I will not be lectured.

What I will admit is that the raw output is bad in a very particular way. Not wrong. Bad. It reads like every other piece of AI prose on the internet, which is to say competent, smooth, and instantly forgettable (except you hear echos of it everywhere in the structure and language). The ideas can be mine. The sentences are nobody’s.

The first time I noticed this, I had drafted a section on entry-level hiring, read it back, and felt nothing. Every sentence was fine. The whole thing had the texture of a hotel lobby. Pleasant, anonymous, and impossible to remember once you have left.

So I started keeping a list of the specific things that made it sound like that. The list became a document, then a real style guide: every pattern to exclude, limit, or keep, with the reason for each. That guide is now a standing instruction the AI reads before it drafts a single sentence for me. It writes against my rules, runs the audit against my rules, and hands me the result. I read what comes back, catch what it missed, and make the calls it cannot. Nothing publishes until it survives that loop.

Why a banned-word search is not enough

The obvious move is to search for the famous AI words and delete them. “Delve.” “Tapestry.” “Leverage.” Fine. Do that. It takes thirty seconds and it is the easiest win available.

It is also nowhere near sufficient, because the strongest tells are not words. It’s the syntax. A reader who has seen a thousand AI posts does not consciously think “ah, the word delve.” They feel a wrongness in the rhythm, the way every paragraph builds the same way, the little rhetorical question that sets up its own answer. You can pass a banned-word search clean and still sound exactly like a machine.

The research backs this up. The most reliably identified AI tell is not vocabulary at all. It is a sentence structure: negative parallelism. “It’s not about the technology, it’s about the people.” “The question isn’t whether AI will change work. The question is how.” Once you see it you cannot unsee it, and AI produces it constantly, because it is a cheap way to manufacture the feeling of insight without any actual contrast underneath.

So my guide hunts syntax, not just words. Here is what it tells the AI to look for, and what I check behind it.

The audit, in full

The guide runs these as ordered passes. Each one is a sweep through the whole draft looking for one specific thing. Boring, mechanical, and it works. I do the same passes on the way back, because the AI applying my rules and me confirming them are two different safeguards, and I want both.

1. The typography sweep. This is the highest-value five minutes and almost nobody does it. Em dashes: zero, always, no exceptions, rewrite the sentence. I should confess that this rule costs me personally. I loved the em dash. For twenty years it was my favorite piece of punctuation, the elegant little bridge I dropped into a sentence whenever a comma felt too weak and a period felt too final. Then the machines learned to love it too, and now it is a tell, and I have had to give it up like a perfectly good habit that turned out to be a crime. I still reach for it. I still grieve. I use a comma and move on. Stray Markdown that survived the copy-paste: literal asterisks where bold should be, a hash mark in front of a header, bracket-paren link syntax sitting in the body. Headers in sentence case, not Title Case That Capitalizes Everything. URLs stripped of tracking junk, because a live link in your published post carrying utm_source=chatgpt.com is the most literal confession available to a writer. Then paste the whole thing through a plain-text editor once to kill the invisible Unicode characters AI sometimes smuggles in. You will not see them. They are there.

2. The negative parallelism hunt. Every “it’s not X, it’s Y” and its whole family. I rewrite each one affirmatively. “This is a management problem” says everything “It’s not a technology problem, it’s a management problem” says, in half the words, with more authority, and without the fingerprint. This single pass does more for the human-ness of a draft than any other.

3. The rhythm check. I read it aloud. If every sentence is the same length, something is wrong, because AI tends toward uniformity and humans do not. A short sentence lands hard after a long one. I also look for clusters of single-sentence paragraphs stacked for fake drama, and I consolidate them. One lone sentence can hit. Five in a row is a robot doing an impression of Hemingway.

4. The vocabulary kill list. Now the words. “Leverage” as a verb, “robust,” “seamless,” “unlock,” “streamline,” “underscore.” Plus the sneakier inflation cluster the 2026 frequency data flagged: “pivotal,” “showcase,” “foster,” “garner,” “a testament to,” “stands as a reminder of.” The pattern under all of them is the same. The word inflates a significance the writer has not earned with a fact. Replace it with the fact.

5. The significance-inflation cut. Any sentence that tells the reader something is important gets deleted. Not softened. Deleted. If a development matters, the specifics show it. “This represents a broader shift in how organizations think about talent” is a sentence that has decided to skip the part where it proves anything. The fact carries the weight, or there is no weight.

6. The caps on the real tools. Some AI patterns are also legitimate rhetorical moves, so I ration them instead of banning them. One tricolon per piece, the rule of three, because AI uses it to fake comprehensiveness. Anaphora twice, maximum. Not every bullet starting with a bolded phrase and a colon. These are tools. The tell is the overuse.

7. The source check. Every statistic verified and hyperlinked at the exact point it appears, not in a cleanup pass later. “Experts argue” and “research suggests” are not citations. They are the absence of one wearing a tie.

That is the whole audit. Seven passes. The AI runs them as it drafts because the rules live in its instructions, and my read-back on a long piece takes maybe twenty minutes, less once the patterns are in your own head and you start catching them on sight.

Subscribe now

The part where I tell on myself

The honest thing to report is that the loop still fails, on both ends, constantly.

I drafted this very piece with AI help, ran the typography sweep on the way back, and found two em dashes. Neither came from the machine. I had typed them myself, by hand, in the lines I revised. Eighteen years of writing habits do not evaporate because I wrote a rule against them, and the guide I built to catch the AI ended up catching me.

The AI misses things too. It left a “the question isn’t whether, it’s how” sitting in an early paragraph, which is a special kind of irony given that the paragraph was about negative parallelism. My read-back caught it. I rewrote it. That is the entire point of running the audit twice, once by the AI against my rules and once by me: neither pass is clean on its own, and the version of me that is confident at nine at night is not a safeguard I trust.

So I will be precise about who does what. I wrote the rules. The AI applies them while it drafts and flags what it can. I read every line, catch what slipped through on both sides, and decide what is actually mine. The machine just makes sure the rules get run every single time, which is more than I managed when the only auditor was me at the end of a long day.

Here’s how you take action

If you write anything with AI in the loop, build your own version of this. It does not have to be elaborate. Start here.

• Write the rules down where the AI can read them. Open a note and list the five things that most make AI writing sound like AI to you: em dashes, the “it’s not X, it’s Y” shape, headers in Title Case, the significance-announcing sentence, the bolded-colon bullet. Then paste that list into your system prompt or project instructions so the model applies it as it drafts, instead of you cleaning up the same mistakes by hand every time. Mine grew from six items to dozens, but it began that small.

• Separate the word passes from the syntax passes. Search-and-delete handles the vocabulary. Only reading aloud catches the rhythm and the structure. Do both, and do them as separate steps, because your brain cannot hunt words and syntax at the same time.

• Always run the typography sweep on the final paste. Em dashes, stray Markdown, Title-case headers, and tracking junk in URLs are the four tells that survive the move from your draft into Substack or LinkedIn. Strip them in the published version, not just the draft.

• Verify every number where it sits. When you cite a statistic, link the primary source in the same sentence, immediately. Do not promise yourself you will go back. You will not go back.

• Audit both ends. The AI will miss things even with your rules in front of it, and you will introduce your own tells in the lines you revise. It will still use words that make you go, huh? Read every draft yourself, no matter how clean the model claims it is. The point is a reliable way to catch the imperfect draft, whoever made it imperfect. That habit is cheaper than the trust you lose when a reader’s gut tells them a machine did all your thinking for you.

The machine drafts fast and now runs my rules while it does it. The taste, the direction, and the final call stay with me, and that is the part nobody can automate. The guide just makes sure the rules get run every time, which is the part I could never be trusted (or find the time) to do alone.

The Prompt Log is a recurring feature of Workforce Rewired. Published when there's something honest to say about the process. If you're building your own version of this, I'd like to hear what's on your list: christina@workforcerewired.co.

Jeff Bezos took the VivaTech stage in Paris on Wednesday and bet against the displacement story. AI will create a labor shortage, not mass unemployment, he argued, because human wants are endless and AI mostly removes the constraints that hold work back. He is now the most prominent name in a small group of optimists, alongside Apollo’s Torsten Slok and a recently softened Sam Altman and Dario Amodei. The timing is pointed. Tech layoffs citing AI passed 115,000 through May, Goldman Sachs counts roughly 16,000 AI-attributed cuts a month, and a Reuters/Ipsos poll this month found 53 percent of Americans fear AI could cost someone in their household a job. Bezos did not engage those numbers. The gap between his forecast and what workers report living through is the thing for HR leaders to hold.

Subscribe now

By the Numbers

• 53% of Americans worry AI could put them or someone in their household out of work, per a Reuters/Ipsos poll of 4,531 people conducted June 3 to 8.

• 71% in the same poll said they are concerned AI will put too many people out of work permanently.

• 115,000+ tech layoffs through May 2026 cited AI as a driver, approaching the full-year 2025 total, per figures Fortune cites in its coverage of Bezos’s remarks.

• ~16,000 U.S. jobs eliminated per month by AI, per Goldman Sachs estimates referenced in the same Fortune report, with entry-level and Gen Z workers absorbing the heaviest impact.

Layoffs and Company Decisions

Bezos Bets AI Creates a Labor Shortage, While Workers Brace for the Opposite

At VivaTech, Bezos said he “totally disagrees” with the view that AI will make humans redundant. His case rests on an old pattern: past industrial revolutions created more work than they destroyed, and people always find new things they want to build once a constraint lifts. In a May CNBC interview he reached for a bulldozer-versus-shovel metaphor, predicted deflation from productivity gains, and waved off displacement fears for radiologists and software engineers. The optimism is worth taking seriously, and it has company. What it skips is the present. Bezos did not address the 115,000 AI-cited tech cuts logged through May, the Goldman estimate of 16,000 jobs lost a month, or the CFO survey pointing to AI layoffs running nine times higher this year than last. His own audience tells a different story. A Reuters/Ipsos poll this month found 53 percent of Americans fear AI could put someone in their household out of work, and 71 percent worry it will displace too many people permanently. Bezos forecasts labor scarcity from a Paris stage. Millions of workers are watching their teams shrink. The long-run case for new jobs does not pay this quarter’s rent, and leaders who repeat the optimism without naming the transition cost ask workers to carry a risk the executives are not pricing. Forecast the shortage if the data supports it. Owe workers an honest account of the months between here and there.

Source: Fortune, June 17, 2026

Why it matters: Executive optimism about long-run job creation sets the tone employees hear, and right now it runs ahead of what those employees report experiencing. HR leaders who echo the labor-shortage forecast without a concrete plan for the displacement happening now widen the trust gap the poll numbers already show.

What Workforce Leaders Are Watching

• If AI does create a labor shortage in five years, the displacement is happening now. What is your company doing to carry affected workers across that gap rather than leaving them to absorb it alone?

• Bezos says human demand for work is endless once AI lifts constraints. For your own teams, which constraints would AI actually remove, and which roles would that expand versus eliminate?

• More than half of Americans now fear AI for their household income. How is that fear showing up in your workforce, in retention, in willingness to adopt new tools, in trust of leadership messaging?

• When founders forecast abundance and workers report contraction, employees notice which one leadership repeats. Is your internal AI message honest about the near-term cost, or only the long-term promise?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

TL;DR: Eighty-nine percent of engineering leaders say AI made their teams more productive. The same leaders admit their metrics miss tech debt, validation time, and burnout, and only 6 percent feel equipped to fix that. The gap between those numbers is where management is quietly breaking. When the work moves inside a model the manager can’t see, output stops telling you who is good at their job. Most companies are responding by buying more surveillance. The ones that pull ahead will rebuild what they measure.

Subscribe now

A manager has always done one private calculation, fast and mostly unconscious: look at what someone produced, and back out how good they are. Clean code, a tight memo, a deck that lands. The output was a proxy for the person. It told you who to promote, who to coach, who to trust with the hard thing next quarter.

That proxy just stopped working, and almost no one has said so out loud.

When a junior analyst hands you a flawless competitive brief, you no longer know what you’re looking at. Maybe she structured the analysis, found the non-obvious angle, and used the model to format it. Maybe she typed three sentences into a prompt and shipped what came back without reading it closely. The brief looks identical either way. The skill behind it is the difference between someone you fast-track and someone you’ve just learned you can’t rely on. The artifact stopped carrying that information.

This is the management problem of the decade, and it is not the one getting attention. The conversation is stuck on whether AI makes people more productive. That question is nearly settled and mostly boring. The harder question is what happens to evaluation, development, and trust when output decouples from the person who produced it.

The numbers don’t add up, and the people reporting them know it

Harness surveyed 700 engineers and managers across five countries for its State of Engineering Excellence 2026 report. The headline looks like a clean win: 89 percent of engineering leaders say developer productivity improved since they deployed AI. Then the same report turns the lights on. Eighty-one percent of those leaders say code review time went up after deploying AI, in many cases sharply. Developers estimate that roughly a third of their day now goes to AI-related work that shows up in no metric at all: reading code written in a style nobody recognizes, tracking down the subtle bug the model introduced, supervising ten agents running in parallel.

So output went up and invisible labor went up at the same time. Both are true. The dashboard only shows you the first one.

The most revealing finding in the Harness data is a contradiction sitting inside the same population. Eighty-nine percent of leaders say their current metrics accurately reflect AI’s impact. Ninety-four percent say those same metrics miss the factors that matter most: tech debt, validation time, developer burnout. Both groups are largely the same people. They are confident in a measurement system they simultaneously describe as blind. Only 6 percent believe they’re equipped to close the gap.

That is not a calibration error you fix with a better chart. High confidence in a system you know is incomplete is a coping mechanism. When a familiar metric is the only thing you have, you keep trusting it past the point where it tells you the truth, because the alternative is admitting you can’t see the work anymore.

Output was always a proxy for performance

Spend twenty years in talent strategy and you learn what performance measurement actually does. It runs on inference. A manager used tickets closed, lines shipped, decks delivered, deals booked as readable signs of the things that matter and are difficult to measure: judgment, initiative, reliability, the capacity to handle ambiguity. The output was legible on its face. The qualities underneath stayed hidden. So a manager read the obvious thing and inferred the rest.

AI severs the link between the two.

Now heavy output can mean a person is creative and fast, or it can mean they are dependent, indecisive, and outsourcing thinking they haven’t done. Light AI use can mean someone is resistant and falling behind, or it can mean they are skilled enough that they don’t need the crutch for routine work.

The Harness and Worklytics analyses both land on the same uncomfortable point: usage data is nearly uninterpretable as a performance signal, because the same number can mean opposite things about two different people.

This is why the self-reported gains deserve a hard look. METR surveyed 349 technical workers in early 2026 and found a median self-reported 1.4 to 2x improvement in the value of their work from AI tools. Striking, until you read METR’s own caution: in a 2025 controlled study, experienced developers believed AI sped them up by 20 percent while it actually slowed them down by 19 percent. People are confidently wrong about their own productivity. If the workers can’t accurately read their own output, the manager reading it secondhand through a dashboard has no chance.

Subscribe now

The reflex is surveillance; it makes the data worse

Faced with work they can no longer see, companies are reaching for the oldest tool in the drawer: watch harder. Roughly 78 percent of employers now run some form of employee monitoring, up from about 60 percent before the pandemic. AI-powered monitoring specifically jumped from 12 percent of enterprises in 2023 to 34 percent in 2026. CNBC reports that almost every Fortune 500 company now tracks AI usage at the group, role, or individual level, mostly because token cost became a budget line, not because anyone proved the tracking reveals who is good at their job.

Here is the problem with answering an inference crisis by counting keystrokes. The thing you can measure cheaply, activity, is exactly the thing AI made meaningless. You end up with a high-resolution picture of motion and no picture of value. Worse, the measured know they’re being measured. Harness found that 54 percent of practitioners fear individual performance evaluation based on AI productivity data, while managers are nearly four times more likely to report no concerns at all.

When measurement feels like surveillance, you don’t get accurate data. You get people optimizing for the dashboard. The analyst learns to generate the activity the system rewards instead of doing the work the business needs. You have spent money to make the signal dirtier.

What actually rebuilds the signal

The companies that come out ahead will stop trying to see the work and start measuring the thing output used to stand in for. That means moving evaluation away from artifacts and toward judgment, which is harder, slower, and one of the few things left that AI hasn’t commoditized.

A few moves separate the organizations rebuilding the signal from the ones buying more cameras:

• Measure shipped value. Harness found most organizations can say how much AI code was accepted, but few can say how much reached production. Acceptance is vanity. Production is the business. Apply the same test everywhere: count what survived contact with reality, and treat raw output volume as noise.

• Evaluate the editing. When the model writes the first draft, the human contribution is the judgment applied to it: what got cut, what got caught, what got challenged. Review those decisions. Ask someone to walk you through what the AI got wrong and how they knew.

• Separate improvement data from performance data, and say so. Developers in the Harness study asked for one thing plainly: keep the usage metrics out of the performance review, be transparent about what’s tracked, and involve them in defining it. None of that is technically hard. It requires the discipline to leave the easiest number in the room alone.

• Promote on problems posed. The work AI can’t touch is deciding which problem is worth solving. That has always been the senior skill. It is still the durable one. Watch who reframes the question, and pay less attention to who clears the queue fastest.

We can no longer build talent strategies on the assumption that output is a reliable read on the person behind it.

That assumption is now fully faulty, and pretending otherwise is how good people get mismanaged and mediocre ones get promoted on borrowed competence. The managers who notice first, and rebuild what they’re actually measuring, will spend the next five years making better calls than the competitors still staring at a dashboard that stopped telling the truth.

Here’s How You Take Action

If you manage people: Pick one direct report this week and stop reviewing their output. Instead, ask them to walk you through one decision the AI got wrong and how they caught it. That conversation tells you more about their judgment than a quarter of clean deliverables.

If you set measurement strategy: Audit your dashboards for the Harness gap. List what you measure, then list what actually drives value, tech debt, validation, the work that didn’t ship. If the two lists don’t match, your confidence in the first list is the problem.

If you’re being measured: Don’t perform for the dashboard. Build a record of the judgment calls you make, the problems you chose, the model output you rejected and why. When output stops proving skill, the people who can show their reasoning win.

For everyone: Ask the uncomfortable question in your next leadership meeting. When this team’s output looks identical whether the person is excellent or just well-prompted, how are we telling the difference? If the room goes quiet, you’ve found the work nobody is doing yet.

Christina Lexa writes Workforce Rewired, on the intersection of workforce transformation, AI, and global talent.

The views expressed here are my own and do not represent the position of my employer or any organization I am affiliated with.

Mark Zuckerberg told Meta staff this week that the company’s AI restructuring went badly in places, an admission that lands differently from the layoff memos that preceded it. In a note dated June 12, he wrote that the changes were disruptive, that leadership “made mistakes and will almost certainly make more,” and that Meta does not expect further company-wide layoffs in 2026. The reorganization he is now trying to stabilize cut roughly 8,000 jobs in May and moved about 7,000 more people into AI workflow roles, touching close to one in five employees. The same week, Meta raised its 2026 capital expenditure guidance toward $145 billion. The contrast between record AI spending and an apology for how the workforce was handled is the story HR leaders should sit with.

Subscribe now

By the Numbers

• 8,000 Meta employees cut in May, with about 7,000 more reassigned into AI workflow roles, together affecting close to one in five of the company’s roughly 78,000 staff, per Reuters.

• No further company-wide layoffs expected at Meta in 2026, per Zuckerberg’s June 12 memo, a stated expectation rather than a guarantee for individual teams.

• Up to $145 billion in 2026 capital expenditure guidance, raised the same week as the stabilization memo, nearly double Meta’s 2025 spend.

Layoffs and Company Decisions

Zuckerberg Admits Meta’s AI Restructuring Went Wrong, Promises Stability

Zuckerberg’s June 12 memo is a rare executive admission that an AI-driven reorganization caused real damage. He called the changes disruptive, said “given the complexity of these changes, we’ve made mistakes and will almost certainly make more,” and committed to “providing as much stability as possible.” Inside the company, the picture is harder than the memo. Meta’s AI division has seen open dissent over its new rules, and employees describe morale at its lowest in years after a spring in which 8,000 people were cut and 7,000 reassigned. Chief Product Officer Chris Cox, at a company meeting, compared the past few months to running a marathon in a hailstorm. The stabilization measures Zuckerberg offered, bigger budgets for team offsites and a July hackathon, read as small against that. The phrase that deserves attention is “made mistakes.” It names the people side of a restructuring that most companies describe only in headcount and efficiency terms. Naming it is the first honest step. Repairing trust with the workers who survived the cuts, and the ones moved into roles they did not choose, is the harder one, and event budgets do not get there.

Source: Reuters, June 12, 2026

Why it matters: An admission of mistakes earns its weight when it changes how the next restructuring is run, beyond how this one gets described after the fact. Leaders moving people around AI should treat reassignment as a change to manage with the same care as a layoff, because the workers reshuffled into unfamiliar roles carry the morale cost long after the headcount math settles.

What Workforce Leaders Are Watching

• Meta reassigned 7,000 people into AI workflow roles. What support are those workers getting to succeed in jobs they did not apply for, and how is Meta measuring whether the redeployment worked or just relabeled the disruption?

• Zuckerberg ruled out further company-wide layoffs in 2026 but left room for team-level cuts. How should employees read a stability promise that holds at the company level and not the team level?

• When a CEO admits an AI restructuring caused harm, what would accountability look like beyond a memo: changes to severance, to redeployment design, to how the next reorganization is planned?

• Meta is spending toward $145 billion on AI while apologizing for how it handled the workforce that spending displaced. Which companies are budgeting for the human transition at any fraction of what they budget for the technology?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

PwC released its 2026 Global AI Jobs Barometer this morning, and the headline finding is a labor market splitting in two. Jobs that demand specific AI skills are growing almost eight times faster than the market overall, and the wage premium for those skills has climbed to 62%. Underneath that average sits a harder pattern: AI-exposed entry-level roles are now seven times more likely to require senior-level judgment and leadership, so the bottom rung is being raised even as some rungs disappear. Two Bloomberg columns this week press on what the Barometer’s growth numbers obscure. Parmy Olson argues the most likely outcome is a quiet erosion of the quality of the jobs that remain. Gautam Mukunda of Yale describes a related cost, using the case of a Carnegie Mellon graduate student whose two years of work an AI system reproduced in five days: when the meaningful part of a job moves to the machine, the motivation to do the rest goes with it.

Subscribe now

By the Numbers

• 62% wage premium now attaches to jobs requiring specific AI skills, up from 56% a year earlier, per PwC’s 2026 Global AI Jobs Barometer.

• Almost 8x faster job growth (69% versus 9% for the market overall) for roles that require AI skills, per PwC.

• 7x more likely that AI-exposed entry-level roles require traditionally senior skills such as judgment and leadership; these roles grew 35% since 2019 while other entry-level roles fell 10%, per PwC.

• 2x the job growth and 42% faster salary growth for “professionalised” roles, where AI handles routine tasks, versus “democratised” roles, per PwC.

• 5 days for an AI system to finish a mathematical-proof verification that took a Carnegie Mellon graduate student more than two years, the case Yale’s Gautam Mukunda uses to argue AI is coming for worker motivation.

Layoffs and Company Decisions

The AI Jobs Story Is Shifting From Headcount to Job Quality

Bloomberg’s Parmy Olson makes the case that the AI employment debate has fixated on the wrong question. The apocalypse Dario Amodei predicted last year, more than half of entry-level white-collar jobs gone within five years, has not arrived on schedule, and Amodei himself has shifted toward a productivity framing: automate 90% of a job and the worker does the remaining 10%. Olson’s point is that the remaining 10% is often the thin, repetitive, or supervisory part, while the substantive work moves to the model. She describes a slow hollowing of the jobs people keep, harder to measure and easier to ignore than a layoff announcement. Amazon and Citigroup cutting workers gets attention. A role that still exists but has lost its craft does not.

Source: Bloomberg, June 15, 2026

Why it matters: Headcount dashboards will read green while engagement and skill development quietly decay underneath them. HR leaders should track what AI is doing to the content of jobs, not only the count, because a stable headcount of hollowed-out roles is its own retention and capability problem.

When AI Takes the Meaningful Part of the Work, Motivation Goes With It

Gautam Mukunda, who teaches leadership at the Yale School of Management, opens with Sidharth Hariharan, a Carnegie Mellon mathematics graduate student who spent more than two years translating a celebrated proof into a form a computer could check. An AI system called Gauss, built by the startup Math Inc., finished the same task in five days. Mukunda’s argument locates the threat to engagement in AI taking the part of a job that made it worth doing. People stay motivated when they own hard problems. Hand the hard problem to a model and leave the cleanup to the human, and the work stops feeling like theirs. He puts the burden on leaders to redesign roles so that people keep meaningful ownership, rather than assuming productivity gains arrive free of motivational cost.

Source: Bloomberg, June 12, 2026

Why it matters: Role design now carries motivational stakes that did not exist when automation handled only routine tasks. Managers who deploy AI on the hardest, most satisfying parts of a job to maximize speed may win the quarter and lose the people, so the design choice is which problems stay human.

Reskilling and Education

PwC: AI Is Building a Two-Track Labor Market, and the Tracks Pay Differently

PwC’s 2026 Global AI Jobs Barometer, published June 15, separates the labor market into two paths. On one track sit “professionalised” roles, where AI absorbs routine tasks and the human work shifts toward judgment, creativity, and leadership. On the other sit “democratised” roles, where AI lowers the skill needed to do the work. The professionalised track is growing twice as fast and paying 42% faster salary growth. Skills employers want for AI-exposed jobs are changing 66% faster than for other jobs, two and a half times the rate of churn PwC measured a year ago. The entry-level finding cuts against the standard fear that AI simply erases junior work. AI-exposed entry-level roles grew 35% since 2019, and they grew by demanding senior-level skills, seven times more often than comparable roles, while entry-level roles untouched by AI declined 10%. Companies in the most AI-exposed sectors posted 34% productivity growth over 2018, against 24% for the least exposed.

Source: PwC, June 15, 2026

Why it matters: The entry-level rung is moving up, which breaks the traditional earn-while-you-learn path into senior work. Organizations that still expect new hires to grow into judgment through years of routine tasks need a faster way to build that judgment, because the routine tasks that used to teach it are gone.

What Workforce Leaders Are Watching

• If PwC is right that AI-exposed entry-level roles now demand senior judgment, how do organizations build that judgment in people who never get the years of routine repetition that used to produce it?

• What metric captures the job-quality erosion Parmy Olson describes, given that headcount, attrition, and engagement surveys can all stay flat while the substance of the work drains out?

• When a manager decides which tasks to hand to AI, what stops the default from being the hardest and most satisfying ones, the choice Gautam Mukunda warns removes the reason people stay?

• A 62% wage premium for AI skills rewards the workers who already have them. Which employers are closing that gap internally rather than buying it on the open market at a premium?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

TL;DR: Anthropic released its most powerful model on a Tuesday and the U.S. government forced it offline by Friday. The lesson is not about one company or one model. It is about what happens to the people and the work when the tools they depend on come in and out of favor faster than any planning cycle can absorb, while the government keeps writing the rules after deployment instead of before. If your workforce strategy assumes the tools you adopt will still be there next quarter, you are planning on sand.

Subscribe now

On Tuesday, June 9, Anthropic released Claude Fable 5, calling it the most capable model it had ever made generally available and noting it was particularly effective at finding software vulnerabilities. Developers wired it into products. Teams rebuilt workflows around it. The usual rush to adopt the new frontier began.

On Friday at 5:21 in the evening, Eastern time, a U.S. government export control directive ordered the company to cut off access for any foreign national, citing national security concerns. Because Anthropic cannot sort foreign nationals from everyone else in real time, the practical effect was a hard shutoff. The company disabled Fable 5 and its controlled sibling Mythos 5 worldwide. Three days from launch to dark.

I covered the launch in Thursday’s daily briefing, the same edition where I noted Anthropic pledging $350 million to study AI’s economic impact and Dario Amodei publishing a tiered framework for how government should respond to AI-driven unemployment. Less than 24 hours after that briefing went out, the government did something nobody’s framework anticipated: it made the most advanced tool on the market unusable for everyone, overnight, over a dispute about a jailbreak the company says it has only seen demonstrated verbally and narrowly.

Then the story got more interesting. The government did not stumble onto this concern on its own. According to the Wall Street Journal, Amazon CEO Andy Jassy raised concerns about the models directly with senior administration officials, and the jailbreak demonstration that drove the directive was run by Amazon's own researchers, who prompted the model into disclosing software vulnerabilities. Amazon is also one of Anthropic's largest investors. So the action that took the most capable tool on the market offline was set in motion by a competitor who is also a backer, talking to the people who hold the regulatory lever. Read that sentence twice.

Anthropic is complying while disputing the rationale, calling it a likely misunderstanding and working to restore access. That fight will resolve one way or another. The fight is not the point. The point is what just got demonstrated to every organization watching: a capability you adopted on Tuesday can be gone by Friday, and the reason can have nothing to do with you, your industry, or anything you did.

This is not a one-off

The instinct is to file this under “unusual government action involving a frontier model” and move on. That would be a mistake, because the pattern is already well established and this is just the most dramatic version of it.

Cast your mind back eighteen months. DeepSeek arrived, impressed everyone with its performance, and within weeks became radioactive. Texas banned it from state devices on January 31, 2025. New York followed on February 10. Virginia, Iowa, Kansas, and others lined up behind them, and Congress took up a bipartisan bill to bar it from federal devices entirely. A tool that looked like the smart, cheap choice in January was a compliance liability by April. Any organization that had built a workflow on it spent the spring ripping it out.

Different trigger, identical shape. A tool enters the market, gets adopted at speed, then gets restricted faster than the organizations relying on it can re-plan. Foreign ownership, security vulnerability, export control, data residency. The specific reason changes, but the pattern doesn’t.

Here is the uncomfortable truth most adoption strategies refuse to name. The frontier moves faster than the rules, and the rules, when they finally arrive, arrive abruptly. That combination turns every tool decision into a bet. Not a bet on whether the tool works. A bet on whether the tool will still be permitted by the time you have built anything that depends on it.

Subscribe now

What this does to the work, and the people doing it

For years I built workforce plans that touched operations around the world, and the planning assumption underneath all of them was continuity. You assume the capabilities your people use today will exist tomorrow, so you train for them, hire for them, and design roles around them. That assumption held for decades because the tools of white-collar work changed on a timescale measured in years. Email. Spreadsheets. The cloud. Each transition gave organizations time to absorb it.

Most of us were already working with a different set of assumptions in the age of AI about the pace of change. However, AI tools change on a timescale measured in days, and now the policy reactions to them do too. When the most capable tool on the market can vanish in 72 hours, the continuity assumption breaks, and a lot of quiet organizational design choices break with it.

Consider what actually happens inside a company when a tool disappears overnight:

• The team that rebuilt a workflow around Fable 5’s vulnerability-finding strength now has a workflow with a hole in it, and no replacement of equal capability to drop in.

• The people who got good at that tool, who developed the prompting instincts and the judgment about when to trust it, have skills suddenly attached to nothing.

• The manager who promised faster delivery on the strength of the new capability has to walk it back, to a customer or to their own leadership.

• The workforce plan that assumed a productivity gain from the tool now assumes a gain that evaporated.

None of those people did anything wrong. They adopted a sanctioned, publicly released tool from a credible company. The capability was real on Tuesday. The exposure was structural, and it sat outside anything they controlled.

This is the case for designing human work around durable capabilities rather than specific products. The instinct to train your people on the hot tool is understandable and partly right, because they do need fluency. But the deeper investment is in the capability the tool happens to deliver this quarter. The judgment to evaluate a vulnerability, not the specific model that surfaces it. The skill of directing and checking AI output, not loyalty to one vendor’s interface. Tools are rentals. The human capabilities around them are what you own, and they are the only part of the stack that survives a Friday-evening directive.

“Tools are rentals. The human capabilities around them are what you own, and they are the only part of the stack that survives a Friday-evening directive.”

Subscribe now

The government is still behind the eight ball

Now the harder half of the story, because the volatility has a source beyond the speed of the technology.

The United States has no comprehensive federal AI law. What it has instead is a patchwork of state statutes, agency guidance, and voluntary standards, with states filling the federal vacuum on their own terms. Through 2025, attempts to impose a federal moratorium on state AI laws failed, including a proposed ten-year freeze that the Senate stripped from a budget bill on a 99-1 vote. In March 2026, the White House released a National Policy Framework for Artificial Intelligence, urging Congress to replace the patchwork with a uniform approach. The framework is non-binding and creates no compliance obligations.

So when a genuine national security concern surfaced about Fable 5, the government did not have a calibrated instrument to reach for. It had an export control directive, a blunt tool built for a different era, and it swung that tool after the model was already in the wild. The restriction did not prevent a risk before deployment. It reacted to one after, and the reaction was so imprecise that it took down the tool for every user on earth to address a concern nominally about foreign nationals.

The Amazon detail sharpens the whole problem. When the only instruments available are blunt and reactive, they become unusually easy for a private actor to aim. A competitor with a relationship to the right officials and a research team willing to build the demonstration can point a national security lever at a rival's product. War-gaming that scenario used to belong in the someday pile. It just happened, in public, to the most capable tool on the market.

I spent eight years serving federal government clients, and I will say this plainly and without criticism of any administration: government moves at the speed of process, and process is slow by design. That slowness is a feature when the job is deliberation and a liability when the job is keeping pace with a technology that ships a new frontier every few months. The mismatch is structural. It will not be fixed by any single law, and it is not going away.

For organizations, the consequence is precise. You are operating in an environment where the rules governing your most powerful tools are written reactively, applied bluntly, and changed without warning. The regulator is not a stable backdrop you can plan against. The regulator is itself a source of volatility, because it acts late and then acts hard.

That is the part most AI adoption conversations skip. Leaders ask whether a tool is good and whether their people can use it. The better question (or at minimum, one asking in addition), the one almost nobody is asking out loud, is whether the tool will still be legal to use by the time it is load-bearing in your operation, and whether you have designed your work so that the answer does not sink you.

What to do with this

The takeaway is not to slow down or to wait for the dust to settle, because the dust will not settle. The takeaway is to build for the volatility you now know is permanent.

Here are three moves, ordered by how much they protect you.

1. First, design human work around capabilities, not products. Ask of every AI investment: if this specific tool disappeared tomorrow, what would my people still know how to do? If the honest answer is “very little,” you have built dependency, not capability. Train for the judgment, the evaluation, the direction of AI output. Those skills move with your people across whatever tool survives the next directive.

2. Second, treat tool adoption like a portfolio, not a marriage. Avoid letting a single model become load-bearing in a critical workflow without a fallback of comparable function. The team that lost Fable 5 on Friday is in a different position than the team that could route the same work to a second capability with a day of reconfiguration. Redundancy costs something. So does a hole in your operation on a Friday night.

3. Third, put policy volatility into your planning assumptions explicitly, the way you already plan for currency risk or supply disruption. Your 2027 plan should name which capabilities you will build on, which of them could be restricted, where that restriction could land, and what your operation does the day it happens. You cannot predict the next directive. You can refuse to be surprised that one is coming.

The tool that vanished this week will probably come back. The condition it revealed is here to stay. The organizations that come out ahead will be the ones that stopped treating their tools as permanent and started treating the human capability around those tools as the only asset they actually own.

For people who want better questions: the next time a vendor or a consultant tells you a new model will transform your operation, ask them what your operation looks like the morning it gets pulled. If they have not thought about it, you have just found the gap in the plan.

Workforce Rewired publishes weekly at the intersection of AI, organizational design, and the future of work. If this piece made you think, share it with someone navigating the same questions.

The views expressed here are my own and do not represent the position of my employer or any organization I am affiliated with.

TL;DR: For thirty years, India’s IT services industry ran the world’s largest white-collar apprenticeship, and the world hired its output: 71 percent of US H-1B approvals, the top source of international students, 11 Fortune 500 CEOs. AI has collapsed the intake at the same moment Western visa policy is narrowing the export channel. This piece traces the double squeeze on the global talent supply and what to do before your 2030 workforce plan meets a pool nobody restocked.

Subscribe now

In fiscal year 2022, India’s IT services industry hired roughly 600,000 new college graduates (often called ‘freshers’). By fiscal 2025, that number had fallen to about 120,000, according to talent analytics firm Xpheno. Four out of five entry doors into the world’s largest technology services workforce closed in three years.

The numbers come from India. The consequences land on every workforce plan, in every country, that assumes experienced technology talent will be available to hire in 2030, because India has been the world’s single largest supplier of it.

Here is the model that just broke. For three decades, TCS, Infosys, Wipro, and their peers hired graduates in six-figure batches, ran them through three to six months of training, and deployed them on the routine work of the global economy: testing, maintenance, support, basic development. The industry called it the pyramid. A wide base of juniors, narrowing toward a small number of senior architects and delivery leaders. The pyramid’s most valuable export was people, with the code almost a byproduct. Every year it converted hundreds of thousands of graduates into experienced professionals, and the entire global market hired from the output.

I have spent years building workforce strategies that touched include India, and I can tell you the assumption was never written down because it never needed to be. Experienced talent was treated like a renewable resource. The pyramid replenished it. You planned around salary inflation and attrition, never around the possibility that the intake itself would stop.

The intake is stopping. AI tools now perform the testing, maintenance, and basic development work that justified the fresher cohorts. TCS cut around 12,200 jobs in 2025, the largest reduction in its history, a reduction Outlook Business read as the collapse of the talent pyramid as an operating model. The head of digital operations at Kimberly-Clark put the trajectory plainly in May: “(The) zero-to-two-years experience ... will go away is my assumption in the next few years.”

The boom that hides the problem

The other half of the same labor market is thriving, and the contrast is the whole story. India’s global capability centers, the captive technology arms of multinationals, added nearly 200,000 net employees in fiscal 2026, almost twice the 110,000 added by IT services firms. The Nasscom-Zinnov 2026 landscape report counts 2,117 centers employing 2.36 million professionals. When I wrote about the GCC model in post-03, the open question was whether the cost arbitrage that built these centers would survive. The answer turned out to be that it did not need to: the centers moved up-market into AI platforms, product engineering, and cybersecurity.

But watch who they are hiring. TeamLease Digital data reported by The Economic Times shows mid-to-senior talent rose from 60 percent of GCC hiring in 2023 to more than 77 percent in fiscal 2026. The fastest-growing employers in the world’s largest technology labor market have, as a group, nearly stopped buying at the entry level.

The strain is already measurable. Quess Corp estimates a 38 to 42 percent gap in AI and data skills across India’s GCC ecosystem, and Deloitte’s Rohan Lobo told The Economic Times that demand for AI specialists has risen more than 300 percent since 2024. Blame for that gap usually lands on universities and training programs. The arithmetic points somewhere else: an industry stopped manufacturing the very people it now cannot find. The senior AI engineer of 2031 is the junior hire of 2026, and if the junior hire of 2026 went unmade, the senior engineer of 2031 goes unmade with her. Recruiting budgets change none of that math, because recruiting only redistributes the talent that already exists

Subscribe now

The double squeeze

What makes this everyone’s problem is what India has been to the rest of the world: the largest exporter of high-skilled talent in modern history. India-born workers received 71 percent of all US H-1B approvals in fiscal 2024, roughly 283,000 people. India is the top source of international students in American universities for the second consecutive year. And the channel reaches the very top of Western business: 11 Fortune 500 companies, with a combined market value above $6.5 trillion, are led by CEOs of Indian heritage, Microsoft and Google among them. Trace those careers back far enough and many begin in the same place: an entry-level seat in the Indian talent system.

Now the channel is narrowing at the same moment the source is shrinking. A $100,000 supplemental fee on new H-1B petitions took effect in September 2025, replacing a previous cost of $2,000 to $5,000 per petition. New international student enrollment in the US fell 17 percent in fall 2025, on top of a 7 percent drop the year before, with the steepest declines among Indian students. Whatever you think of the policy choices, the workforce arithmetic stays the same: the US, Europe, and Canada built their technology and leadership benches partly on talent India developed, and the development and the export are constricting at once.

In post-02, I argued that AI is severing the apprenticeship layer inside American companies. The standing answer to that problem has always been import: if you stop growing your own seniors, hire someone else’s. India’s numbers close that exit. Both jaws of the vise are tightening on the same future hire.

The uncomfortable truth

Global talent strategy at most large companies rests on an assumption nobody checks: that experienced talent is a renewable resource replenished by someone else. Someone else trains the juniors, absorbs the cost of the unproductive first two years, and converts graduates into professionals. Your job is to show up at year five with a competitive offer or a visa petition and collect. That someone else has left the market.

The organizations that come out ahead will switch from buying to making, and a few are showing what that looks like. Infosys held its fresher target at 20,000 this year, with its CFO describing the intake as an investment in future capacity. Accenture’s Julie Sweet has argued that younger employees adapt to AI tools fastest, which reframes the entry-level hire from a cost center into an advantage. The old pyramid trained people by handing them five years of routine work, and that work is gone. Whatever replaces it has to be deliberate: smaller cohorts, AI-fluent from day one, given real problems with real stakes and enough supervision to build judgment on purpose rather than by osmosis. That is harder to run than a training bootcamp and a bench. It is also the only mechanism left that manufactures the talent everyone’s 2030 workforce plan assumes will be available.

The pyramid worked because someone was willing to pay for the base. For thirty years, that someone was the Indian IT services industry, and the whole world collected the dividend. The question your organization should be asking is the one the industry just stopped answering: who pays for the base now?

Subscribe now

Here’s How You Take Action

1. Audit your dependency on talent you did not develop. Pull your hiring plan for the next three years and mark every role that assumes an experienced external hire, then note how many of those hires have historically come through immigration channels. That share of your plan is a double bet: on a supply pool that is no longer being restocked, and on an import channel that is narrowing. Name the number and put it in front of your leadership team.

2. Price the make-versus-buy decision honestly. Recruiting premiums for experienced AI talent are rising with a 300 percent demand spike behind them. Run the comparison: five years of escalating buy-side premiums versus the cost of a deliberate early-career development engine. The bootcamp-and-bench model is dead, so run the comparison against a structured modern apprenticeship rather than against nostalgia.

3. If you operate global capability centers, stress-test your 2030 bench. Take your center’s current fresher intake and project your senior staffing five years out. If the intake is near zero, your growth plan depends entirely on poaching from a market where every competitor is running the same play.

4. Find the practice loops your AI deployments deleted, and rebuild them on purpose. Every routine task AI absorbed was doing double duty as someone’s training ground. Identify where your early-career people used to build judgment, and design replacement experiences with real stakes and real supervision. Osmosis retired along with the routine work, so judgment development is now a design problem, and it belongs on someone’s roadmap by name.

Workforce Rewired publishes weekly at the intersection of AI, organizational design, and the future of work. If this piece made you think, share it with someone navigating the same questions.

The views expressed here are my own and do not represent the position of my employer or any organization I am affiliated with.

The Wall Street Journal put the AI jobs question to 16 prominent economists, including Nobel laureate Daron Acemoglu, and got near-unanimity on productivity alongside a three-way split on employment. Separately, Anthropic pledged $200 million to research AI’s economic impact while CEO Dario Amodei published a tiered framework for how government should respond to AI-driven unemployment. New Ipsos data adds a ground-level view: more than a third of American workers have yet to touch AI on the job, and the age divide is wide.

Subscribe now

By the Numbers

• 15 of 16 economists surveyed by the Wall Street Journal say AI will meaningfully lift labor productivity; none said it will not

• On whether AI eliminates more jobs than it adds, the same panel split: 8 expect no net change, 5 expect net losses, 2 expect net growth

• $200 million: Anthropic’s new Economic Futures Research Fund, paired with a $150 million fellowship program for early-career professionals

• 38% of U.S. workers say they are not using AI at work at all; among workers over 55, that figure is 59%

• 1 in 10 workers say they worry their employer wants to replace them with AI tools

Layoffs and Company Decisions

Sixteen Top Economists Agree AI Lifts Productivity, Then Split Three Ways on Jobs

The Wall Street Journal surveyed 16 economists on AI and the future of work, a panel that included Daron Acemoglu and two former chairs of the White House Council of Economic Advisers. Fifteen said AI will meaningfully lift labor productivity. On net employment, eight expect no change, five expect losses, and two expect growth. Acemoglu, voting net loss, pointed to the China import shock and robot adoption as precedents for sudden, geographically concentrated displacement. Harvard’s Jason Furman, in the no-change camp, called the current aggregate labor-market effect “small to zero,” and MIT’s David Autor put real impacts five to ten years out while warning that institutions are unprepared. The disagreement lands in a market where companies attributed 40% of May’s announced job cuts to AI even as payrolls grew by 172,000. Jed Kolko of the Peterson Institute offered one reconciliation: a CEO can more proudly blame AI for a layoff than admit to over-hiring after the pandemic. A New York Times Magazine expert panel published the same day reached a similar destination, with former White House AI adviser Dean Ball concluding, “We need better empirical economic data. You can’t create policy remedies for a problem you don’t understand.”

Source: Wall Street Journal | Related: New York Times Magazine

Why it matters: Executives are attaching AI rationales to headcount decisions faster than economists can verify them, and the experts themselves cannot agree on the direction of the net effect. Before an AI justification reaches a press release or a WARN notice, workforce leaders should pressure-test whether the technology actually changed the work.

Policy and Government

Anthropic Commits $350 Million to AI Economic Research as Amodei Proposes Unemployment Response Tiers

Anthropic announced a $200 million Economic Futures Research Fund on Wednesday to back empirical research, policy trials, and program evaluation on how AI reshapes labor markets, alongside a $150 million national fellowship program aimed at helping early-career professionals extend AI’s benefits beyond coastal tech hubs. CEO Dario Amodei published a companion essay proposing a tiered government policy framework keyed to three unemployment scenarios: 5%, 10%, and an unspecified “unprecedented” level. The lower tiers lean on wage insurance, capital accounts, tax incentives, and an expanded social safety net; the unprecedented tier contemplates basic income, sovereign wealth models, and equity-sharing mechanisms. The framework also calls for government authority to block deployment of models posing significant risk of catastrophic harm.

Source: Associated Press | Anthropic policy framework (PDF)

Why it matters: An AI lab is now drafting the displacement safety net before Congress has produced one, and tying its proposals to specific unemployment thresholds that can be tracked. Watch whether the research fund produces evaluations policymakers actually adopt, and whether other labs match the commitment or let Anthropic carry the policy conversation alone.

Reskilling and Education

Ipsos: A Third of Workers Have Not Touched AI on the Job, and Workers Over 55 Are Sitting Out at Twice the Rate

New Ipsos Consumer Tracker data, published June 8, finds 38% of full- and part-time workers are not using AI at work at all, with a sharp generational split: 26% of workers under 35 abstain, against 59% of workers over 55. Among those who do use the tools, sentiment leans positive but stays unsettled. 44% say AI makes them more productive, 38% say the tools are a good start but not ready to produce finished work, and 34% say AI cuts time spent on tasks they dislike. One in five report that time saved on some tasks just means more tasks assigned. One in ten worry their employer wants to replace them with AI.

Source: Ipsos Consumer Tracker

Why it matters: Adoption mandates assume a workforce that is already on board, and this data shows a third sitting out entirely, concentrated among the most experienced employees. Training programs that treat the 55-plus cohort as an afterthought will widen the adoption gap rather than close it.

What Workforce Leaders Are Watching

• If 16 of the most-cited economists in the field cannot agree whether AI adds or eliminates jobs, what evidence standard should govern the workforce planning assumptions going into 2027 budgets?

• Anthropic pegged its policy framework to unemployment thresholds of 5% and 10%. Which internal indicators would tell an HR leader that displacement is accelerating inside the company before any national number moves?

• Workers over 55 are opting out of AI at more than twice the rate of younger colleagues. Does the training design treat them as a distinct audience with distinct incentives, or assume one program fits all?

• Both expert panels this week ended on the same demand for better measurement. What would it take to instrument a single organization’s AI impact on roles, tasks, and headcount, ahead of any federal reporting requirement?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

The Wall Street Journal put the AI jobs question to 16 prominent economists, including Nobel laureate Daron Acemoglu, and got near-unanimity on productivity alongside a three-way split on employment. Separately, Anthropic pledged $200 million to research AI’s economic impact while CEO Dario Amodei published a tiered framework for how government should respond to AI-driven unemployment. New Ipsos data adds a ground-level view: more than a third of American workers have yet to touch AI on the job, and the age divide is wide.

Subscribe now

By the Numbers

• 15 of 16 economists surveyed by the Wall Street Journal say AI will meaningfully lift labor productivity; none said it will not

• On whether AI eliminates more jobs than it adds, the same panel split: 8 expect no net change, 5 expect net losses, 2 expect net growth

• $200 million: Anthropic’s new Economic Futures Research Fund, paired with a $150 million fellowship program for early-career professionals

• 38% of U.S. workers say they are not using AI at work at all; among workers over 55, that figure is 59%

• 1 in 10 workers say they worry their employer wants to replace them with AI tools

Layoffs and Company Decisions

Sixteen Top Economists Agree AI Lifts Productivity, Then Split Three Ways on Jobs

The Wall Street Journal surveyed 16 economists on AI and the future of work, a panel that included Daron Acemoglu and two former chairs of the White House Council of Economic Advisers. Fifteen said AI will meaningfully lift labor productivity. On net employment, eight expect no change, five expect losses, and two expect growth. Acemoglu, voting net loss, pointed to the China import shock and robot adoption as precedents for sudden, geographically concentrated displacement. Harvard’s Jason Furman, in the no-change camp, called the current aggregate labor-market effect “small to zero,” and MIT’s David Autor put real impacts five to ten years out while warning that institutions are unprepared. The disagreement lands in a market where companies attributed 40% of May’s announced job cuts to AI even as payrolls grew by 172,000. Jed Kolko of the Peterson Institute offered one reconciliation: a CEO can more proudly blame AI for a layoff than admit to over-hiring after the pandemic. A New York Times Magazine expert panel published the same day reached a similar destination, with former White House AI adviser Dean Ball concluding, “We need better empirical economic data. You can’t create policy remedies for a problem you don’t understand.”

Source: Wall Street Journal | Related: New York Times Magazine

Why it matters: Executives are attaching AI rationales to headcount decisions faster than economists can verify them, and the experts themselves cannot agree on the direction of the net effect. Before an AI justification reaches a press release or a WARN notice, workforce leaders should pressure-test whether the technology actually changed the work.

Policy and Government

Anthropic Commits $350 Million to AI Economic Research as Amodei Proposes Unemployment Response Tiers

Anthropic announced a $200 million Economic Futures Research Fund on Wednesday to back empirical research, policy trials, and program evaluation on how AI reshapes labor markets, alongside a $150 million national fellowship program aimed at helping early-career professionals extend AI’s benefits beyond coastal tech hubs. CEO Dario Amodei published a companion essay proposing a tiered government policy framework keyed to three unemployment scenarios: 5%, 10%, and an unspecified “unprecedented” level. The lower tiers lean on wage insurance, capital accounts, tax incentives, and an expanded social safety net; the unprecedented tier contemplates basic income, sovereign wealth models, and equity-sharing mechanisms. The framework also calls for government authority to block deployment of models posing significant risk of catastrophic harm.

Source: Associated Press | Anthropic policy framework (PDF)

Why it matters: An AI lab is now drafting the displacement safety net before Congress has produced one, and tying its proposals to specific unemployment thresholds that can be tracked. Watch whether the research fund produces evaluations policymakers actually adopt, and whether other labs match the commitment or let Anthropic carry the policy conversation alone.

Reskilling and Education

Ipsos: A Third of Workers Have Not Touched AI on the Job, and Workers Over 55 Are Sitting Out at Twice the Rate

New Ipsos Consumer Tracker data, published June 8, finds 38% of full- and part-time workers are not using AI at work at all, with a sharp generational split: 26% of workers under 35 abstain, against 59% of workers over 55. Among those who do use the tools, sentiment leans positive but stays unsettled. 44% say AI makes them more productive, 38% say the tools are a good start but not ready to produce finished work, and 34% say AI cuts time spent on tasks they dislike. One in five report that time saved on some tasks just means more tasks assigned. One in ten worry their employer wants to replace them with AI.

Source: Ipsos Consumer Tracker

Why it matters: Adoption mandates assume a workforce that is already on board, and this data shows a third sitting out entirely, concentrated among the most experienced employees. Training programs that treat the 55-plus cohort as an afterthought will widen the adoption gap rather than close it.

What Workforce Leaders Are Watching

• If 16 of the most-cited economists in the field cannot agree whether AI adds or eliminates jobs, what evidence standard should govern the workforce planning assumptions going into 2027 budgets?

• Anthropic pegged its policy framework to unemployment thresholds of 5% and 10%. Which internal indicators would tell an HR leader that displacement is accelerating inside the company before any national number moves?

• Workers over 55 are opting out of AI at more than twice the rate of younger colleagues. Does the training design treat them as a distinct audience with distinct incentives, or assume one program fits all?

• Both expert panels this week ended on the same demand for better measurement. What would it take to instrument a single organization’s AI impact on roles, tasks, and headcount, ahead of any federal reporting requirement?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

New Cognizant research, presented at Fortune’s COO Summit on Sunday, finds that 93% of jobs are already touched by AI disruption, six years ahead of schedule, and executives are now drawing hard lines about which decisions AI will not make on their behalf. BCG’s annual AI at Work survey, covering nearly 12,000 workers, documents the productivity time that has materialized: 42% of regular AI users save a full workday per week. The gap is in what organizations do with it. Separately, workers are learning that the AI budget is coming out of their paychecks, as companies including Teradata and TTEC suspended raises and 401(k) matches to redirect capital toward AI infrastructure.

Subscribe now

By the Numbers

• 93% of jobs are now impacted by AI, according to Cognizant’s updated New Work, New World analysis, six years ahead of the firm’s 2032 projection.

• 30% of jobs now face existential change, up sharply from Cognizant’s prior assessment.

• 74% of frontline employees are now regular AI users, up 23 percentage points year-over-year, per BCG’s fourth annual AI at Work survey of 11,749 workers across 14 markets.

• 42% of regular AI users save a full workday or more per week; 66% say they received little or no guidance on how to use that time strategically, per BCG.

• More than half of 866 business leaders surveyed plan to cut employee compensation, including bonuses, raises, and equity, to fund AI investments, per a Resume Builder survey cited in Fortune.

Layoffs and Company Decisions

Cognizant: AI Disruption Arrived Six Years Early. Executives Are Drawing New Lines.

New research from Cognizant, presented at Fortune’s COO Summit last week, finds that 93% of jobs are already impacted by AI and 30% face existential change, reaching thresholds the firm had not projected until 2032. Even construction, previously considered safe, now carries a 12% AI exposure score, up from 4% three years ago. C-suite roles score 60%, up from 25%. At the same summit, Cisco Chief People Officer Francine Katsoudas described a pilot the company killed: an AI agent designed to diagnose workplace conflict from employee communications. The decision to cancel it was deliberate. “We believe that’s a conversation a leader needs to have,” she said. Lattice CEO Sarah Franklin drew the same line from a governance angle, arguing that organizations cannot simply let any AI into their workflows without accountability for what it does there.

Source: Fortune, June 8, 2026

Why it matters: The pace of disruption is outrunning most workforce planning cycles. The Cognizant data challenges assumptions that blue-collar roles have a meaningful runway, and the Cisco/Lattice responses illustrate what governance actually looks like in practice: killing a tool that works because it crosses a human-judgment line. HR leaders need explicit frameworks for where AI will not be used, not just where it will.

BCG: The Time Is There. The Strategy Is Not.

BCG’s fourth annual AI at Work global survey, covering 11,749 frontline employees across 14 markets, finds 74% of workers are now regular AI users, and 42% save a full workday or more per week. The problem is what happens to that time. Two-thirds of those workers say their employers gave them little or no guidance on how to use the reclaimed hours. Half are not redirecting it toward higher-value work. BCG’s David Martin, who leads the firm’s People and Organization practice, told Fortune the root cause is straightforward: leaders are not communicating clearly why AI should be used or toward what end. “Senior leaders are really struggling to articulate what the vision and strategy is on AI,” he said. The data also surfaces what BCG calls a “Joy Paradox”: 67% of regular AI users report higher job satisfaction, but 41% also report increased cognitive load. The report also marks the decline of tokenmaxxing: companies that required employees to hit AI usage metrics are now pulling those incentives as compute costs have spiked without commensurate productivity returns.

Source: Fortune / BCG, June 3-5, 2026

Why it matters: The productivity gap is no longer a technology problem. BCG’s data shows companies that pair AI deployment with clear strategy see 25-percentage-point higher business impact than those that issue mandates without direction. Workforce leaders who have rolled out tools without a reinvestment plan for the time those tools free up are leaving the most measurable gains on the table.

Workers Are Paying for the AI Race Out of Their Own Compensation

Teradata told its 5,100 employees in January that there would be no annual salary increases in 2026. The reason, per CEO Steve McMillan’s internal memo: the company is reallocating the raise budget to AI investments. TTEC went further, suspending 401(k) matching for its 15,000 U.S.-based employees through year-end to fund AI certifications and automation tools. A separate Resume Builder survey of 866 business leaders found that more than half plan to cut some form of employee compensation, including bonuses, equity grants, and raises, to redirect capital toward AI. Chief career advisor Stacie Haller told Fortune the pattern reflects companies racing without a clear destination: “They have really no idea what they’re going to need in a workforce afterwards.” Workers who take the cut but see no productivity upside or role clarity at the end of it are making a rational decision to leave. In a low-hire, low-fire market where employers count on job-hugging to absorb the friction, that may not show up immediately. It will.

Source: Fortune, June 6, 2026

Why it matters: Compensation cuts to fund AI spending is a bet that workers won’t notice or won’t leave. Both are wrong assumptions, especially for high performers with options. HR leaders who allow this tradeoff without a clear roadmap for what workers get on the other side are accelerating the very attrition they are trying to manage around.

What Workforce Leaders Are Watching

• When Cognizant’s revised exposure scores put C-suite roles at 60%, how does that change the board-level conversation about AI governance accountability, and who owns the decision about where AI is not allowed to operate?

• BCG’s data shows 66% of AI users get no guidance on what to do with the time AI saves them. Which organizations have actually answered that question, and what does a reinvestment framework look like in practice?

• If more than half of companies are cutting raises and benefits to fund AI spending, what is the retention exposure among workers who absorb the compensation reduction but see no corresponding investment in their own development or role clarity?

• Cisco killed an AI conflict-analysis tool because the output required a human judgment call. How are other large employers building that same discipline, and what criteria determine when AI insight crosses into a decision that belongs to a manager?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

The May jobs report beat every forecast on Friday, and the federal government spent the weekend proposing to take equity stakes in the AI companies reshaping that same labor market. Separately, Congress dropped a 269-page AI governance draft that would freeze state worker protections for three years, and three major labor unions rejected it within hours. IBM reported reaching 22 million learners in its AI training programs. Four distinct developments. None of them point in the same direction.

Subscribe now

By the Numbers

• 172,000: nonfarm payroll jobs added in May, more than double the 80,000 consensus forecast (BLS)

• 107,000: financial activities jobs lost since the sector’s May 2025 peak, including 22,000 in May alone (BLS)

• 524,000: increase in long-term unemployed workers (27+ weeks jobless) over the past year (BLS)

• 40%: share of May 2026 announced job cuts attributed to AI by the cutting companies, the highest monthly share ever recorded (Challenger Gray)

• 87,714: AI-attributed job cut announcements in the first five months of 2026, more than the total for all of 2025 (Challenger Gray)

• 22 million: learners IBM has reached toward its 30 million by 2030 target; company tripling entry-level U.S. hires in 2026 (IBM/Fortune)

Layoffs and Company Decisions

The May Jobs Report Looked Strong. The Details Did Not.

The Bureau of Labor Statistics reported 172,000 nonfarm payroll jobs added in May, well above the 80,000 Wall Street forecast and roughly in line with April’s revised 179,000. Leisure and hospitality led gains with 70,000 additions, followed by local government (55,000) and health care (35,000). The unemployment rate held at 4.3 percent.

Beneath that, financial activities shed 22,000 jobs in May and is down 107,000 from its May 2025 peak. Transportation and warehousing is down 92,000 from its February 2025 peak. Long-term unemployment rose by 524,000 over the past year; those jobless 27 weeks or more now account for 27.5 percent of all unemployed people, a figure that keeps climbing. Separately, Challenger, Gray & Christmas reported that employers cited AI as the reason for 40 percent of May job cuts, the highest monthly share ever recorded and up from 7 percent in January. AI-attributed cut announcements in the first five months of 2026 already exceed the full-year 2025 total.

The two data sets do not contradict each other. Layoff announcements are not the same as jobs lost in the payroll survey, and total employment can grow while specific sectors contract. What the combined picture shows is a labor market that is adding jobs in services and government while quietly hollowing out finance, logistics, and back-office functions, precisely the sectors most exposed to AI automation. Workers who lose those roles are staying unemployed longer.

Sources: BLS Employment Situation Summary, June 5, 2026; Challenger, Gray & Christmas Job Cut Report, May 2026; CNBC, June 5, 2026

Why it matters: Long-term unemployment rising by half a million in a year is not a rounding error. Those workers are concentrated in sectors contracting under automation pressure, and the retraining infrastructure for them remains thin. The headline employment number keeps organizational leaders from feeling urgency about a displacement problem that is already accumulating.

The White House Wants Equity in the AI Companies Reshaping the Workforce

The Trump administration is actively pursuing government equity stakes in OpenAI, Anthropic, and xAI as part of a broader effort to anchor U.S. competitiveness in frontier AI. President Trump endorsed the approach publicly; Sam Altman privately pitched a version of the arrangement before a planned White House meeting with tech executives. OpenAI is projecting a $14 billion net loss in 2026 even as Anthropic approaches $47 billion in annualized revenue.

The arrangement is framed in national competitiveness terms, not workforce terms. No reporting has identified worker protection, displacement mitigation, or reskilling investment as conditions attached to any equity stake discussion. The government would be a financial stakeholder in companies whose products are simultaneously the leading stated cause of U.S. job cut announcements.

Sources: Washington Post, June 5, 2026; Fortune, June 5-6, 2026

Why it matters: When the federal government holds equity in frontier AI companies, its financial interest aligns with those companies’ commercial success, not with their workforce impact. HR leaders and workforce policy advocates should track whether any conditions on worker protection are ever attached to these arrangements, because none have been proposed yet.

Policy and Government

Congress Dropped a Federal AI Bill. Unions Said No Within Hours.

Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page discussion draft of the Great American Artificial Intelligence Act on June 4. The bill creates a federal AI governance framework, establishes a Center for AI Standards and Innovation with $100 million per year in funding, and requires large frontier AI developers to publish risk plans before model releases. On workforce, it directs the Labor Department to convene an AI Workforce Research Hub and includes layoff disclosure requirements.

The bill’s central controversy is a three-year preemption of state laws specifically governing AI model development. States would retain authority over AI deployment and use, but any state law targeting how AI is built would be frozen. AFT President Randi Weingarten, AFA-CWA President Sara Nelson, and the AFL-CIO issued a joint “hard no” statement the same day, calling the bill “a giveaway to the AI industry and a handful of trillion-dollar companies, at the expense of American workers.” Public Citizen called it “a disastrous proposal that Big Tech is celebrating.” The bill landed to near-universal rejection from labor, consumer advocates, and a formal House Democratic commission. Brad Carson of Americans for Responsible Innovation said the preemption provision was “a generational mistake” that converts the current state-law floor into a federal ceiling.

Sources: Axios, June 4, 2026; AFT Press Release, June 4, 2026; Public Citizen, June 5, 2026

Why it matters: Nineteen states have enacted AI employment laws; dozens more are advancing them. A three-year federal freeze would eliminate the most active regulatory layer protecting workers from algorithmic management, biometric surveillance, and AI-driven termination. The workforce provisions in the bill are real but thin relative to what the preemption clause removes. Leaders in multistate organizations need to watch this closely; if the bill advances, the compliance patchwork they have been building would be reset.

Subscribe now

Reskilling and Education

IBM Has Reached 22 Million Learners. The Finding That Matters Is Not the Number.

IBM VP and Chief Impact Officer Justina Nixon-Saintil reported that IBM has reached 22 million learners through its SkillsBuild and related programs, on track toward the company’s 30 million by 2030 goal. IBM is also tripling its entry-level U.S. hires in 2026 and has launched an AI Builders Challenge giving students hands-on development experience. IBM Bob, the company’s AI assistant, is now accessible to 20,000 educational institutions.

The data point Nixon-Saintil flagged that deserves more attention: 67 percent of executives surveyed said the biggest barrier to AI adoption inside their organizations is mindset, not skills. IBM’s own research identifies the internal belief problem as larger than the technical training gap. The company’s response has been to design programs that combine skill-building with framing shifts rather than treating the two as separate interventions.

Source: Fortune, June 3, 2026

Why it matters: If 67 percent of executives cite mindset as the primary barrier, skills programs that ignore the belief dimension will underdeliver. Organizations running AI training programs only will hit a ceiling. The leaders who figure out how to change what employees believe about their own role in an AI-augmented workplace will get more out of the same training investment.

What Workforce Leaders Are Watching

• Long-term unemployment is up 524,000 in a year, concentrated in sectors under the most automation pressure. Which organizations in those sectors have a formal off-ramp for workers who cannot quickly re-enter, and which are simply letting the BLS number accumulate?

• The federal government may soon hold equity in OpenAI, Anthropic, and xAI. If that happens, what changes about the political calculus on federal AI worker protection legislation? Financial interest rarely aligns with regulatory ambition.

• The Great American AI Act’s workforce provisions are modest and its preemption clause is sweeping. For HR leaders in multistate operations who built compliance programs around state AI employment laws: what is the contingency plan if those laws are federally frozen for three years starting in 2027?

• IBM’s finding that 67 percent of executives name mindset, not skills, as the primary AI adoption barrier raises a harder question: what does a change management program for AI adoption actually look like at scale, and who in most organizations owns that work?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

The same day employers announced a record number of AI-driven layoffs, private-sector hiring came in stronger than expected. Those two facts are not in conflict. The Challenger data captures announced intent; the ADP data captures actual hires. What today shows is a labor market in which the displacement story and the resilience story are both true simultaneously, just running in different industries and at different career stages. The companies citing AI for cuts are concentrated in tech and fintech. The companies doing the hiring are in healthcare, trade, and construction. Separately, Congress dropped its most substantive federal AI bill to date, a 269-page bipartisan draft that would freeze state AI laws for three years and create a Labor Department hub to track workforce impacts.

Subscribe now

By the Numbers

• 97,006 job cuts announced in May 2026, the highest May total since 2020, up 16% (Challenger, Gray & Christmas / Bloomberg)

• 40% of all May cuts attributed to AI, the highest monthly share since Challenger began tracking in 2023.

• 38,242 tech sector cuts in May alone, the sector’s highest single month since August 2024, accounting for 123,653 tech cuts year-to-date, up 66% from this point in 2025.

• 87,714 total AI-attributed cuts announced so far in 2026, already surpassing the 54,836 attributed to AI in all of 2025.

• 122,000 private-sector jobs added in May, beating the 110,000 forecast and the strongest monthly total since January 2025.

Layoffs and Company Decisions

AI Is Now the Leading Stated Reason for U.S. Job Cuts, and It Is Not Close

Employers announced 97,006 job cuts in May, the highest May total since the pandemic, according to Challenger, Gray & Christmas. AI was cited as the reason in 38,579 of those cuts, 40% of the monthly total, the highest share since the firm began tracking AI as a layoff rationale in 2023. For the year, AI has been named in 87,714 announced cuts, already exceeding everything attributed to the reason across all of 2025. The tech sector led, announcing 38,242 cuts in May, its worst single month since August 2024. Andy Challenger, the firm’s chief revenue officer, put it directly: “AI is now the leading reason companies give for cutting jobs and the primary industry citing it is Technology.” He added that cuts tied to acquisitions and mergers are also climbing sharply, which he reads as companies repositioning for an AI-driven economy through structural consolidation, not just headcount reduction.

Source: Bloomberg, June 4, 2026; Challenger, Gray & Christmas May 2026 Report, released June 4, 2026.

Why it matters: The share of cuts attributed to AI jumped from 7% in January to 40% in May, a trajectory that tells you this is not noise. HR and workforce leaders who are still treating AI displacement as a future scenario are already behind the data. The harder question now is how to distinguish structural workforce change from AI-washing, and the Challenger methodology cannot answer that.

The Macro Labor Market Is Stabilizing. The Sectoral Story Is More Complicated.

Private employers added 122,000 jobs in May, ADP reported Wednesday, beating the 110,000 economists forecast and marking the strongest month of hiring since January 2025. Eight of the ten sectors ADP tracks posted gains. “Hiring was more broad-based in May than we’ve seen in the last few years,” ADP chief economist Nela Richardson said. Skanda Amarnath of Employ America told Fortune the pickup reflects two forces largely unrelated to AI: years of suppressed hiring finally reversing, and the easing of the immigration headwind from 2025. One sector bucked the trend. Information, ADP’s classification covering software publishing, data processing, and telecommunications, shed 9,000 jobs in May, the steepest decline of any industry, and workers who kept those jobs received the slowest wage growth in the economy at 4.0%. The job-openings data from Tuesday told a similar two-track story: total openings climbed to 7.6 million in April, but hires actually fell and quits were flat, the low-hire, low-fire pattern that has defined the market for over a year.

Source: Fortune, June 3, 2026; Bloomberg, June 3, 2026.

Why it matters: A stabilizing headline number masks a real divergence: healthcare and trades are hiring, tech and information are shedding. Workers displaced from the information sector are not easily reabsorbed into the sectors that are growing. Workforce leaders should treat the macro headline as context, not comfort.

Subscribe now

Policy and Government

Congress Drops Its Most Substantive Federal AI Bill, With a Catch for State Protections

Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft Thursday called the Great American Artificial Intelligence Act. The bill is the most detailed federal AI framework yet to surface in Congress. Its workforce provisions include grants for AI-literacy curriculum, scholarships for students studying the technology, and a requirement for the Labor Department to convene an AI Workforce Research Hub to research and evaluate the technology’s impact on the workforce, including the experience of affected workers. It also contains new disclosure requirements in layoff notices tied to AI decisions. The bill’s most contested provision would preempt state AI development laws for three years. Worker advocacy groups, including Public Citizen and Americans for Responsible Innovation, pushed back immediately, arguing the preemption strips the state-level protections many workers currently rely on. The bill is a discussion draft, meaning Congress is soliciting public feedback before formal introduction.

Source: Roll Call, June 4, 2026. [See source note below.]

Why it matters: A federal AI framework with a Labor Department workforce hub and layoff-disclosure requirements is exactly what HR and institutional designers have been asking for. The three-year preemption of state laws is the trade-off: federal floors replace the state patchwork, but for employers who have been hiding behind regulatory ambiguity, the clock is now running.

Reskilling and Education

For Over a Century, More School Meant Better Jobs. AI May Break That Bargain.

Harvard economists Claudia Goldin and Lawrence Katz documented what they called the race between education and technology: as machines replaced farm and factory work, demand grew for educated knowledge workers, and each generation stayed in school longer than the last. AI may reverse that dynamic. Unlike prior technologies, it reaches into white-collar work, not around it. A new analysis in Chalkbeat reviewed the evidence and found the picture genuinely uncertain: a recent NBER working paper showed that with AI assistance, the gap in problem-solving performance between high- and less-educated workers shrank substantially. If AI can close the performance gap, the wage premium that justified a four-year degree comes under real pressure. Chalkbeat’s Matt Barnum notes that while some claim AI has already decimated entry-level graduate prospects, the weight of current data does not support that narrative fully, but the direction of change leaves students preparing for a future no one can accurately describe.

Source: Chalkbeat, June 4, 2026. [See source note below.]

Why it matters: If the credential-to-job pipeline weakens, the reskilling programs organizations are building need to answer a harder question than “how do we add AI skills?” They need to answer whether the education system that produces the workforce they plan to develop is still producing workers with the right baseline. That question does not yet have an answer.

What Workforce Leaders Are Watching

• When 40% of announced layoffs cite AI as the reason, at what point does HR need a formal methodology to distinguish legitimate workforce redesign from AI-washing in its own organization’s decisions?

• The Great American AI Act’s Labor Dept Workforce Research Hub would, if enacted, create the most detailed federal dataset yet on AI’s workforce impact. How should employers with active AI deployment programs prepare now for the disclosure requirements that are likely to follow?

• If ADP’s stabilizing macro data masks a deep sectoral split between trades/healthcare hiring and tech/information shedding, which workforce planning assumptions in technology-heavy organizations are built on aggregate numbers that no longer describe what is actually happening?

• If a college education’s economic return is under pressure from AI closing performance gaps between educated and less-educated workers, what does that mean for talent development programs built around degree requirements as a hiring filter?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.

TL;DR: Ninety-six percent of organizations are running AI agents. Only 21% have mature governance over them. The gap between those two numbers is where the next generation of organizational failures will be born. When an AI agent acts and something goes wrong, most companies have no clear answer to the most basic management question: who is responsible? The courts are starting to answer it for them.

Subscribe now

There is a management question so foundational that most leaders have never had to think about it consciously: when something goes wrong, who owns it?

In a human workforce, the answer is embedded in the org chart. An employee makes an error; a manager is accountable for the error; a system exists to investigate, correct, and prevent recurrence. The accountability chain is imperfect, and often contested, but it exists. Organizations have spent a century building it. It is the scaffolding under every performance conversation, every incident review, every compliance audit.

Now introduce an agent that schedules meetings, screens applicants, drafts contracts, flags supplier risks, initiates procurement workflows, and responds to customers on the company’s behalf. Ask the same question: when this agent makes a mistake, who owns it? Most large organizations do not have an answer.

In February 2024, the British Columbia Civil Resolution Tribunal ruled against Air Canada after its chatbot gave a passenger incorrect information about bereavement fare refund eligibility. Air Canada argued, with apparent seriousness, that the chatbot was a “separate legal entity” responsible for its own actions. The Tribunal rejected this in terms that should be posted in every enterprise AI deployment room: the company bears responsibility for all information on it shares with customers, regardless of whether a human or an algorithm produced it.

In March 2026, Nippon Life sued OpenAI after ChatGPT helped a claimant draft 44 post-settlement legal filings, including at least one with a fabricated case citation. The fabricated filings cost Nippon Life approximately $300,000 in legal fees to respond to. The lawsuit is not primarily about the hallucination. It is about who bears the cost when AI outputs cause harm in a legal proceeding.

The cases are not isolated. Edelson PC has filed wrongful death and personal injury suits against OpenAI and Google across at least a dozen cases. HSB introduced AI Liability Insurance in March 2026, which is notable not because insurance solves the problem, but because insurance companies do not invent products for risks that do not exist.

What the cases have in common is not the severity of the harm. It is the accountability vacuum underneath each one. The AI acted. Something went wrong. The company had no clear owner of the outcome. Courts are filling the gap the companies left open.

The data on organizational readiness is not ambiguous. According to a 2026 analysis aggregated across enterprise AI deployments, 96% of organizations are already running AI agents. Only 21% have a mature governance model for autonomous AI agents. That means roughly three out of every four companies running agents cannot tell you, with confidence: who approved this agent’s scope of authority? Who monitors its outputs? Who is alerted when it acts outside expected parameters? Who is called when something goes wrong?

The Deloitte 2026 State of AI in the Enterprise report is specific about what mature governance requires and how many organizations have it: approximately 80% lack clear boundaries defining which decisions agents can make independently versus which require human approval; real-time monitoring that tracks agent behavior and flags anomalies; and audit trails capturing the full chain of agent actions. Those three requirements are not exotic. They are the minimum conditions for managing anything consequential in an organization. We apply them to financial transactions, to regulatory filings, to clinical decisions. We are not applying them to the agents we are giving authority to act.

An Accenture and Wharton joint report published in March 2026 stated the problem cleanly: “Intelligence may be scalable, but accountability is not.” The researchers found that more than 50% of working hours across the American economy are now subject to reshaping by AI agents. The accountability structures most organizations rely on were designed before any of that was true.

Here is the mechanism by which this becomes expensive. AI agents do not make isolated errors. They make errors inside workflows, and workflows are connected. A Roborhythms analysis of 2026 enterprise AI deployments describes the cascade pattern: an AI agent mislabels a supplier’s risk rating, which triggers an automated contract review flag, which pauses a procurement approval, which delays a product launch. No single point of failure. A chain of automated reactions, each one reasonable given the preceding step, the whole chain resting on an original error that no one caught because no one was watching.

More than half of all agents deployed in enterprise environments run without security oversight or logging. Only 24.4% of organizations have full visibility into which AI agents are communicating with each other. Twenty-eight percent of US firms report zero confidence in the data quality feeding their agents, which means those agents are reasoning correctly over wrong inputs and producing confidently wrong outputs.

This is not a technology problem. The technology is doing what it was deployed to do. The problem is a management design problem. Specifically: the absence of the accountability infrastructure that organizations build around every other consequential process.

I spent nearly a decade building workforce and organizational design functions from the ground up. The pattern I saw repeatedly was this: new capabilities get deployed faster than the management systems that surround them. The capability is visible. The governance gap is invisible until it becomes a crisis. AI agents are following the same pattern, at a much larger scale and with much faster deployment timelines.

Subscribe now

In post-07, I wrote about the governance gap between “we have agents” and “we manage agents.” The data has sharpened since then. What was a gap is becoming a liability, in the literal sense: the EU AI Act’s full enforcement begins August 2, 2026. High-risk AI systems used in employment, credit decisions, education, and similar contexts are subject to mandatory risk management, human oversight mechanisms, and audit trails. Organizations deploying agents in those domains that have not built governance infrastructure are not just operating carelessly. Starting this August, in the EU, they are operating unlawfully.

The EU AI Act approach is notable because it assigns liability to the deployer, not the developer. Companies that deploy AI systems bear responsibility for their operation. The “our vendor’s AI did it” defense has the same legal credibility as Air Canada’s “the chatbot is a separate entity” argument. Courts and regulators are not interested in the internal supply chain of the AI product. They are interested in who decided to deploy it and who was supposed to be watching. Third-party risk management just took on a lot more weight too.

The Accenture-Wharton framing is useful here: accountability is not scalable. You cannot apply more accountability by simply deploying more agents. Every agent added to an organization’s workforce is an additional node of potential error, an additional actor whose outputs need someone responsible for reviewing them. The governance cost grows with every deployment. Most organizations have not accounted for that cost.

The organizations that are getting this right share a specific discipline. Before deploying any agent with consequential authority, they answer four questions: What decisions can this agent make without human approval? Who monitors its outputs and on what cadence? Who is notified when it acts outside expected parameters? Who owns the incident response when something goes wrong?

Those four questions are not technically complex. They are organizationally complex because answering them requires cross-functional alignment among technology, legal, compliance, operations, and business leadership before the agent goes live, not after. Most organizations reverse that sequence. They deploy first, build governance when something breaks.

Forty percent of agentic AI projects were canceled or paused as of February 2026, with governance friction cited as one of the top blockers. The lesson most organizations are drawing from this is that governance slows deployment. The correct lesson is that deployment without governance is not deployment. It is exposure.

The Air Canada ruling was about $812. The Nippon Life suit is about $300,000. The next case may be about a financial transaction an agent initiated, a hiring decision an agent influenced, a contract an agent approved. The scale of the error will match the scope of the authority organizations have handed to agents. That scope is expanding every week.

The question is not whether AI agents will make mistakes. Every workforce makes mistakes. The question is whether the organization that deployed the agent has built the accountability infrastructure to catch errors early, assign responsibility clearly, and correct course before a manageable problem becomes a lawsuit. Most have not.

Here’s How You Take Action

1. Run the accountability audit. For every AI agent currently operating in your organization with consequential authority, ask the four questions: What can it decide without approval? Who monitors its outputs? Who gets alerted when it acts anomalously? Who owns incident response? If you cannot answer all four for a given agent, that agent is ungoverned.

2. Separate deployment authority from governance authority. The team that builds or procures an agent should not be the team that approves its governance model. Conflating the two creates a structural incentive to undercount risk. Someone outside the deployment team needs to sign off that the governance infrastructure is in place before the agent goes live.

3. Map your EU AI Act exposure by August 2. If your organization operates in the EU or deploys AI in employment, credit, education, or similar high-risk domains, full enforcement begins August 2, 2026. Start with which agents touch those domains and work backward through the compliance requirements.

4. Build the incident response protocol before you need it. The organizations best positioned when an AI error escalates are the ones that decided in advance how to respond. Who is notified? Who investigates? Who communicates externally? What is the remediation path? Don’t wait until an incident’s post-mortem to document this.

5. Challenge the deployment velocity. Every new agent added to your organization’s workforce adds governance cost. If your procurement or technology teams are deploying agents at a rate that outpaces your governance infrastructure’s ability to absorb them, the gap is widening, not closing. Deployment velocity is a governance risk metric.

Christina Lexa writes about workforce strategy, organizational design, and the human infrastructure that organizations build or fail to build around new capabilities. Workforce Rewired publishes weekly at the intersection of AI, organizational design, and the future of work. If this piece made you think, share it with someone navigating the same questions.

The views expressed here are my own and do not represent the position of my employer or any organization I am affiliated with.

Editor’s note: I really hate the word “tokenmaxxing”.

The Federal Reserve Bank of New York published new research this week finding that remote work, not AI, drove most of the rise in entry-level unemployment since the pandemic. The finding lands the same week Apollo’s chief economist declared “zero evidence” AI has killed jobs at all, and Cognizant’s CEO announced he is hiring 24,000 graduates in 2026 while competitors shrink their junior pipelines. Separately, COOs at Okta and Cisco publicly disagreed at the Fortune COO Summit about whether AI agents belong in an org chart or a workflow, exposing how far most organizations still are from answering that question with any governance behind it.

Subscribe now

By the Numbers

• Two-thirds of the post-pandemic rise in unemployment among young college graduates is explained by remote work, not AI exposure, according to new Federal Reserve Bank of New York research.

• The unemployment rate for workers ages 20-24 has fallen every month year-over-year in 2026 after rising steadily the previous two years, as companies resume entry-level hiring.

• Cognizant plans to hire 24,000 to 25,000 fresh graduates in 2026, up 20% from 2025, bucking the sector-wide trend of junior role elimination.

• Apollo chief economist Torsten Slok cites ADP employment data showing AI spending is growing employment, not shrinking it, even as Goldman Sachs separately tracked 21,900 AI-attributed U.S. layoffs in April 2026 alone.

Layoffs and Company Decisions

Apollo’s Chief Economist Says There Is “Zero Evidence” AI Is Killing Jobs

Torsten Slok, chief economist at Apollo Global Management, published a direct rebuttal to the dominant displacement narrative this week, citing ADP employment data as evidence that AI is creating work, not eliminating it. Slok argues the Jevons paradox is playing out in real time: cheaper AI tools generate more demand for implementation, integration, and oversight roles. His claim lands amid data showing AI-attributed layoffs are rising month over month, with Goldman Sachs tracking 21,900 in April alone. Critics, including Forrester, note that companies may be using AI as cover for cost cuts they would have made anyway. The debate has not resolved. What has shifted is that a credentialed mainstream economist is now publicly willing to call the displacement story unsupported by aggregate employment data.

Source: Fortune, June 1, 2026

Why it matters: HR leaders defending headcount decisions in boardrooms need to know this argument is gaining traction among economists. The “AI made us do it” framing is under scrutiny from multiple directions. Leaders who deploy it without substance behind it are taking on reputational risk they may not have priced in.

Fortune COO Summit: Executives Can’t Agree Whether AI Agents Are Colleagues or Software

At Fortune’s COO Summit in Scottsdale this week, Okta President and COO Eric Kelleher and Cisco’s Chief People Officer Francine Katsoudas offered 180-degree disagreements on a question that turns out not to be philosophical: how to classify AI agents in your organizational structure. Kelleher has named his agents Leo, Sloan, Hank, and Walker; they appear in business reviews alongside human staff. He argues that naming agents transforms the relationship from tool to colleague, and that the framing shift changes how managers engage. Katsoudas countered that agents belong in workflows, not org charts, and that treating them as colleagues obscures accountability. Both are describing the same underlying problem: most organizations learned to experiment with AI long ago and are now stuck, unable to redesign the actual work around it. Kelleher’s proposed intervention is blunt: push token budgets down to people managers, forcing them to make concrete tradeoffs between human and AI labor allocation. Katsoudas is focused instead on ensuring that humans remain the named accountable actors, particularly where consequential decisions occur.

Source: Fortune, June 2, 2026

Why it matters: The framing question has structural consequences. How an organization classifies AI agents determines who owns their outputs, who is accountable when they err, and how work gets distributed across human teams. Organizations that defer this classification are not staying neutral; they are making a choice by default.

Subscribe now

Reskilling and Education

Cognizant’s CEO Is Hiring 24,000 Graduates This Year. He Calls AI Tokenmaxxing a “Vanity Metric.”

Ravi Kumar, CEO of Cognizant, appeared at the Fortune COO Summit with a hiring plan that cuts against the prevailing direction in technology services. Cognizant will hire 24,000 to 25,000 fresh graduates in 2026, up 20% from 2025, adding two new entry-level roles built around AI proficiency: Frontier Certified Engineer and Frontier Business Operator. Kumar’s framing centers on “learnability” over experience, and he was direct about the industry trend he is pushing back on: measuring AI performance by token consumption is a “vanity metric” that distracts from actual business outcomes. The strategy is a deliberate pyramid-broadening play, positioning Cognizant to absorb market share as competitors narrow their junior hiring pipelines. Kumar also addressed the middle-manager question, arguing AI is remaking that tier into “player-coaches” who both execute and develop the people below them.

Source: Fortune, June 1, 2026

Why it matters: Every competitor cutting junior headcount is handing Cognizant a pipeline advantage. Workforce leaders should track whether Kumar’s bet on entry-level hiring yields a talent and capability edge in 18-24 months, or exposes Cognizant to margin pressure when clients expect leaner delivery models.

New York Fed: Remote Work, Not AI, Drove Most of the Rise in Young Graduate Unemployment

New research from the Federal Reserve Bank of New York, covered by NPR and Bloomberg this week, identifies remote work as the primary driver of rising unemployment among college graduates under 29, not AI. Unemployment in that cohort rose 20% after the pandemic while falling slightly for older graduates. Remote work grew fourfold across that same period. When researchers analyzed a Fortune 500 tech company that shifted to remote work post-pandemic, hiring of young people dropped sharply as managers became reluctant to bring in employees requiring close training and mentoring. AI exposure did not explain the divergence. Bloomberg Opinion columnist Allison Schrager, writing separately on the same research, pointed to an emerging reversal: unemployment for workers 20-24 has declined year-over-year every month in 2026 as return-to-office mandates and post-restructuring hiring pull entry-level candidates back into the market.

Source: NPR / Bloomberg Opinion, June 1-2, 2026

Why it matters: Organizations designing junior talent pipelines around AI risk have been focused on the wrong variable. The research puts a concrete lever back in employers’ hands: in-person and hybrid onboarding structures matter more right now than AI literacy curricula for entry-level hiring outcomes.

What Workforce Leaders Are Watching

• If remote work, not AI, explains most of the junior talent pipeline disruption, which structural decisions around in-person onboarding and hybrid policies do organizations need to revisit before attributing entry-level hiring problems to automation?

• When senior leaders publicly disagree about whether AI agents are colleagues or software, what accountability structures and governance policies are they operating without, and who in the organization is currently answering that question?

• As Apollo declares zero evidence of AI-driven job loss and Goldman tracks record monthly AI layoff figures, organizations citing AI as the driver of cuts face increasing scrutiny from economists, regulators, and employees. What due diligence does that require before the next restructuring announcement?

• Cognizant is expanding junior hiring while competitors contract it. For organizations that have already cut entry-level pipelines, what is the realistic recovery timeline for rebuilding the talent development layer those roles provided?

This briefing was prepared automatically by the Workforce Rewired research assistant. All stories include direct source links.