TGR ISSUE 18
Sections
☽☼
60-Second Read What's New Stephen's Take The Reason Code The New Pyramid The Debate Signal Map Industry Pulse Global & Regulatory Voices Implications Archive
Issue 18 — September 11, 2026

The Great Recomposition

The New Pyramid. A reason code gives you a cause and a number. What it cannot see is that the pyramid is not shrinking, it is growing wings, and where you put them is the last thing a competitor can copy.
H1H2H3H4H5
AI drives the task ⟶ shared agency ⟶ human essential

On 1 October, Connecticut becomes the first state to make employers say on a mass-layoff filing whether the cuts were related to their use of AI. None of the new rules says what qualifies. Three weeks before that, AI dropped out of the top three reasons for US job cuts after leading for five months straight, while payrolls beat expectations threefold. So the attribution becomes a legal document at the exact moment it stopped holding still. That matters less than what it cannot see. A reason code gives you a cause and a number. What is actually happening is a change of shape: the base moving up into orchestration, the middle widening where judgment and closeness to the customer live, and the coordination layer tucking into a waist. Not a smaller pyramid, and not the same one with fewer people in it.

4th
AI's Rank Among Stated Layoff Reasons, August
After Five Straight Months at No. 1. Challenger
3×
Reported AI Productivity vs. What Their Own Numbers Imply
~750 CFOs. Atlanta & Richmond Fed with Duke
0.14
Replacement-to-Enhancement Ratio, Management Roles
Admin Support Is 2.03, the Only Group Above 1.0
+162K
August Payrolls Against a 31K Twelve-Month Average
BLS. Unemployment Steady at 4.1%
StartThe Issue
1 of 13
All sections
Next →60-Second Read
The 60-Second ReadIssue 18 · September 11, 2026 · 52+ institutions
  1. AI led the layoff-reason table for five months, then fell to fourth in one. Challenger recorded 3,462 August cuts attributed to AI, the lowest monthly total since December 2025, ending a run that began in March. Restructuring took the top spot at 16,173. AI still leads year to date at 116,175, about 22% of all cuts. A cause that moves from first to fourth in thirty days, with nothing changing in the technology, is behaving like a label rather than a mechanism.
  2. In nineteen days that label becomes a legal filing, and nobody has defined it. From 1 October, Connecticut SB 5 requires every employer filing a WARN-standard mass-layoff notice to say whether the layoff is related to its use of AI or other technology change. California's Executive Order N-6-26 directs agencies to measure first and build a dashboard. The bipartisan GAAIA discussion draft would add a similar WARN disclosure at the federal level. None of the three says what counts as caused by AI. Compliance is about to manufacture a data series out of an undefined term.
  3. The same problem shows up inside the company, and you can measure it. A survey of roughly 750 financial executives by the Atlanta and Richmond Feds with Duke found CFOs reporting 1.8% AI-driven labor productivity growth for 2025 against 0.6% implied by their own revenue and employment numbers. For 2026 it is 3.0% against 1.8%. What they believe runs about three times what they can show. We are not measuring AI. We are measuring the story we tell about it.
  4. And the reason that generates the layoff notice is the one with no return. In the same data, productivity gains line up with growth motives: building new or better products, serving customers more effectively. They do not line up with cost reduction, workforce development or capital upgrading, and the coefficient on cutting non-labor costs runs negative. Executives put cutting labor costs dead last among their own reasons for investing, at 2.0 on a 0 to 4 scale. Release is the pathway that shows up in the announcement and not in the results.
  5. What a reason code cannot see is the shape. The same executives expect routine clerical work to fall 2.19 percentage points of the workforce by 2028 while skilled-technical rises 1.35, with large firms shedding clerical share and smaller firms adding technical roles. Their replacement-to-enhancement ratio, built from their own descriptions of what AI does to each occupation, puts Office and Administrative Support at 2.03, the only group above 1.0, against Management at 0.14 and Architecture and Engineering at 0.10. That is not a pyramid getting shorter. That is weight moving out of the base and into the middle.
  6. Which is the argument of this issue. The pyramid survives, but its weight moves. The base tugs up as execution roles move into orchestration, the same people now directing and checking the thing that does the task. The middle widens into wings, where closeness to the customer, real expertise and judgment absorb the hard cases machines hand back. The coordination layer tucks into a waist. Fewer layers, less height, two wide bands. You can buy capacity, so everyone has it. Where you put your people is a decision only you make.
← PreviousCover
2 of 13
All sections
Next →What's New
What's New

Issue 18: The New Pyramid

Two data releases and one legal deadline landed inside three weeks, and together they change what this publication should be measuring. The attribution we have all been reading turns out to be unsteady, is about to become mandatory, and was never able to describe the event anyway.

The Anchor

AI Fell From First to Fourth Among Layoff Reasons in a Single Month

Challenger's August report, published 3 September, records 3,462 job cuts attributed to artificial intelligence, the lowest monthly figure since December 2025, ending a five-month run in which AI was the leading stated cause. Restructuring took first place at 16,173, its highest month since January. Total announced cuts were 52,881, up 58% from July but the lowest August since 2022, with year-to-date announcements of 529,914 running 41% below last year. AI still leads year to date at 116,175, about 22% of the total. Hold both facts. AI is the leading annual reason and it moved four places in thirty days. The second fact is the one worth reasoning about, because nothing in the technology changed between July and August. What changed was the sentence firms chose to write. Twenty of the thirty industries Challenger tracks are running below last year, and Andy Challenger's own read is that employers are now planning to add workers, with 46% of those plans coming from manufacturing.

The Research

CFOs Report Three Times the AI Productivity Their Own Numbers Show

A survey of roughly 750 financial executives, run by the Federal Reserve Banks of Atlanta and Richmond with Duke and published as NBER working paper 34984, asked CFOs to attribute productivity, revenue and employment changes specifically to AI. Reported labor productivity growth from AI came in at 1.8% for 2025. The figure implied by the same respondents' reported revenue and employment was 0.6%. For 2026 the gap holds at 3.0% against 1.8%. The authors read this as a classic productivity paradox with revenue arriving late, and note that each year's reported figure roughly equals the following year's implied figure, which would make the belief early rather than wrong. That reading deserves to stay open. It does not change the operational problem, which is that decisions get made on the reported number. Two other findings from the same data run through this whole issue. Capital deepening explains only about 15% of the productivity gain at the typical company, because most AI spend is an operating expense rather than capital investment. And half of all firms say plainly that AI will replace no roles at all.

The Deadline

On 1 October, "AI Did It" Becomes a Line You File With the State

Connecticut SB 5, signed in May, takes effect on 1 October. Any employer filing a WARN-standard mass-layoff notice has to tell the Department of Labor whether the layoff is related to its use of AI or other technology change. California's Executive Order N-6-26 runs the opposite sequence, putting no immediate obligation on private employers and instead telling state agencies to investigate and build a dashboard. The bipartisan Great American Artificial Intelligence Act discussion draft, released in June, carries a Workforce title that would add a similar federal WARN disclosure where AI is a "substantial factor." None of the three says what counts as caused by AI. Within a year there will be an official series counting AI-related layoffs, and people will read it as measurement. It will actually be a pile of legal judgments made under uncertainty about an undefined term, in the same year that term moved from first to fourth in a month. That is not an argument against disclosure. It is an argument about what the resulting number can carry.

The Mechanism

The Motives That Pay Are the Ones That Never Generate a Layoff Notice

The most consequential result in the Fed data is a set of regressions almost nobody has picked up. Across specifications, the AI investment motives that line up with real revenue-productivity gains are building new or better products and services and reaching or serving customers more effectively. The motives that do not line up, showing imprecise and frequently negative coefficients, are cost reduction, workforce development, capital upgrading and decision support. Cutting non-labor costs runs negative. Executives themselves rank cutting labor costs last among their own stated motives at 2.0 on a 0 to 4 scale, behind production efficiency at 2.9 and labor productivity at 2.7. Because this is a relationship between self-reported variables rather than a level, it is the part of the survey that holds up best against the attribution problem described above. Set against the founding argument of this series, it is the strongest confirmation yet available. The companies getting productivity from AI are the ones moving freed capacity into growth, not the ones releasing it.

The Shape Data

Admin Support Is the Only Occupation Group Where Replacement Beats Improvement

The same executives were asked, in open text, which roles AI replaces and which it improves. Mapped to occupation groups, the ratio of replacement to enhancement mentions is the cleanest task-level picture published this year. Office and administrative support sits at 2.03. Nothing else is above 1.0. Business and financial operations sits at 0.83, balanced, which points to tasks moving around inside jobs rather than jobs going away. Computer and IT 0.60, legal 0.47, production 0.31, sales 0.30, management 0.14, architecture and engineering 0.10. The forward projections agree. Routine clerical work falls 2.19 percentage points of the workforce by 2028 while skilled-technical rises 1.35, with large firms shedding clerical share and smaller firms adding technical roles. That last difference matters more than it looks. It means a real share of the reallocation is happening between companies rather than inside them, which is exactly the movement an aggregate employment number cancels out.

The Counter-Signal

The Labor Market Went the Other Way the Same Week

August payrolls came in at 162,000 against expectations in the mid-fifties and a prior twelve-month average of 31,000, the strongest month in five. Unemployment held at 4.1%, participation rose to 61.6%, and June and July were revised up by a combined 55,000, turning July's reported loss into a gain. Underneath: food services up 59,000, local government education up 42,000, manufacturing up 16,000, and information down 23,000, concentrated in computing infrastructure, data processing and publishing. Unemployment among college graduates held at 2.7%. Corroborating evidence from an unlikely sector arrived the day before. An analysis of regulatory filings found wealth management firms disclosing AI use grew headcount 15% year over year against 8% for non-disclosers, with non-advisory staffing up 14.2%, more than double advisor growth. The adopters are hiring, and hiring disproportionately into support and operations. That is the orchestration wing filling up.

The Adoption Picture

Deployment Is Nearly Universal, Returns Are Not, and the Gap Is Becoming a Workforce Policy

Gallup's Q2 reading puts organizational AI adoption at 47%, up six points in a quarter, with 52% of US workers now using AI in their role and 15% using it daily. Against that, a vendor survey published this month reports 97% of executives deployed AI agents in the past year while only 29% see significant ROI, 79% report adoption problems, and, the number worth pausing on, 92% say they are actively cultivating "AI elite" employees while 60% plan layoffs for non-adopters. That is a company sorting its people into a wing and a waiting room. Disclosure: this survey comes from a company that sells enterprise AI software and has a direct commercial interest in the conclusion that adoption separates winners from losers, so treat the direction as plausible and the magnitudes as marketing. The more defensible version of the same point comes from Strada's survey of nearly 1,500 executives, which found that firms with a coherent company-wide AI plan report increases in entry-level hiring, while firms that only partly integrated AI, using it mainly to automate routine tasks, report reductions. Stopping halfway is the expensive place to stop.

← Previous60-Second Read
3 of 13
All sections
Next →Stephen's Take
Stephen's Take

Reading the Signal Through the Noise

What I got wrong about layoff attribution, why the shrinking-pyramid forecast is the most expensive mistake on the table, and what to do with the capacity you free up.

12 ROI› 13 Agency› 14 Redesign› 15 On-Ramp› 16 Leverage› 17 Teaching› 18 Shape
Stephen Wroblewski

Stephen's Take

The New Pyramid · Why the Shape Is the Strategy

The New Pyramid — Why the Shape Is the Strategy

First, a correction. For several issues I treated AI attribution in layoff announcements as a signal. I quoted the share of cuts citing AI. I quoted the monthly rank. August tells me that instrument is weaker than I thought. A cause does not move from first place to fourth in thirty days when nothing about the technology changed. Some of what we were all reading as adoption was companies picking a better-sounding sentence. I should have caught it sooner. Here is what I think we should read instead.

01 · The Position

The pyramid is not flattening. Its weight is moving.

Most workforce projections I see still draw the same shape with fewer people in it. Cut the base, keep the proportions, book the saving. That is what a headcount model produces, because a headcount model has only one dimension to move. It is also the costliest mistake available right now, and the reversal data is the receipt. About a third of managers who cut a role for AI have hired back for it. Most leaders who ran the play now say it was wrong. A third of the companies that reversed spent more on restaffing than the cut ever saved. Those firms did not misjudge the technology. They shrank a pyramid when the job was to change its shape.

What is actually happening inside companies that are reinventing rather than trimming is a redistribution. The base pulls upward as execution roles are redeployed into orchestration: the same people, now setting up, directing and checking the thing that does the task. The middle widens into a band where judgment, proximity and expertise absorb the variance machines hand back. The coordination layer, whose job was moving information up and instructions down, pulls into a waist. Fewer layers, less overall height, two wide bands, and a detachable edge for peak work. That silhouette is not a pyramid. And this is redeployment, not recruitment, which is exactly why a headcount report shows nothing and a reason code sees nothing.

02 · The Argument

Capacity is purchasable. Wing placement is not.

This is the part I would fight for. Your competitor rents the same models you rent, from the same few providers, at prices that keep converging. The Fed data shows this is now structural rather than temporary: capital investment explains only about 15% of AI productivity gains at the typical company, because most AI spending is subscription and service rather than owned asset. Everyone draws from the same central pool. Capacity is common by construction, and nothing common can set you apart.

What cannot be matched is where you chose to put people. Close to which customers. Close to which risks. Close to which exceptions. Close to which craft. I have taken to calling those bands the wings, and they matter because they are a judgment about where your value actually comes from, written as headcount in a layer rather than as a line in a budget. Two companies with the same technology and the same cost structure can have completely different wings, and you will see the difference in defect rates, in retention, and in what a customer says when something goes wrong. The pyramid was a shape you inherited. The airplane is a shape you choose. The wings also do not land at the same level everywhere. A regulated manufacturer grows them around deviation and qualification calls. A retailer grows them at store and category level. A bank grows them around credit judgment and client continuity. If they landed in the same place for everyone, this would be a template, and templates do not differentiate anybody.

03 · The Decision Nobody Owns

Capacity freed is an enterprise asset. Treat it like one.

Here is where most programs quietly fail, and it has nothing to do with technology. When AI frees capacity in a function, that capacity is almost always treated as belonging to the budget holder who freed it. Finance frees eight thousand hours, so finance decides what happens to those hours. Usually finance gives them back as a cost saving, because that is the only move a single function can be rewarded for. Multiply that across twenty functions and you get a company that spent three years automating and has nothing to show for it except a slightly smaller payroll and the same operating model.

Capacity freed is an enterprise asset, not a functional one. Whether a given block of it gets rotated to similar work, redeployed into new work, or released out of the company is a decision that belongs at the enterprise level, because only the enterprise can see where that capacity is worth the most. The function that freed it cannot make that call well, and should not be asked to. This is not a governance nicety. It is the difference between an automation program and a reinvention.

And the choice has to be made against something specific: your own value equation. Not a benchmark, and not what the industry is doing. Every company creates value in a particular way, out of a particular mix of growth, quality, pace, risk, cost and customer intimacy, weighted differently than its competitors weight them. Freed capacity should flow toward whichever part of that equation is genuinely yours. If your advantage is speed to market, capacity goes into cycle time and the wings form around the decisions that gate it. If your advantage is trust, capacity goes to the people closest to the customer at the moment something breaks. The Fed data makes this concrete rather than philosophical: the AI motives that actually produce measured productivity are building better products and serving customers better, while the cost-reduction motives show weak and often negative results. The companies that convert freed capacity into their own value equation get the return. The ones that hand it back as savings are, on their own numbers, not getting one.

04 · The Caveat

The evidence here is directional, and I will not dress it up.

The Fed CFO paper carries most of this issue, and its method is to ask executives to attribute outcomes to AI themselves. That is a reason code, which is the thing I spend the first half of the issue warning about. I would rather say so than have you find it. Three things keep it usable. The self-attribution is checked against the same respondents' revenue and employment numbers, which is how the gap becomes visible at all. The panel's forecasts track what those firms actually go on to report. And the finding I lean on hardest, which motives pay, is a relationship between variables rather than a level, and relationships survive a uniform reporting bias in a way levels do not. So treat the levels as soft and the relationships as solid. It is also a working paper without peer review. The aggregate employment number people will pull from it, a projected decline of roughly half a percent for 2026, is a model built on self-reported expectations rather than anything that has happened, and I would not put it in a board pack without that sentence attached. The strongest case against this entire issue is still the Budget Lab at Yale, which finds the occupational mix shifting no faster than it did for the PC or the internet. I run it at full strength. My answer is that a steady national mix can sit on top of violent movement inside companies, because the national number averages firms that are moving in opposite directions. The survey shows exactly that, with large firms shedding clerical share while smaller firms add technical roles.

05 · The Move

Draw your shape before somebody makes you file your reason.

In three weeks a Connecticut employer filing a mass-layoff notice has to say whether AI was involved, with no definition of the term to answer against. More states will follow and the federal draft already carries the same clause. The work to do before that lands is not a compliance exercise. It is a design exercise, and it is the same work either way. Name your wings as bands of people rather than as functions. Work out whether your entry level is being redeployed into orchestration or simply removed, because a headcount report shows those as identical and a capability plan shows them as opposites. Find out who absorbs the variance the machines hand back, and whether their time is funded as real work or running on goodwill. Ask what your waist is for once routine coordination automates. Then decide, at the enterprise level and against your own value equation, where the freed capacity goes. Capacity freed is not headcount reduced. This is the first issue where I can point at evidence rather than conviction. A leader who cuts to the reason code books a saving and files a sentence. A leader who changes shape builds something no competitor can buy, and gets no credit for it in the quarter it happens. That second one is still the job.

— Stephen Wroblewski, Managing Director, Accenture Talent & Process Reinvention

← PreviousWhat's New
4 of 13
All sections
Next →The Reason Code
The Centerpiece · Diagnosis

The Reason Code

A reason code gives you a cause and a count. In nineteen days it becomes something you file with a state labor department, and it still will not describe what actually happened to your organization.

The month the reason moved

March to July
No. 1
AI, five months running

The run

For five months straight, AI was the single most-cited reason in US job-cut announcements. It became the default explanation on earnings calls, in press releases, in the trackers that aggregate them, and in a lot of the analysis built on those trackers, including some of ours.

August
4th
3,462 cuts, lowest since Dec 2025

The drop

In one month AI fell to fourth, behind restructuring at 16,173. Nothing changed in the technology between July and August. What changed was the sentence companies decided to write. A cause does not move four places in thirty days. A fashion does.

Two caveats, and both point the same way. First, AI is still the leading year-to-date reason at 116,175 cuts, about 22% of the 2026 total, so this is a collapse in the monthly rate and not a repudiation of the year. Second, August was a light month in a light year: 52,881 cuts, the lowest August since 2022, with year-to-date announcements down 41%. Small denominators move fast. Neither caveat rescues the reason code, because the question was never whether AI causes some cuts. It clearly does. The question is whether the field measures anything steady enough to reason from. For five months it looked like it did. In August it did not.

What the same month did to the other number

The week the reason fell, the labor market went the other way. August payrolls came in at 162,000 against expectations in the mid-fifties and a prior twelve-month average of 31,000. Unemployment held at 4.1%. Participation rose to 61.6%. June and July were revised up by a combined 55,000, which turned July's reported decline into a gain. The composition underneath is the interesting part: food services up 59,000, local government education up 42,000, manufacturing up 16,000, and information down 23,000, concentrated in computing infrastructure, data processing and publishing.

The level held. The mix moved. Those are different events, and only one of them has a reason code.

This is the pattern the series has circled since Issue 11. Total employment is a poor instrument for a recomposition, because a recomposition barely moves the total by definition. Work shifts between categories and the sum hardly notices. The Budget Lab at Yale has made the strongest version of this case all year, finding the US occupational mix shifting no faster than it did when the PC or the internet arrived. We run that at full strength, and this issue agrees with more of it than any issue before. Where we part company is on what follows. A steady mix at national scale is completely compatible with violent redistribution inside individual companies, because the national mix averages across firms moving in opposite directions.

What executives report against what their own numbers show

The most useful new evidence this cycle comes from an unusual place. The Federal Reserve Banks of Atlanta and Richmond, working with Duke, surveyed roughly 750 financial executives between November 2025 and January 2026 and published the results as NBER working paper 34984. It asks CFOs to attribute changes in productivity, revenue and employment specifically to AI, then checks that attribution against their own arithmetic.

2025, what they said
1.8%
Reported AI-driven labor productivity growth

What they believe

What the executive thinks AI did to output per worker. This is the number that reaches the board pack, the investor call, and eventually the reason code on the layoff notice.

2025, what it implies
0.6%
Implied by their own revenue and employment

What they can show

AI-attributed revenue growth minus AI-attributed employment change, from the same person in the same survey. For 2026 the gap holds at 3.0% against 1.8%. Belief runs about three times evidence.

The authors read that gap generously, and they may well be right. Some of it is revenue that has not landed yet and quality improvements that do not show up in a top line. They point out that each year's reported figure roughly equals the next year's implied figure, sector by sector, which would make the belief early rather than wrong. That reading deserves to stay open. It does not fix the practical problem. A three-to-one gap between what leaders believe a technology has done and what their own numbers show is exactly the condition under which a confident, undefined attribution ends up inside a workforce decision.

A note on this paper's method, because it is the issue's own subject

This paper's core technique is asking executives to attribute outcomes to AI themselves. That is a reason code. Its main strength, separating the AI effect from everything else acting on a company at once, is the very mechanism this issue warns about, and it would be dishonest to lean on the findings without saying so. Three things keep it usable. The self-attribution gets checked against each respondent's own revenue and employment numbers, which is how the gap becomes visible instead of staying hidden. The panel's forecasts have been checked externally, and its year-ahead revenue expectations track what those same firms report later and what the national accounts show. And the finding that matters most here, which motives actually pay, is a relationship between self-reported variables rather than a level, so it holds up under a uniform reporting bias far better than any level drawn from the same data. Treat the levels as soft and the relationships as solid. It is also a working paper and has not been peer-reviewed.

A correction we owe you. This publication has cited this same survey since Issue 9, through secondary coverage, using its headline number: roughly 502,000 AI-driven job cuts projected for 2026 against 55,000 in 2025. Having now read the paper itself, that figure needs a caveat we did not give it. It is a modeled aggregate built from what executives expect, weighted up to the economy. It is not a count of anything that has happened, the confidence interval runs from about 78,000 to 925,000, and the paper's own authors note that new firm formation could make the net effect smaller still. The number is still worth knowing. It should not be quoted the way we quoted it, and the Issue 9 entry now carries the caveat.

Why this gets worse before it gets better

On 1 October 2026, Connecticut SB 5 takes effect. Any employer filing a WARN-standard mass-layoff notice with the state has to say whether the layoff is related to the firm's use of AI or other technology change. California's Executive Order N-6-26, signed in May, puts no immediate obligation on private employers and instead tells state agencies to investigate and build a dashboard, measuring first and governing later. The bipartisan Great American Artificial Intelligence Act discussion draft released in June carries a Workforce title that would require similar WARN disclosure where AI is a "substantial factor." None of them says what the term means.

A field with no definition, filed under penalty, aggregated into a public series, and quoted back at you as evidence

This is not an argument against the disclosure. Being transparent about technological displacement is reasonable policy, and the direction of travel is clearly toward more of it. It is an argument about what that data will and will not support. Within a year there will be a state-published count of AI-related layoffs and it will get treated as measurement. It will actually be a pile of legal judgments made by counsel under uncertainty, in a year when the same attribution moved from first to fourth in a single month.

The practice now has a name, AI washing, and the Budget Lab's Martha Gimbel has been its most prominent skeptic, arguing that however you cut the data the macroeconomic effect is not visible. Oxford Economics' Ben May and Revelio Labs' Lisa Simon make the same case from the company side: firms using technology change to dress up corrections they would have made anyway. The reverse case is just as instructive. When Amazon announced roughly 16,000 more role reductions, it credited cutting bureaucracy rather than AI, turning down the fashionable reason at the moment it was most available. Mislabeling runs in both directions, which is the point. The field is discretionary.

The sharpest version of the objection comes from the person with the least reason to make it. Speaking to Channel NewsAsia in May, Nvidia chief executive Jensen Huang called the narrative linking AI to job losses "just too lazy", on the straightforward ground that the timeline does not work: generative tools only became broadly useful recently, so cuts made a year or two earlier cannot honestly be pinned on them. He added that some executives reach for AI to sound smart, and that frightening people this way is irresponsible. Google DeepMind's Demis Hassabis has made a similar criticism. Read Huang's interest before you read his argument. He sells the compute, and a world where AI expands what companies attempt is a larger market for him than one where it shrinks payrolls. That does not make the timeline objection wrong, and it is the same objection the August data produced independently. It does mean the quote belongs here as corroboration, not as authority. Underlying announcements and economist commentary are compiled here.

What this means if you are building a workforce strategy: stop treating the attribution as the finding. Whether a cut gets coded to AI tells you almost nothing about whether the work was really recomposed, whether capacity was freed and where it went, or what shape your organization is now in. Those are the questions with operational content, and none of them fit in a disclosure field. The rest of this issue is about the one that matters most.

← PreviousStephen's Take
5 of 13
All sections
Next →The New Pyramid
The Centerpiece · Prescription

The New Pyramid

The pyramid is not disappearing and it is not simply flattening. Its weight moves. You get a narrow top, a tucked waist, and two wide bands that never used to be wide. Those bands are where a company becomes hard to copy.

Why the usual forecast is wrong

The default projection is that AI shrinks the pyramid. Cut the base, hold the proportions, keep the geometry. That is the shape a headcount model produces, because a headcount model only has one dial. It is also the most reliable way to destroy value with this technology. Every reversal in the record is a company finding out it shrank a pyramid when it should have changed the shape. About a third of hiring managers who cut a role mainly because of AI have already rehired for the same or a similar one. A majority of leaders who ran that play now say it was the wrong call. And a third of the companies that reversed spent more restaffing than the cut ever saved.

A proportional cut is not a redesign. It is the old shape with fewer people in it.

What the evidence actually shows is not scaling. It is redistribution, running as four movements at once, in companies that are really reinventing rather than trimming.

The inherited shape and the recomposed one

Widths show the share of people in each band, not cost or level. The shape on the right is a model and not a forecast. Where the bands sit differs by company, which is the whole point.

Inherited: execution-led Volume creates value Direction Management Supervision Direct execution Supervision is overhead. The base is the product. Redeployed, not recruited Recomposed: judgment-led Judgment creates value Direction holds, slightly wider Judgment & exceptions wing Deep expert waist Orchestration wing Remaining execution Peak & variable Fewer layers, less height, two wide bands. Widths = share of people Wing placement varies by company

The four movements

Movement one
↑
The base tugs up

Execution becomes orchestration

Direct execution shrinks, but the people in it do not leave. They move. The entry role stops being do the task and becomes set up, direct and check the thing that does the task. That is a promotion in the work without being a promotion in level, and it is the most misread of the four, because on a headcount report it looks like nothing happened and on a reason code it looks like a cut. This is redeployment, not recruitment, and the difference decides whether you are making a hiring decision or a design decision.

Movement two
↔
The middle widens into wings

The hard cases become the work

Machines take the ordinary case. What comes back is the hard one: the exception, the ambiguous call, the customer who needs something the policy does not cover, the defect the model did not flag. Handling that takes closeness to the work, real expertise and judgment, and it has always been the thinnest band on the chart. In a company that is reinventing it becomes the widest human layer in the building. This is not adding managers. Supervision shifts from managing activity to managing outcomes and exceptions, which is a different job with deeper expertise behind it, not more of the same one.

Movement three
→←
The waist tucks in

Coordination loses its reason to exist

One layer existed mainly to move information up and instructions down. Routine coordination is the most automatable work in the enterprise, more automatable than most execution, because it is almost entirely information movement. When that goes, the layer built around it has no remaining reason to be a layer. This is where the frozen middle sits, and it is why the waist is not a cut in the usual sense. The function got absorbed while the bands directly above and below it grew.

Movement four
⋅⋅⋅
A detachable band appears

Peak capacity leaves the permanent base

Surge and seasonal capacity that used to sit permanently in the execution base becomes contracted, seasonal or platform-sourced. The reason is structural rather than financial. When the base is orchestration rather than execution, buying surge is a different transaction. You are no longer renting hands for a known task, you are renting throughput against a system somebody inside already directs. That band belongs outside the silhouette, which is where it is drawn.

The evidence that this is redistribution and not reduction

The clearest support comes from the same Fed data, and it is the part least affected by the attribution problem, because it concerns relationships between categories rather than levels. Executives were asked to describe, in their own words, which roles AI replaces and which it improves. Those descriptions were mapped to occupation groups, and the ratio of replacement mentions to enhancement mentions gives the sharpest task-level picture available.

Occupation group Replace : improve Where it lands in the shape
Office & administrative support2.03The waist and the remaining base. The only group where replacement wins
Business & financial operations0.83Splits. Tasks move around inside the job rather than the job going away
Computer & information technology0.60Orchestration wing
Legal0.47Judgment wing
Production0.31Orchestration wing
Sales0.30Judgment wing, close to the customer
Management0.14Judgment wing and the top band
Architecture & engineering0.10Deep-expert waist. Narrow, and almost never replaced

Read down that column and the shape draws itself. One group above 1.0. Everything else improved by a factor of somewhere between one and ten. The forward projections in the same survey agree. Routine clerical work falls 0.76 percentage points of the workforce in 2026 and 2.19 points by 2028, while skilled-technical rises 0.62 and then 1.35. The movement is also not the same across companies. Large firms shed clerical share while smaller firms hold it and add technical roles, which the authors read as reallocation happening less inside companies and more across the economy. That difference is the shape change showing up at national scale, invisible in an aggregate that cancels it out.

Why the wings are the strategy

Here is the part worth arguing over. Two competitors can run the same frontier models, rent the same capacity, and automate the same ordinary case, and increasingly they do, from the same three vendors at converging prices. The Fed data shows why that is now permanent. At the typical company, capital deepening explains only about 15% of AI-driven productivity gains, because most AI spend is an operating expense, subscriptions and services and training, rather than capital investment. Large firms put 55% of AI spend into operations, small firms 64%. Everybody rents capacity from the same central pool, so capacity is common by design.

You can buy capacity, so everyone has it. Where you put your people is a decision only you make.

What a competitor cannot copy is where you chose to put people. Close to which customers, which risks, which exceptions, which craft. Those choices decide what your company is good at in a way that matched spending never reproduces, because they are not a purchase. They are a judgment about where your value actually gets made, written into the org as headcount. Two companies with identical technology and identical cost structures can have completely different wings, and you will see the difference in retention, in defect rates, and in what customers say when something goes wrong.

The wings also do not land at the same level in every company. That is not a weakness in the model, it is the model. In a regulated manufacturer the wing forms around qualification and deviation calls, several levels below the executive team. In a specialty retailer it forms at store and category level, close to the assortment decision. In a bank it forms around credit judgment and keeping relationships intact. In a hospital system it forms at the bedside. If the wing landed in the same place for everyone it would be a template, and a template by definition does not differentiate you. The part that generalizes is the geometry: narrow top, tucked waist, two wide bands, detachable peak. Where the bands sit is your strategy, and it is the question the reason code will never ask you.

Rotate, redeploy, release, mapped onto the shape

These three moves have run through this publication since Issue 15. The shape is what makes them concrete, because each one does something different to the silhouette. The Fed evidence also gives them very different returns. And the decision among them belongs at the enterprise, not with the budget holder whose function happened to free the capacity.

Rotate: move capacity to the same kind of work somewhere else. Keeps the shape. Buys time, changes no geometry, and is the right call only when the shape is already right.
Redeploy: move capacity into different or new work. This is the movement that builds the wings, and it is the only one the productivity data rewards. Motives tied to building products and serving customers better are the strongest and most consistent correlates of real revenue-productivity gains in the survey.
Release: free capacity out of the organization. A legitimate move and sometimes the right one. Also the move with no measured productivity return in this dataset. Cost-reduction motives show imprecise and often negative coefficients, and executives themselves rank cutting labor costs last among their reasons for investing, at 2.0 on a 0 to 4 scale against 2.9 for production efficiency.

Set against the founding argument of this series, that is about as direct a confirmation as survey evidence gets. Capacity freed is not headcount reduced, and we can now say something firmer than that it is a values position. The companies treating freed capacity as a cost event are, by their own numbers, not the ones getting the productivity. The ones getting it moved that capacity into growth. Release is the pathway that shows up in the announcement and disappears from the results.

The same finding turned up in blunter language, from an unexpected direction. Asked at Nvidia's GTC conference why so many companies announce job cuts and credit his chips as the reason, Jensen Huang answered: "Because you're out of imagination." Companies with imagination do more with more, he argued; where leadership has run out of ideas, extra capability simply goes unused. He framed it as a failure of vision rather than a limit of the technology, which is the Fed regression stated in plain English by the man selling the shovels. Which is also the caveat. A world where AI expands what companies attempt is a bigger market for Huang than one where it trims payrolls. He is arguing against displacement from a position that benefits from arguing against it. What makes it worth quoting is that he lands on the same place as a Federal Reserve survey with no such interest.

Four questions to test whether you are changing shape or just shrinking

Each of these is answerable this quarter, and each one fails loudly if the answer is a shrug.

1. Where are your wings, named as bands of people rather than as functions? Point at the layers where closeness and judgment create value a competitor with identical technology could not reproduce. If the answer is a function name rather than a band of people, you have described an org chart, not a shape.
2. Is your entry level being redeployed or removed? Redeployment means the same people, new task content, orchestration instead of execution, with the job architecture and pay bands updated to match. Removal means the band thins and nothing takes its place. On a headcount report these look identical. On a capability plan they are opposites.
3. Who is handling the hard cases the machines send back, and is their time funded as real work? Exception handling, escalation review and correcting the model are the wing's actual workload. If that runs on the goodwill of people who also carry a full delivery load, the wing exists on your diagram and not in anybody's calendar.
4. Who decides where freed capacity goes? If the answer is the function that freed it, you will get release by default, because that is the only pathway visible in their budget. Freed capacity is an enterprise asset. Put the pathway decision above the function and make one person accountable for it.

What would prove this wrong

The honest limits, because a model earns its keep by being checkable. The strongest counter-case is still the Budget Lab position that the occupational mix is moving no faster than in earlier technology waves. If that holds at company level as well as national level, then what we are calling a shape change is ordinary churn with new vocabulary. The geometry here comes from a mix of survey self-attribution, forward expectations to 2028 that nobody has had to deliver yet, and patterns we see in live workforce projections rather than realized outcomes. The band widths illustrate; the direction of movement is the claim. What would settle it is unglamorous and completely obtainable: three years of company-level data on whether the people who left direct execution turn up in orchestration roles at the same employer, or do not turn up anywhere. That series does not exist. The reason-code series will exist by this time next year, and it will not answer the question. Which is the one thing worth carrying out of this issue.

← PreviousThe Reason Code
6 of 13
All sections
Next →The Debate
The Flashpoint

The Great AI-Economy Debate

Each issue adds a fault line and keeps the earlier ones live. Issue 18 adds one above all of them, because it decides how the rest get argued. Is the change AI produces a count, meaning jobs up or down with a cause attached, or a shape, meaning people moving between bands in a way no count can see? The automation and augmentation axis, the reversibility axis and the capital reallocation axis all stay live beneath it.

The New Lead Axis — Issue 18

Counting vs. Shape

The Counting Read

"Count the jobs. Name the cause. That is the measurement."

Click for the counting case →
The Shape Read

"The total barely moves. Where people sit moves a great deal."

Click for the shape case →
The Counting Read: layoff trackers, WARN disclosure rules, headcount models

A Number and a Cause

The dominant frame, and from 1 October the legal one. AI's workforce effect is a number of jobs, up or down, with a cause you can name. It is readable, comparable across companies, and now it is filable. Every forecast built on "X% of roles are exposed" is a counting claim.

Connecticut SB 5 writes the counting read into law from 1 October 2026
AI cited in 116,175 US job cuts year to date, about 22% of the total
Its strength: you can audit it, and a measure you cannot audit is not a measure
Its weakness: the same cause ranked first for five months and fourth in the sixth, with nothing changing in the technology
Second weakness: it nets to near zero nationally at the exact moment company-level movement is largest
The Shape Read: Fed CFO Survey occupational composition data

A Redistribution Between Bands

The alternative view is that the total is the least informative number available, because a recomposition barely moves the total by definition. What changes is where people sit relative to each other: which band widens, which tucks in, and where in the structure judgment concentrates. A count cannot show you that. Only a shape can.

Routine clerical share expected down 2.19 points by 2028, skilled-technical up 1.35
Replacement-to-improvement ratio: admin support 2.03, management 0.14, engineering 0.10
Large firms shed clerical share while smaller firms hold it and add technical roles, so the movement runs across the economy and not only inside companies
Total AI employment effect for 2026 modeled at under 0.4%, which is an expectation drawn from self-reported data and not a realized outcome
Its weakness: shape is harder to audit than a count, and no standard company-level series for it exists yet
Still Live — Corroborated in Issue 15

Automation vs. Augmentation

The Automation Read

"Remove the human. The agent does the whole task now."

Click for the substitution case →
The Augmentation Read

"Recompose around the human. Partnership beats either alone."

Click for the worker-desire case →
The Automation Read — Layoff trackers / agentic-AI claims

Remove the Human From the Loop

The aggressive read: agentic AI can now execute end-to-end white-collar tasks, so the efficient move is full automation (H1) and headcount removal. It is the logic behind the ~186K 2026 cuts and the "AI made the role obsolete" announcements.

2026 tech layoffs past 186,000 — ≈978/day, ~2× the 2025 pace
53% of layoff events explicitly cite AI / automation
Cloudflare, Coinbase, Intuit framing cuts as AI making roles "obsolete"
The case's weakness: Gartner shows the deepest cutters get no return premium
And Stanford shows H1 is the level workers desire least across most tasks
The Augmentation Read — Stanford WORKBank / Digital Economy Lab

Recompose the Work Around the Human

The evidence-led read: value and worker desire both peak at Equal Partnership (H3). The job isn't to subtract people — it's to redesign tasks so human judgment and AI execution compound. This is the path the returns data and the workforce both endorse.

H3 (Equal Partnership) is the most-desired agency level in 47 of 104 occupations
Workers want more human agency than experts deem necessary on 47.5% of tasks
46% of tasks rated positively for automation — appetite is real, but targeted
69.4% want AI to free time for high-value work — augmentation, not exit
Skill demand shifts toward coordination, coaching, and judgment alongside agents
Aligns with Gartner: returns come from recomposition, not headcount subtraction

The Talent Strategist's Take: The Workforce Already Settled This One

For the first time, the debate has a referee that isn't a CEO or an analyst — it's 1,500 workers rating their own tasks. The verdict is decisive: the automation extreme (H1) is the least-desired state across most occupations, while Equal Partnership (H3) is the most-desired. This isn't "don't automate" — workers welcome automation on 46% of tasks. It's that the unit of decision is the task, not the title, and the destination for most work is partnership, not removal. The organizations winning the recomposition will automate the green-light tasks honestly and design the rest toward H3 — because that is simultaneously what workers want and where Gartner's returns actually show up. The two questions leaders treat as a trade-off turn out to share one answer.

Still Live — Issue 17

Does the Learning Regenerate on Its Own?

The Emergent Read

"New work creates new learning. It always has."

Select for the self-correcting case →
The Institutional Read

"Learning is manufactured. Remove the factory and nothing replaces it."

Select for the deliberate-build case →
The Emergent Read — Yale Budget Lab / historical transition evidence

The Ladder Rebuilds Itself

The historical case: every general-purpose technology destroyed an apprenticeship and grew a new one without central planning. Junior work does not vanish, it relocates — to prompt design, output validation, agent supervision, exception handling — and juniors learn there instead. Panic about the ladder mistakes a change of rungs for the loss of a ladder.

Yale Budget Lab: US occupational mix shifting no faster than during the PC or internet transitions
No relationship found between an occupation’s AI exposure and its employment or unemployment duration
BLS layoffs-and-discharges rate remains close to normal; the softening looks cyclical, a hiring pause
Ramp/Revelio: heaviest AI spenders grew entry-level headcount 12% — new rungs appearing where investment is real
The case’s weakness: it explains the aggregate while the damage is age-specific and concentrated at 22–25
The Institutional Read — German establishment panel / apprenticeship literature

Learning Is Manufactured, Not Emergent

The structural case: professional judgment has never been a by-product of task exposure alone. It requires supervision, feedback with consequence, and a progression ladder — an institution. Where one exists, AI intensifies it. Where none exists, removing the tasks removes the only accidental substitute, and nothing appears in its place because nothing was ever there.

German panel: AI adoption raises new apprenticeships 14% at training firms; zero effect on whether firms start
The professions that kept their on-ramps are those where the apprenticeship is legally mandated, not chosen
2026 review of 18 studies: the informal post-degree apprenticeship system no longer reliably exists
32% of managers who cut a role to AI have rehired; 48% have canceled projects for lack of qualified people
The case’s weakness: it is largely European evidence, and building an institution is a five-year cost with no in-cycle return
Still Live — Issue 16

Capacity Confidence vs. Decision Reversibility

The Capacity Read

"The buildout is real. Position the workforce for it now."

Select for the forward-positioning case →
The Reversibility Read

"Match the irreversibility of the decision to the confidence in the input."

Select for the underwriting case →
The Capacity Read — Infrastructure commitments & capex trajectory

Position for the Buildout

The forward-positioning case: capability is compounding, the infrastructure commitments are enormous and specific, and the organizations that restructure early will hold a durable cost and speed advantage over those that wait for proof. On this read, hesitating is itself a decision — and an expensive one.

Nvidia reported working on more than $750B in fresh AI infrastructure deals in late July 2026
AI capex estimated near 2% of US GDP; roughly 2,800 data centers planned in the US
Challenger through July: AI cited in 112,713 cuts (~24%), top reason for five consecutive months
The case's weakness: the purest levered expression of it fell ~78% in three weeks
Updated for Issue 17: the sector has partly recovered since late July — which cuts against the drawdown proving anything about diffusion, exactly as we cautioned
And a meaningful share of the demand signal may be vendor-financed rather than end-customer funded
The Reversibility Read — Decision economics under uncertainty

Underwrite the Decision, Not the Forecast

The underwriting case: you do not need to resolve the capability question to make a good decision. You need to match the reversibility of each move to the confidence in the input behind it. Software licenses, process redesign and pilots are reversible — move fast. Deleting a graduate intake is not — hold a floor.

A public book can be unwound in a single block trade; a cohort cannot be rebought at any price
Citadel Securities: if compute's marginal cost exceeds labor's, substitution does not occur — an economic ceiling
Forrester on the selloff: a repricing of expectations, not evidence of weakening demand — which cuts both ways
IBM's move: concluded AI can do most of the old junior work, then rewrote the jobs and tripled intake
The case's weakness: holding a floor costs real money now against a benefit that arrives in a decade
Still Live — Issue 11

Capital Reallocation vs. Capability Replacement

The Capital Reallocation Read

"$725B in 2026 capex. Salaries are the only flexible cost."

Click for the balance sheet read →
The Capability Replacement Read

"Agentic AI is genuinely doing white-collar tasks at scale."

Click for the technology read →
The Capital Reallocation Read — Bloomberg / Invezz / Big Tech filings

Salaries Fund the Capex

Big Tech's 2026 capital expenditure is the largest concentrated capital reallocation in modern corporate history. The cuts aren't really about AI productivity — they're about funding the buildout on shareholder timelines. AI is the cover story, not the cause.

$725B 2026 capex (Google + Amazon + Microsoft + Meta) — up 77% YoY from $410B
More than the entire global oil and gas industry spends on exploration
Microsoft Q2 2026 capex: $37.5B in a single quarter, alongside 8,750 buyouts
Meta capex: $125–145B (~$370M/day on data centers) — 8,000 cuts effective May 20
Bloomberg: ~half of "AI-attributed" cuts result in same roles rehired offshore at lower wages — labor repricing, not displacement
Sam Altman (OpenAI) and Babak Hodjat (Cognizant) both publicly admit AI washing
The Capability Replacement Read — Yale CELI / Goldman / Sonnenfeld

Agentic AI Is Genuinely Doing Work

The capital reallocation read has merit, but it understates real capability change. Agentic AI is genuinely performing knowledge work — and the entry-level roles that train future seniors are the canary.

Goldman: AI suppressing ~16,000 U.S. jobs/month, concentrated in routine white-collar roles
Yale Sonnenfeld/Tian: "AI won't kill your job — it will kill the path to your first one"
Stanford 2026 AI Index: software developers aged 22–25 down nearly 20% since 2022
Anthropic Economic Index: customer service, business sales, automated trading API workflows doubled Nov 2025 → Feb 2026
Salesforce CEO Marc Benioff on cutting 4,000 customer support roles: "I need less heads"
NACE 2026 Winter Salary Survey: starting CS salaries up ~7% YoY — high-skill demand rising while entry-level openings vanish

The Talent Strategist's Take: Both Are True — and the Conflation Is the Problem

This is the most important strategic distinction of the cycle. Both reads are partially correct. Capital reallocation is real — Bloomberg's offshore-rehiring data and Altman/Hodjat's public admissions confirm it. Capability replacement is also real — Goldman's 16K/month and Anthropic's API doubling confirm it. The strategic failure happens when leaders conflate them. A "capital reallocation" cut requires honest workforce communication and offshore strategy. A "capability replacement" cut requires task-level redesign and reskilling investment. Treating both as the same — which most boardrooms are doing right now — guarantees mismanagement of both. The consultancy work that matters most this quarter: helping clients tell the two apart in their own portfolio, then governing each honestly.

The Corporate Fork — Continued from Issue 10

Oracle vs. IBM: Same Technology, Opposite Conclusions

The Oracle Path — Cut Humans, Fund Machines

"Your role has been eliminated as part of a broader organizational change."

Click to explore the case study →
The IBM Path — Triple Down on Humans

"The companies that doubled down on entry-level hiring will be the most successful."

Click to explore the counter-model →
The Oracle Path — March 31, 2026

Cut Humans to Fund Machines

Oracle eliminated up to 30,000 employees — 18% of its global workforce — in a single morning via 6 AM email. Now extended through Issue 11: Microsoft's first-ever buyout (8,750), Meta's 10% cut (8,000), Salesforce's 4,000 customer support eliminations. The Oracle template is now the industry default.

Oracle: 30,000 cut to free $8–10B for AI infrastructure — $156B total buildout (TD Cowen)
Microsoft: First-ever voluntary buyout in 51 years — 8,750 (7% of U.S.) — same outcome, softer optics
Meta: 8,000 cuts effective May 20; recruiting/HR absorbing 35–40% of cuts
Salesforce CEO Marc Benioff on 4,000 customer support cuts: "I need less heads"
Nike: 1,400 in tech department; Snap: 1,000 (16% of workforce); Block: 4,000 (40%)
Reputational lesson learned: every CHRO is now redesigning the cuts to be "voluntary"
The IBM Path — February 2026

Triple Down on Humans Alongside AI

IBM tripled entry-level hiring in the U.S. in 2026 — explicitly for the roles other companies say AI can replace. CHRO Nickle LaMoreaux personally rewrote every job description. Issue 11 reinforcement: NACE 2026 Winter Salary Survey shows CS major starting salaries up ~7% YoY — the IBM-style bet on early-career investment is being priced in by the labor market.

Tripled entry-level hiring "across the board" while 37% of companies plan to replace those roles with AI
Redesigned junior roles away from coding toward customer judgment, human interaction, AI oversight
CHRO: "The companies 3–5 years from now that are most successful doubled down on entry-level hiring"
CEO Krishna: "People are talking about layoffs. We are the opposite."
NACE 2026 Winter Salary Survey: CS major starting salaries up ~7% YoY despite headlines
Sonnenfeld/Tian (Yale CELI, Apr 29): "Kill the layoffs we can't see" — the entry-level path being severed is the long-term cost

The Talent Strategist's Take: The Oracle/IBM Fork — Still the Defining Choice

The Oracle path didn't slow this quarter — it spread, and it learned. Microsoft's first-ever buyout is the same balance sheet decision in better packaging. Meta's 8,000 starting May 20 is the same mechanism with calendared distance. The IBM path, meanwhile, just got two new institutional validators: the NACE salary data showing the labor market is rewarding early-career investment, and Sonnenfeld/Tian naming the long-term cost of severing the apprenticeship pipeline. The leadership question hasn't changed — but the time horizon for the consequences has shortened.

Continued from Issue 9

Expectation vs. Evidence — The Tension Deepens

The Expectation Side — Duke/NBER/Fed CFO Survey

"Cutting on Potential, Not Performance"

Click to explore 5 key findings →
The Evidence Side — Anthropic Labor Market Paper

"Limited Evidence AI Has Affected Employment"

Click to explore 5 key findings →
The Expectation Side — Duke/NBER/Fed CFO Survey

"Cutting on Potential, Not Performance"

750 CFOs surveyed. AI-driven layoffs projected 9× higher in 2026 (~502,000 vs. 55,000 in 2025). But productivity perceptions exceed actual results. Updated for Issue 18: the 502,000 is a modeled aggregate from executive expectations, not realized cuts, with a range of roughly 78,000 to 925,000.

502,000 projected AI-driven job cuts in 2026, up from 55,000 in 2025 (modeled expectation, range ~78K–925K)
Only 44% of CFOs plan AI layoffs — the other 56% are holding steady
"It's not really showing up yet in revenue" — study co-author John Graham
59% of companies framing ordinary cost-cutting as "AI-driven" for investor appeal
Citrini's "Ghost GDP" is materializing — but from expectation, not from proven capability
The Evidence Side — Anthropic Labor Market Paper

"Limited Evidence AI Has Affected Employment"

Anthropic's own research — using a new observed-exposure framework — finds no systematic unemployment increase for AI-exposed workers since late 2022.

New measurement: observed AI use (actual Claude traffic) vs. theoretical exposure
No clear causal link between AI exposure and unemployment in CPS or DOL data
Augmentation slightly increasing over automation across Claude platform usage
Task concentration declining: top 10 tasks went from 24% to 19% of conversations
"The track record of past approaches gives reason for humility" — the paper's opening line

The Talent Strategist's Take: The Evidence Finally Caught Up

For three issues this box documented a paradox: firms cutting on expectation while the measured evidence for AI-driven job loss stayed thin. Issue 13 resolves it — not by proving the cuts were justified, but by exposing that they were never really about capability. Three new pieces of evidence landed at once: Stanford's WORKBank showed workers want partnership (H3), not removal; Gartner showed the deepest cutters earned no return premium; and NBER found 90% of executives admit AI had zero employment impact at their own firm — even while headlining it externally. The expectation-driven story has finally met hard evidence, and the evidence says the cuts were balance-sheet pressure dressed in technology vocabulary. Now disclosure law (CA SB 951, Colorado, Warner–Hawley) is arriving to make firms name which system, if any, actually did the work.

The Empirical Clash — Strengthened

The Task-Level Evidence Converges

Stanford Digital Economy Lab

"Canaries in the Coal Mine"

Click for full findings →
Google Chief Economist + Iscenko & Millet

"The Timing Doesn't Fit"

Click for full findings →
Stanford Digital Economy Lab

"Canaries in the Coal Mine"

Using ADP payroll data covering millions of workers, Stanford found a 16% relative employment decline for workers aged 22–25 in the most AI-exposed occupations since late 2022. The effect is concentrated where AI automates (not augments) labor. Holds after controlling for firm-level shocks, remote work, and pandemic effects.

Google Chief Economist + Iscenko & Millet (2026)

"The Timing Doesn't Fit"

Google's chief economist argues the decline began April 2022 — six months before ChatGPT — coinciding with the Fed's rate-hiking cycle. His paper with Iscenko reaches the same conclusion: macro tightening, not AI, explains the employment shifts. Calls AI "the most profoundly transformative technology" but says labor displacement is premature.

The Talent Strategist's Take: The Empirical Question Has Moved

The Stanford/Google timing argument — did the decline start before or after ChatGPT — is now a side debate. Issue 13 shifts the empirical center of gravity entirely: from counting displacement to mapping agency. The question that matters is no longer just "is AI cutting jobs?" but "what do workers want AI to do, and where does value actually come from?" And on that, the evidence now converges hard: MIT CSAIL on capability (47–73% task success), HBS on augmentation rising over automation, Stanford WORKBank on worker desire peaking at Equal Partnership, and Gartner on returns coming from recomposition, not subtraction. Four institutions, four methods, one answer: the value is in augmentation toward partnership — and the cuts racing toward full automation are aimed at the corner of the map that workers want least and returns reward least.

← PreviousThe New Pyramid
7 of 13
All sections
Next →Signal Map
Signal Map · Four Views, One Section

Where the Evidence Agrees, Splits and Sits

Convergence, divergence, job categories and the institutional matrix are one question asked four ways, so they run as one section. Eleven convergent themes, nine live disagreements, eight job categories, and the full comparative matrix across 52+ institutions.

Across 52+ institutions and competing frameworks, these themes show up with remarkable consistency. Issue 18 adds an eleventh, and it is the one this publication has been building toward since the first issue.

Convergence 1

Task Transformation, Not Job Elimination REINFORCED

Now validated from both the capability side and the worker-desire side. MIT CSAIL tested 11,500 real-world tasks across 40+ AI models: success rates range from 47% to 73% — confirming AI advances at the task level, not the job level. HBS/Srinivasan measured it in job postings: automatable tasks ↓17%, augmentation tasks ↑22%. New for Issue 13: Stanford's WORKBank audited 844 tasks and found Equal Partnership (H3) is the most-desired agency level in 47 of 104 occupations, with workers wanting more human agency than experts deem necessary on 47.5% of tasks. Anthropic's Economic Index v5 independently shows augmentation rising over automation. Five institutions, one conclusion: the transformation is recomposition, not elimination.

Convergence 2

Entry-Level Workers Bear Disproportionate Impact REINFORCED

The convergence keeps strengthening. Goldman Sachs now estimates AI is suppressing ~16,000 U.S. jobs/month, concentrated where Gen Z is overrepresented. Yale's Jeffrey Sonnenfeld and Steven Tian reframed it: "AI won't kill your job — it will kill the path to your first one." Stanford's 2026 AI Index (devs aged 22–25 down 20%), ServiceNow's McDermott (30%+ grad unemployment forecast), HBS/Burning Glass (18M entry-level jobs at risk), and Duke CFO survey (502K projected cuts, modeled) all converge on the same finding. The IBM counter-model stands as the only major institutional contrarian.

Convergence 3

The Deployment Gap Remains Massive

MIT's 5× gap is reinforced by the NBER finding and the SF Fed's productivity paradox assessment. Only 17% of firms use AI (Fed). The CEPR study found without complementary investment — especially in training — AI adoption alone is insufficient. Anthropic's April 2026 update shows experienced users get dramatically better results than newcomers — the skills gap is widening even within Claude's own user base. Solow's paradox is back, with a fresh twist: the people building the technology don't know how to use it.

Convergence 4

The Macro Feedback Loop Is the Genuine Risk FORMALIZED

Academic work has now formalized it. Falk & Tsoukalas (Wharton/Boston U, March 2026) — "The AI Layoff Trap" — proves rigorously that competitive demand externalities trap rational firms in an automation arms race that displaces workers beyond what is collectively optimal, eroding the very consumer demand they all depend on. They conclude that wage adjustments, free entry, capital income taxes, UBI, upskilling, and Coasian bargaining all fail to correct this — only a Pigouvian automation tax can. Citrini's "Ghost GDP" resonated because it intuited a real structural risk; the AI Layoff Trap paper now provides the formal model. Both the Fed's "jobless boom" scenario and KPMG's growth-labor decoupling observation point in the same direction.

Convergence 5

Human Capital Investment Is the Multiplier

The CEPR study provided the clearest evidence: an extra point of training investment amplifies AI productivity by 5.9×. PwC's 56% wage premium for AI skills, WEF's finding that AI credentials offset age and education disadvantages, and Anthropic's experienced-vs-newcomer gap all point to the same conclusion. New for Issue 11: WEF Davos 2026 — wages for AI roles up 27% since 2019; firms struggling to recruit because skill acquisition lags demand.

Convergence 6

Capital Reallocation Is the Mechanism Behind the Cuts

The buyout quarter exposed the math. $725B in 2026 Big Tech capex (up 77% YoY) exceeds the entire global oil and gas industry's exploration budget. Microsoft alone spent $37.5B on AI infrastructure in a single quarter while announcing 8,750 buyouts. Bloomberg analysis: ~half of "AI-attributed" cuts result in offshore rehiring at lower wages — labor repricing, not displacement. OpenAI's Sam Altman: "There's some AI washing where people are blaming AI for layoffs that they would otherwise do." Cognizant's Babak Hodjat: "AI becomes the scapegoat from a financial perspective." When the technology vendors themselves admit it, the rhetorical cover is gone — and the underlying convergence on capital-as-driver becomes impossible to deny.

Convergence 7

The Headcount Math Fails From Both Directions

Accenture's task-level model proves what the series has argued, now in the operational core: demand for core US supply-chain roles rises +19% (1.34M) by 2035 while the labor force grows +3.2% (221K) — a ~1.1M gap hiring can't close — and modeling full automation on a pharma planning function targeting a 33% cut netted only ~6%. 52% of tasks can only be augmented; agentic AI touches 81% but automates just 30%. This converges with Anthropic's Economic Index (augmentation rising over automation), MIT CSAIL's task-level success ceilings, and Stanford's worker-desire data. Across capability, preference, and now operational modeling, the conclusion is identical: technology removes tasks, not roles. People are repositioned, not replaced.

Convergence 8

Employment Weakens Where AI Automates — and Holds Where It Assists

The lead debate axis of this publication has become an empirical finding. Reporting the H1 ledger, Challenger, Gray & Christmas cited Stanford Digital Economy Lab research showing employment has weakened in occupations where the technology automates tasks, while holding up in roles where AI helps employees do their job. That is the automation-vs-augmentation split, measured in payroll rather than argued in forecasts — and it converges with Anthropic's Economic Index (augmentation share rising on Claude.ai while automation dominates programmatic API traffic), Accenture's finding that 52% of supply-chain tasks can only be augmented, MIT CSAIL's task-level success ceilings, and Stanford WORKBank's worker-desire data. The practical consequence is sharp: how you deploy determines whether headcount holds, independent of how much you deploy. Two firms with identical AI spend and opposite deployment philosophies will show opposite employment curves. Challenger's forward flag: finance is the next sector to watch.

Convergence 9

The Effect Concentrates at the Entry Rung

Institutions that agree on almost nothing else now agree on where the labor-market effect is showing up first. Stanford Digital Economy Lab and ADP put workers aged 22–25 in the most AI-exposed occupations roughly 13% below peers, with the gap widening about half a point a month. PwC finds AI-exposed entry-level postings seven times more likely to demand judgment, leadership and creativity — the "seniorised" junior role. MIT's Andrew McAfee names the mechanism as the loss of the apprenticeship ladder. Anthropic's labor analysis lands independently in the same place. Even the employers expanding intake — IBM, Cognizant, Amazon — agree on the diagnosis and differ only on the response: IBM's CHRO states plainly that entry-level jobs from two or three years ago can now largely be performed by AI. Nobody credible now argues the entry rung is unaffected. The argument has moved entirely to what to do about it.

Convergence 10

Supervision Capacity Is the Real Constraint ISSUE 17

An agreement forming across sources that do not usually agree. Every serious framework for AI-assisted work — Stanford’s agency scale, the augmentation literature, vendor guidance, and our own — converges on a human who validates, challenges and overrides machine output. That role is not junior and it is not free. IBM’s move to point new hires at the 6% of HR cases AI could not handle only works because someone senior is reviewing closely. The German panel finds continuing-training resources shifting toward higher-skilled employees as AI adoption sustains — the same signal from the other end. Robert Half finds leadership among the three hardest capabilities to source. The binding constraint on AI deployment is turning out to be the supply of people qualified to supervise it, and that supply is produced by the very apprenticeship layer being compressed. Organizations are budgeting for tools and licenses while the scarce input is senior attention.

Convergence 11

Growth Motives Produce the Productivity; Cost Motives Do Not NEW

The convergence this publication has argued toward since Issue 1, now supported from three directions at once. The Fed CFO survey finds revenue-productivity gains lining up with growth motives, meaning building products and serving customers better, while cost reduction, workforce development and capital upgrading show imprecise and often negative coefficients. Executives themselves rank cutting labor costs last among their own motives at 2.0 on a 0 to 4 scale. The ECB survey of roughly 5,000 firms found intensive AI users about 4% more likely to add staff, with the hiring concentrated in firms using AI for R&D and growth, while the minority using it mainly to cut costs hire less. Strada surveyed nearly 1,500 executives and found firms with a coherent company-wide AI plan reporting increases in entry-level hiring, while firms that only partly integrated, automating routine tasks and stopping there, report reductions. Three datasets, three methods, one finding. The cost lever does not pay and the growth lever does. Capacity freed is not headcount reduced, and it is no longer only a values position.

Where the evidence still splits. Issue 18 adds the one about to become a legal question rather than an analytical one: whether the AI attribution in layoff data measures anything at all.

Divergence 1

Speed of Displacement: Overnight or Decadal?

Citrini models a 2-year spiral. Citadel says decades. The NBER study sides with slow adoption. Stanford's data shows measurable effects already. The SF Fed notes the 1970s IT boom took 20+ years. The honest assessment: the speed is genuinely unknown.

Divergence 2

Expectation vs. Evidence: The Adoption Paradox

Oracle's 30,000-person cut — while posting $3.7B quarterly income — is the expectation economy at full execution. Gallup reveals the paradox underneath: half of U.S. workers never use AI, yet companies restructure as if deployment is universal. MIT CSAIL tested 11,500 tasks and found no category near full automation. Companies are making irreversible workforce decisions based on timelines (12–18 months, per Suleyman) that the empirical evidence doesn't support.

Divergence 3

Will New Jobs Absorb Displaced Workers?

Citadel and PwC say yes — human wants are elastic, 60% of current US jobs didn't exist in 1940. Citrini says this time is different — AI is a general labor substitute. Dario Amodei agrees. The WEF models all four possibilities.

Divergence 4

Is Engineering the First Casualty — or the Most Resilient Role?

The displacement narrative assumes coding goes first. SignalFire's hiring data says the opposite: big-tech hiring is down ~25% versus 2019, but engineering only ~11%, and engineers are now 55% of new hires (up from 46%). If AI were substituting for engineering talent, that hiring would crater first — it hasn't. Against this, Dario Amodei projects AI writing most code, and layoff trackers show engineering-heavy cuts. The real split is methodological: layoff data is a lagging, noisy signal often driven by capital reallocation; hiring data is forward-looking demand. Which one tells the truth about AI substitution is genuinely unresolved — and the answer decides whether "learn to code" survives as advice.

Divergence 5

Is the Entry-Level Collapse Caused by AI — or Merely Correlated With It?

The single most consequential open question in the field, and the honest answer is that it is not settled. Brynjolfsson, Chandar and Chen document a ~13% relative decline for 22–25s in the most exposed occupations that older workers in the same roles don't share — a specificity hard to explain by macro alone. Their February update rules out interest-rate exposure as the explanation. Against that: under the broadest controls the effect only becomes significant from 2024; the authors themselves label the dashboard correlational; pandemic over-hiring and its unwind remain partly in the data; and the July 22 finding on Gen Z women points to occupational sorting rather than AI. SignalFire finds engineering the most resilient function — awkward for a clean substitution story. The strategic point is that the decision does not wait on the answer: the bench thins on either causal account, so the on-ramp needs rebuilding either way. But anyone claiming the causal question is closed — in either direction — is ahead of the evidence.

Divergence 6

Does the Entry Rung Shrink, or Just Change Shape?

The sharpest live disagreement in the market, and both camps have real evidence. Shrinking: PwC's US organization reportedly planning around a third fewer traditional associates by 2028; entry roles across law, consulting and banking down ~35% since 2023; 48% of surveyed hiring managers saying they would rather invest in AI tools than hire and train a graduate. Changing shape: IBM tripling US entry-level hiring after rewriting the job descriptions; Dropbox expanding new-graduate programs 25%; Cognizant targeting 24,000–25,000 graduates on a "broader pyramid" strategy; the agency Brainlabs growing an entry cohort from 19 in 2023 to 64 in April 2026; NACE projecting Class of 2026 hiring up 5.6%. Our read: this is not a contradiction in the data, it is a fork in strategy. The same technology supports both decisions. Which is precisely why it should be underwritten as a decision rather than absorbed as a forecast.

Divergence 7

Is AI Demand Independently Funded — and Does Compute Cost Cap Substitution?

New this issue, and it sits underneath every workforce forecast. Nvidia was reported in late July to be working on more than $750B in fresh AI deals, reviving the argument that vendor financing is inflating apparent demand. Defenders counter that in a supply-constrained market, pairing long-term purchase commitments with financing is how buyers lock in capacity — a virtuous circle, not a circular one. Sitting alongside it is Citadel Securities' boundary argument: displacing white-collar work at scale requires orders of magnitude more compute than current utilization, and if automation expands rapidly the marginal cost of compute rises — so where that cost exceeds the marginal cost of human labor, substitution simply does not occur. It is the strongest available rebuttal to linear displacement and it is not a sentimental one. It also cuts both ways for planners: an economic ceiling on substitution is good news for the workforce and bad news for anyone who has already cut against the unbounded version.

Divergence 8

Does Displaced Learning Regenerate on Its Own? ISSUE 17

The deepest live disagreement in the field, and it determines whether redesign is a design task or an institution-building one. Regenerates: Yale’s Budget Lab finds the US occupational mix shifting no faster than during the PC or internet transitions and no relationship between AI exposure and employment; Ramp and Revelio Labs find the heaviest AI spenders grew entry-level headcount 12%, suggesting new rungs form where investment is genuine; historically every general-purpose technology destroyed one apprenticeship and grew another without anyone planning it. Must be manufactured: the German panel shows AI raising apprentice intake only where a training institution already exists and having no effect on whether one is created; the sectors retaining their on-ramps are those where apprenticeship is legally mandated; a 2026 review of eighteen studies concludes the informal post-degree apprenticeship system no longer reliably exists. Our read: the aggregate evidence and the institutional evidence are not actually in conflict — new rungs do form, but disproportionately inside firms that already had the machinery to build them. That is regeneration for the well-equipped and nothing at all for everyone else, which at an economy level looks like recovery and at a firm level looks like a widening gap.

Divergence 9

Is the AI Attribution in Layoff Data Measurement or Narrative? NEW

A disagreement that was academic until Connecticut made it law. Measurement: AI has been the leading year-to-date reason for US job cuts at 116,175 announcements, about 22% of the total. Companies have specific, checkable reasons to name it, and the new disclosure rules assume the attribution carries information. Narrative: the same reason moved from first to fourth in a single month with nothing changing in the technology. Oxford Economics and Revelio Labs both argue some firms are relabeling overhiring corrections as technology change, because it turns a retrenchment into a strategy. And the Fed CFO data shows executives reporting roughly three times the AI productivity their own revenue and employment numbers show, which is the same problem measured inside the company. Our read: both are true and the proportion is unknowable from outside, which is the problem. A measure that is part signal and part fashion, in unknown proportion, cannot carry the weight now being put on it, and from 1 October it gets filed with a state labor department under an undefined term. Treat the attribution as a disclosure obligation, never as a finding.

White Collar / Knowledge Work

AI Task Exposure57–80%
Actual Displacement (2026)6–7%
Deployment Gap5×

The Citrini/Citadel debate crystallized this category's central tension: theoretical exposure is enormous (McKinsey: 57%, OpenAI: 80%), but actual displacement is modest (Goldman: 6–7%, NBER: 80% of CEOs see no impact). The SF Fed's two-speed economy observation is critical: knowledge sectors (26% of output) drove 50% of GDP growth.

Physical / Frontline Work

AI Task Exposure15–25%
Actual DisplacementMinimal
Demand GrowthStrong

Physical work remains the near-term moat. DeepMind estimates 5–10 years before protection narrows. Citadel cited data center construction as a localized hiring boom. OECD reports 94% of construction firms face difficulty sourcing workers. WEF projects farmworkers as top absolute growth category. 40% of young graduates choosing non-automatable careers. Accenture's supply-chain model sharpens this at task level: across operational roles, 52% of tasks can only be augmented — logisticians and planners see most of their work enhanced rather than replaced, while inspection and physical-handling tasks stay largely untouched. The frontline isn't a displacement story; it's a redesign-and-coordination story.

Entry-Level / Early Career

Employment Decline (AI-Exposed)16%
Codified Knowledge Overlap w/ LLMsHigh
Pipeline Risk (10-Year Horizon)Critical

The most contested category. Stanford's 16% decline (disputed by Google) is reinforced by the Dallas Fed's codified-vs-tacit framework: LLMs replicate what young workers learned in school. IBM's decision to triple young hires is the first corporate counter-move. Recent grad unemployment rose to 5.5% (Goldman).

Creative / Content Work

Freelance Demand Decline20–50%
Quality AugmentationHigh
Premium for OriginalityRising

Cornell research found writing/translation demand on freelance platforms fell 20–50% in AI-substitutable categories — one area where displacement is measurable and immediate. But a split is emerging: commoditized content collapsing while premium creative work commands higher rates. IMF: top researchers using AI boosted output 44%, but 82% reported less creativity — the augmentation trade-off.

Management / Leadership

AI Task Exposure35–50%
Role Redefinition PressureHigh
Agent Management DemandSurging

Deloitte: 36% of managers expect to manage digital agents within 5 years. Microsoft: 82% of leaders plan agent deployment in 18 months. The manager role is shifting from supervising humans to orchestrating human-agent teams. KPMG research suggests agentic AI could flatten hierarchies and reduce middle management — but increase demand for leaders who can govern AI systems.

Technical / Engineering

AI Task Exposure60–75%
Actual DisplacementLow (senior) / High (junior)
Wage Premium for AI Skills56%

The Dallas Fed framework is clearest here: computer systems design employment is down 5% since ChatGPT launch, but wages are up 16.7%. AI substitutes for codified/junior tasks while complementing experienced judgment. Citi reports 9% software development productivity improvement. J.P. Morgan research found cloud, web search, and computer systems design stopped growing after ChatGPT launched.

Sales / Customer-Facing

AI Task Exposure40–55%
Call Center ImpactSignificant
Relationship-Based SalesProtected

Stanford's Canaries paper identified customer service as one of the two occupations (alongside software development) showing the sharpest decline. Klarna reorganized support into hybrid human-AI teams. AI resolved 14% more call center issues per hour (Brynjolfsson). But complex, relationship-driven sales — where tacit knowledge and trust matter — remains protected by the Dallas Fed's codified/tacit divide.

Administrative / Clerical

AI Task Exposure65–80%
Displacement RiskHighest of Any Category
Adaptive CapacityLowest

WEF identifies clerical roles as leading absolute decline. Brookings found clerical workers have the lowest adaptive capacity — fewest transferable skills, professional networks, and savings to absorb disruption. OECD: 90% of US firms deploying algorithmic management tools. This category has the starkest gap between exposure and readiness. Policy intervention is most urgent here. Accenture's supply-chain model puts numbers on the ceiling: the most codified clerical roles — shipping, receiving, and inventory clerks — carry the highest automatable share, but even there full automation lands only when several technologies combine, and new exception-handling and oversight tasks backfill much of the gap. The exposure is real; the clean headcount cut still isn't.

Select any institution name to access the underlying research directly. Use the filters to focus by category or outlook.

Category: All Academic Consultancy Multilateral Frontier AI Finance Central Bank
Outlook: All Alarmed Cautious Optimistic
InstitutionCategoryHeadlineWhite CollarEntry-LevelOutlook
Atlanta & Richmond Fed / Duke NEWCentral BankReported AI productivity ~3× the implied figureRoutine clerical −2.19pp by 2028; skilled-technical +1.35ppAdmin support only group where replacement > enhancement (2.03)Cautious
Strada Institute NEWAcademic1,498 executives: AI raising entry-level hiring 2.7× more than cutting it40%+ say AI raised analytical work asked of juniorsPartial integrators report reductions; full-plan firms report increasesOptimistic
OECD Employment Outlook 2026 NEWMultilateralAdjustment runs through joblessness and new entrants, not worker movementDisplaced workers rarely move into the roles createdNew roles go disproportionately to younger entrantsCautious
Connecticut Dept. of Labor NEWGovernmentSB 5: AI attribution required on WARN filings from 1 Oct 2026Applies to employers with 100+ staffNo statutory definition of "caused by AI"Cautious
Challenger — August 2026 NEWLabor DataAI falls to 4th reason (3,462); restructuring leads at 16,173Tech leads YTD cuts at 155,126AI still leads YTD at 116,175 (~22% of all cuts)Cautious
MITAcademic11.7% automatable; 5× gapHighest exposure; cost constraints-13% since 2022Cautious
MIT NANDAAcademic95% of AI pilots fail; GenAI DivideIntegration, not models, is the bottleneckNot measuredExecution crisis
Stanford DELAcademic16% decline (ages 22–25)Software dev, customer svc hardest hitCanaries confirmed; now 16%Concerned
Stanford SALT LabAcademicH3 partnership most-desired; 844 tasksWorkers want augmentation, not removal47.5% want more agency than expertsWorker-centered
OxfordAcademic47% at risk (original)Geography reversingNot focusFoundational
BrookingsThink TankNo apocalypse yetMetro most exposedMild hiring difficultySkeptical
McKinseyConsultancy57% automatable; $2.9TAdmin most exposedNot focusBullish value
AccentureConsultancy1.1M supply-chain gap; ~6% net at full AITask-level model; 52% augment-onlyPeople repositioned, not replacedRedesign imperative
BCGConsultancy72% managers; 51% frontlineAdoption gapShadow AI riskUneven
PwCConsultancy4× productivity; 56% premiumWages growingPremium at all levelsMost optimistic
WEFMultilateral+78M net; 22% churnTech fastest growing39% skill transformNet positive
IMFMultilateral40% global; 60% advancedHigh-wage cognitiveCollege-educated adaptInequality risk
OECDMultilateral27% high-risk; task-partialCognitive most exposedLow-skilled vulnerablePolicy urgent
AnthropicFrontier AILimited employment impact; augmentation ↑Task diversification; skill-biasedHiring slowing; not displacingMeasured caution
OpenAIFrontier AI80% workers affectedWriting, coding19% face 50%+Opportunity
Goldman SachsFinance6–7% displacedTech hiring declining3pp rise (20–30 tech)Transitory
MicrosoftTech82% plan agentsKnowledge transforming66% won't hire w/o AIAgent era
GartnerAnalyst32M/year by 20280% IT unaugmented 2030Mentoring breakingChaos, not apocalypse
Dallas Fed Iss.7Central BankCodified vs. tacitWages +16.7%Employment -5%Dual impact
SF FedCentral BankProductivity paradox returnsLimited macro AI effectNot measuredCautious
NBER Iss.8Academic90% execs: no impact yetSolow paradox returnsNot measuredAdoption lag
CEPR Iss.9European Research+4% productivity; no job lossTraining = 5.9× multiplierNot measuredInvestment-dependent
Google Econ Iss.9Tech/EconomicsRate hikes, not AITiming challenges StanfordDisputes AI causationOptimistic
Citrini Iss.9Finance/ScenarioGhost GDP; 38% crashWhite-collar spiralFirst casualtiesAlarmed
Citadel Iss.9Finance/RebuttalJobs up 11%; Keynes lessonDemand risingNot addressedBullish
PIIE Iss.9Policy Think Tank"Still in the first inning"Contradictory findingsDataset-dependentUncertain
Duke/NBER/Fed Iss.9Academic/Fed502K AI layoffs projected, 9× (modeled expectation)Cutting on expectation, not resultsPrimary targetsExpectation-driven
Tufts Digital Planet Iss.9Academic9.3M jobs at risk; $757B incomeWriters 57%, Programrs 55%Geographic concentrationRisk-mapping
HBS / Srinivasan Iss.9AcademicAutomatable −17%; Augment +22%Task-level recomposition18M entry jobs at risk (w/ Burning Glass)Task-level
ADP Research Iss.9Workforce Data22% feel safe; 39K workers surveyedFrontline 18% feel safe; C-suite 35%Universal anxietyPsychological crisis
MIT CSAILAcademic11,500 tasks; "rising tide"47–73% task success rangeSeveral years from full automationTask-level validation
Stanford HAI 2026Academic53% GenAI adoption; 400+ pg reportExpert-public gap: 73% vs 23%Devs 22–25 down 20%Nuanced
GallupWorkforce Data50% never use AI; adoption paradoxLarge orgs: more cuts than hires46% prefer current methodsAdoption gap
Oracle (Event)Corporate21K cut (13%); AI named in SEC filing"Reductions to our workforce" — 10-KReallocation now on the recordDisclosure era
NewmarkReal EstateOffice employment flat (+0.3%) thru 2030First non-recession flat since 1944AI headwind to office demandStructural shift
Indeed Hiring LabLabor DataData-center postings surge as tech postings fall71% of postings from top 10 firms; ~25% install & maintenanceDemand shifting, not vanishingRecomposition
RandstadStaffingRobotics techs +107%; cooling +67%; automation +51%Constraint is specialized talent, not chips or capitalTrades demand rising sharplyShortage-led
NACELabor DataClass of 2026 hiring projected +5.6%35% of entry-level jobs now require AI skillsGraduate demand holdingCounter-signal
German Estab. Panel I17AcademicAI adoption → +14% new apprenticeshipsIntensive margin only; no effect on whether firms trainTraining intensifies where it existsInstitution-dependent
Yale Budget Lab I17AcademicOccupational mix shifting no faster than PC/internet eraNo link between AI exposure and employmentSoftening looks cyclical, not structuralSkeptical of shock
Ramp / Revelio Labs I17Labor DataHigh-intensity AI spenders: headcount +10.2%, entry-level +12%~22,000 firms; concentrated in well-funded tech-forward firmsSpend intensity predicts growthCounter-signal
Robert Half I17Labor Data66% increasing H2 hiring; 63% delayed, 48% canceled projectsHardest to find: domain knowledge, software, leadershipCapability shortage bindingConstraint shifted
US Dept. of Labor I17Government~$243M into AI-focused Registered ApprenticeshipsAI Literacy Framework; Innovation Portal; no degree requiredPolicy rebuilding the on-rampInstitution-building
← PreviousThe Debate
8 of 13
All sections
Next →Industry Pulse
Industry Pulse

Nine Sectors, Nine Different Wings

The geometry holds everywhere. The placement does not. Every sector card carries a headline signal, its position on the Human Agency Scale, and the direction that sector is taking on junior intake. Select any sector for the full read, including a new Where the Wings Go assessment: the band in that sector where closeness, expertise and judgment concentrate, and how far down the structure it sits.

At a glance — displacement risk across all nine sectors
Data Ctr
Health
Pharma
Public
Travel
Retail
Fin Svcs
Prof Svcs
Tech
Lower riskHigher risk
Financial Services & Banking
Entry roles −35% since 2023 across law, consulting & banking
Narrowing
Professional Services & Legal
PwC planning ~1/3 fewer associates by 2028
Narrowing
Media, Technology & High Tech
149,023 cuts YTD, +67% — yet IBM is tripling junior intake
Redesigning
Data Centers & AI Infrastructure
~25% of openings are install & maintenance, +42% pay
Broadening
Health Care
Licensure keeps the apprenticeship structurally intact
Broadening
Pharmaceuticals & Life Sciences
~6% modeled maximum supply-chain scenario — not realized
Redesigning
Consumer Products & Retail
Walmart's $1B skills-based pathway is the largest on-ramp rebuild
Broader access
Public Services & Government
Graduate schemes frozen while AI literacy becomes a criterion
Narrowing
Travel & Hospitality
Front line resilient; the rung above it is thinning
Redesigning
← PreviousSignal Map
9 of 13
All sections
Next →Global & Regulatory

Financial Services & Banking

The sector sits at the intersection of the Citrini/Citadel debate — because it is both an AI beneficiary and a potential casualty. Brookings reports the industry is shifting from credentials (MBAs, CFAs) to practical AI fluency. Morgan Stanley advisors now work alongside GPT-4-powered copilots trained on proprietary knowledge. At Klarna, support teams have been reorganized into hybrid human-AI units. BNP Paribas ESG analysts use GenAI to synthesize sprawling unstructured data.

PwC projects up to 50% productivity boost through human/AI collaboration in the back-office. Bloomberg Intelligence projects up to 200,000 back- and middle-office roles could be cut over 3–5 years (~3% of sector workforce). Wells Fargo's 2026 headcount planning already factors in a smaller workforce. Citi reports 9% software development productivity improvement. McKinsey estimates genAI could deliver $200B–$340B in annual value for banking. 50 of the world's largest banks announced 160+ AI use cases in 2025 alone.

Among the most critical findings: NVIDIA's 2026 survey of 800+ financial services professionals found 89% say AI has helped increase revenue and decrease costs. Yet the MIT NANDA report found the sector has among the highest internal-build failure rates — purchased solutions succeed ~67% of the time vs. ~33% for internal builds.

Impact Assessment
Displacement Risk
7/10
New Role Creation
6/10
AI Adoption Speed
8/10
Reskilling Urgency
9/10
Agency Read — Where the Work Wants to Sit

H2–H3 Financial services splits sharply: high-volume back- and middle-office tasks sit low on the scale, while advisory, risk sign-off, and client trust demand partnership or more. The win is automating the screening — not the judgment.

Green-light · KYC/AML screening · reconciliation · report draftingRed-light · advisory judgment · model-risk sign-offR&D · real-time risk explanation
Role Impact Map
Roles at Risk
  • Claims processors & routine underwriters
  • Compliance checkers (KYC, AML screening)
  • Basic financial analysts & report writers
  • Middle-office operations & reconciliation
  • IT support & ticket-based service desks
Roles Growing
  • AI risk governance & model validation
  • Algorithmic auditing & fairness testing
  • Hybrid analyst-engineer roles
  • AI ethics & regulatory compliance
  • Prompt-to-trade systems architects
Where the Wings Go · Issue 18

The wing forms around credit judgment and keeping relationships intact, and it sits deep in the structure rather than near the top. Reconciliation, first-pass memos and covenant checks are the clerical band, the one occupation group where replacement really does win. What survives and widens is the banker who can price an exception, read a borrower in a bad quarter, and defend a decision to a regulator. Orchestration sits just below, with analysts directing model output, checking it, and owning the escalation. The waist is the review layer that existed to aggregate and forward. Test: if your credit committee is the only place judgment lives, you have a top band, not a wing.

Talent Strategist's Outlook: Financial services will be among the first sectors to fully recompose its workforce around human-agent teams. The transition from credentials to capability is already underway. The biggest risk isn't job loss — it's the speed of role redefinition outpacing reskilling programs.

Professional Services & Legal

The sector with the most to lose and the least structural protection. Consulting, law, audit and tax are built on a leverage pyramid in which junior capacity is simultaneously the product, the margin and the training system — which is why AI hits all three at once. Reporting in July 2026 put entry-level roles across law, consulting and investment banking down roughly 35% since 2023.

The public commitments diverge sharply from the public rhetoric. PwC argues forcefully that AI is removing the routine work through which junior employees learned by doing, and that entry roles now require judgment, creativity, leadership and interaction much earlier — while its US organization has reportedly planned to reduce traditional tax and assurance associate hiring by roughly one-third by 2028. Baker McKenzie has cut 1,000 roles. McKinsey has piloted AI-fluency interviews. PwC's own labor-market analysis found AI-exposed entry-level positions in the US were seven times more likely to require capabilities traditionally associated with more experienced employees.

The counter-move exists and it is not sentimental. Accenture reports apprentices at 20% of US and Canada entry-level hiring across the past four fiscal years, with more than 2,500 hired since 2016. Cognizant plans 24,000–25,000 fresh graduates in 2026 — about 20% more than 2025 — on an explicit "broader pyramid" strategy that uses AI to make juniors productive on more complex client work sooner. Same technology, opposite pyramid.

Impact Assessment
Displacement Risk
8/10
New Role Creation
5/10
AI Adoption Speed
9/10
Reskilling Urgency
10/10
Agency Read — Where the Work Wants to Sit

H3–H4 The deliverable is advice, and advice carries accountability that cannot be delegated to a model. Research, synthesis and first drafts sit low on the scale; scoping, client judgment, opinion and sign-off sit high. The trap is that the low-scale tasks were the curriculum.

Green-light · research · document review · first-draft synthesisRed-light · audit opinion · legal advice · client judgmentR&D · supervised junior validation workflows
Where the Wings Go · Issue 18

The wing sits on the floor and in the yard, not in the planning office. Routing, scheduling and load building are the ordinary case and they automate fast. What does not automate is the exception: the damaged consignment, the compliance stop, the customer whose problem does not match any ticket category. This is where the base tug-up is most visible and most misread. The operative who becomes an orchestrator of automated flow is doing a materially different job at the same grade. Redeployment, not recruitment. The peak band is already detachable here, which is why the new geometry shows up in logistics earlier than almost anywhere else.

Role Impact Map
Roles at Risk
  • First-year audit and tax associates
  • Document review and discovery paralegals
  • Research analysts and deck builders
  • Due-diligence and data-room reviewers
  • Proposal and RFP drafting support
Roles Growing
  • AI assurance and model-audit specialists
  • Junior validators of machine-generated work
  • Client-facing analysts from year one
  • Legal-engineering and workflow design
  • Apprenticeship and capability leads

Talent Strategist's Outlook: Professional services will resolve this before anyone else because the economics force it — a leverage model cannot quietly lose its leverage layer. The firms that win the 2030s will be the ones that redesigned the associate role around supervising and validating machine output, paid and promoted it as the harder job it is, and held a floor under intake through the cycle. The firms that treated junior headcount as a cost line will discover around 2032 that they have no partners to promote.

Media, Technology & High Tech

Ground zero for AI workforce impact — and the Oracle/IBM fork is now the sector's defining case study. Oracle's 30,000-person cut (March 31) — the largest single AI-attributed layoff in history — was a profitable company ($3.7B quarterly income) eliminating humans to fund $156B in AI infrastructure. Wall Street rewarded it. Cornell research found writing and translation demand on freelance platforms fell 20–50% in AI-substitutable categories. Yet Citadel/Indeed data shows software engineering postings up 11% YoY.

2026 trackers: tech layoffs have passed ~186,000 year-to-date at ≈978/day — roughly double the 2025 pace, with 56% of events citing AI. Oracle (21K, AI named in its SEC filing), Amazon (16K), Dell (11K), Block (4K), Meta (2K+), Atlassian (1.6K) lead the cascade. IBM's decision to triple young hires — and redesign every entry-level role around what AI can't do — is the counter-signal that may define the winning talent strategy. Dallas Fed data shows computer systems design employment down 5% since ChatGPT launch, while wages in the same sector rose 16.7% — the clearest example of the codified/tacit split.

Impact Assessment
Displacement Risk
9/10
New Role Creation
8/10
AI Adoption Speed
10/10
Reskilling Urgency
10/10
Agency Read — Where the Work Wants to Sit

H3 desired vs H1 deployed — the agency gap is at its widest here. Leaders are racing toward full automation, yet the Dallas Fed shows the tacit, system-level judgment of senior engineers is exactly what commands a wage premium. The deployment marker (left) sits well below where the work — and the value — actually wants to be.

Green-light · QA · boilerplate code · content first-drafts · tier-1 ITRed-light · system architecture · senior (tacit) engineeringR&D · human-AI workflow design
Role Impact Map
Roles at Risk
  • Content writers & marketing copywriters
  • Junior/mid-level software developers
  • QA testers & manual code reviewers
  • IT support & tier-1 helpdesk
  • Data entry & basic data analysis
Roles Growing
  • AI systems architects & MLOps engineers
  • Prompt engineers & AI interaction designers
  • AI product managers & strategy leads
  • Responsible AI & AI safety roles
  • Human-AI workflow designers
Where the Wings Go · Issue 18

The wing is the line supervisor with deep process knowledge, and it widens for a reason the sector keeps relearning. Automated quality systems catch the faults they were trained on. Ford's three-year round trip with veteran engineers is the reference case, and the lesson is not that automation failed, because the systems stayed. Experienced people working with automation is simply a better configuration than either one alone. The deep-expert waist is narrow, specialized and almost never replaced. Engineering carries the lowest replacement ratio in the survey at 0.10. Scheduling and production admin is where the clerical contraction lands.

Talent Strategist's Outlook: Tech is experiencing the most visible and rapid recomposition — and the Oracle/IBM fork is the clearest distillation of the choice facing every company. Oracle chose extraction: cut humans, fund machines, show improved margins. IBM chose investment: redesign roles, triple young hires, build the pipeline. The Dallas Fed's codified/tacit framework is most predictive here: junior engineers with textbook knowledge face displacement while senior engineers with system-level judgment command premium wages. MIT's task-level data (47–73% success range) gives leaders the granular evidence to redesign, not just reduce.

Data Centers & AI Infrastructure

The physical foundation of the AI economy is booming — and it's labor-intensive. Citadel specifically cited data center construction as driving localized hiring booms that counter the displacement narrative. OECD reports 94% of construction firms face difficulty sourcing workers. This sector perfectly illustrates DeepMind's cognitive-physical divide: AI automates knowledge work, but the physical infrastructure to run AI requires human construction workers, electricians, and engineers.

Stanford HAI's James Landay warned that "you can't tie up all the money in the world on this one thing" — flagging the speculative bubble risk. But in the near term, AI infrastructure investment is one of the purest job creation stories in the economy. WEF reports 40% of young graduates are now choosing non-automatable careers in construction, plumbing, and electrical work — partly driven by the data center boom.

Impact Assessment
Displacement Risk
1/10
New Role Creation
9/10
AI Adoption Speed
3/10
Reskilling Urgency
4/10
Agency Read — Where the Work Wants to Sit

H4–H5 The physical foundation of AI is human-essential: construction, electrical, and cooling work is low-capability for AI and high-desire for people. Near-pure augmentation, minimal displacement.

Green-light · facility monitoring · cable labelingEssential-human · trades · electrical · cooling/HVAC
Role Impact Map
Limited Risk
  • Routine facility monitoring (some automation)
  • Basic cable management & labeling
Roles Growing
  • Construction trades & heavy equipment operators
  • Electrical engineers & power grid specialists
  • Cooling system & HVAC engineers
  • Fiber-optic technicians & network architects
  • Site reliability engineers (SREs)
  • Facility managers & sustainability leads
Where the Wings Go · Issue 18

The most distorted shape in the economy, and the least safe to generalize from. Construction and commissioning absorb enormous labor while the operating footprint per megawatt stays thin, which creates a temporary base that is a project workforce rather than a permanent one. The durable wing is reliability and incident engineering, the people who own the two-in-the-morning anomaly no runbook covers. Everything else here is orchestration of heavily instrumented systems. Do not read this sector's capex-driven headcount as a signal about anyone else's workforce.

Talent Strategist's Outlook: Data centers are the rare net job creator in the current AI cycle. The irony is profound: the technology that displaces knowledge workers requires an army of physical workers to build and maintain. This sector is the strongest argument for the trades pipeline and a key recruitment target for displaced white-collar workers willing to retrain.

Health Care

Health care is one of the clearest augmentation stories. Stanford HAI reports AI-powered medical devices have increased 37× since 2015. Chronic staffing shortages mean AI fills gaps rather than displacing workers. Administrative tasks consume 37% of physician time — the primary target for AI intervention. Nurses using AI can now perform work previously limited to doctors, expanding scope of practice.

Brynjolfsson's research found AI resolved 14% more issues per hour in healthcare call centers. OECD reports 94% of healthcare systems and construction firms face severe difficulty sourcing workers. Stanford HAI warns of a "tsunami of noise" as hospitals are inundated by AI startup pitches — a typical hospital receives hundreds of vendor proposals with limited evaluation frameworks to assess them. The need is not more AI tools; it's better AI governance and integration capability.

Impact Assessment
Displacement Risk
2/10
New Role Creation
8/10
AI Adoption Speed
5/10
Reskilling Urgency
6/10
Agency Read — Where the Work Wants to Sit

H3–H5 Staffing shortages make AI additive, so workers welcome it on the administrative load — but diagnostic accountability and patient care are essential-human and must stay there.

Green-light · coding/billing · scheduling · routine pre-screeningRed-light · diagnostic accountability · patient careR&D · clinical AI validation
Role Impact Map
Roles at Risk
  • Medical coding & billing specialists
  • Radiology pre-screening (routine reads)
  • Administrative scheduling & records
  • Insurance claims processing
Roles Growing
  • AI-augmented clinicians & diagnosticians
  • Health informatics & clinical data scientists
  • Remote patient monitoring specialists
  • AI safety, evaluation & clinical validation leads
Where the Wings Go · Issue 18

The wing is at the bedside and it is not moving, which is why healthcare keeps showing up in this publication as the counter-example. Documentation, prior authorization, coding and scheduling are the clerical band and they are shrinking fast. Clinical judgment, patient closeness and handling ambiguity make up the widest human layer in the sector and arguably in the economy. The distinctive risk here is not displacement but the waist: care coordination and utilization review functions whose core work is moving information between parties, and which will get absorbed whether anyone plans it or not.

Talent Strategist's Outlook: Health care is uniquely positioned — staffing shortages mean AI is additive, not substitutive. The sector's challenge is governance and integration, not displacement. Invest in clinical AI evaluation frameworks and data infrastructure before scaling AI tools.

Pharmaceuticals & Life Sciences

One of the clearest augmentation stories in the AI landscape. AlphaFold (Google DeepMind) won the Nobel Prize for Chemistry, demonstrating AI's transformative potential in drug discovery. Domain expertise remains essential — but AI dramatically accelerates the research cycle from years to months. Clinical trial design, regulatory document preparation, and pharmacovigilance are all being transformed.

The IMF found top researchers using AI boosted output by 44%, but 82% reported less job satisfaction due to diminished creativity — the augmentation trade-off in action. Stanford HAI warns of a "tsunami of noise" as hospitals and pharma companies are inundated by AI startup pitches with limited evaluation frameworks. Accenture's biopharma analysis found approximately 55% of total workforce hours are impacted by digital and physical agents across a biopharma enterprise.

Impact Assessment
Displacement Risk
3/10
New Role Creation
7/10
AI Adoption Speed
7/10
Reskilling Urgency
6/10
Agency Read — Where the Work Wants to Sit

H3–H4 The clearest augmentation story — and a cautionary one: the 44% output gain came with an 82% drop in satisfaction, exactly the signal that scientists want to keep creative agency, not cede it.

Green-light · regulatory drafting · routine assays · PV case processingRed-light · scientific creativity & judgmentR&D · AI clinical-trial design
Role Impact Map
Roles at Risk
  • Regulatory document drafters
  • Basic lab technicians (routine assays)
  • Pharmacovigilance case processors
  • Medical writing (first-draft level)
Roles Growing
  • AI-augmented researchers & computational biologists
  • AI clinical trial designers
  • Real-world evidence analysts
  • AI safety & validation leads (GxP)
Where the Wings Go · Issue 18

The wing forms around qualification, deviation and regulatory judgment, several levels below the executive team, in the people who decide whether a batch ships and can defend that call to an inspector. Document generation, literature review and submission assembly are the shrinking band. This sector shows the central claim as clearly as any. Two competitors can run identical models over identical data, and the difference in what they get out depends entirely on where the judgment sits relative to the process. That placement is not for sale and it does not transfer between companies.

Talent Strategist's Outlook: Pharma is the model for "AI as accelerator, not substitute." The 44% productivity boost / 82% satisfaction decline trade-off must be managed explicitly. Companies that redesign roles around augmentation rather than automation will attract and retain top scientific talent.

Consumer Products & Retail

Retail faces a dual dynamic: back-office and supply-chain AI is advancing rapidly, while customer-facing physical roles remain resilient. Citrini's scenario specifically highlighted the risk of agentic commerce — where machine-to-machine purchasing bypasses human consumers entirely. Citadel countered that humans generate demand; machines do not.

WEF projects delivery drivers and warehouse workers among the fastest-growing categories globally. Cornell research found writing/translation demand on freelance platforms fell 20–50% in AI-substitutable categories — affecting the content marketing engine that drives retail. The sector is seeing AI deployed in demand forecasting, dynamic pricing, inventory optimization, and personalized marketing at scale.

Impact Assessment
Displacement Risk
5/10
New Role Creation
5/10
AI Adoption Speed
6/10
Reskilling Urgency
6/10
Agency Read — Where the Work Wants to Sit

H2–H3 Demand forecasting and campaign drafting are genuine green-light work; brand, experience, and last-mile human roles are where workers — and customers — want people kept firmly in the loop.

Green-light · demand forecasting · inventory · copy first-draftsRed-light · experience design · store & brand craft
Role Impact Map
Roles at Risk
  • Merchandising analysts & demand forecasters
  • Customer service agents (chat/phone)
  • Back-office inventory & logistics planners
  • Marketing copywriters & email campaign managers
Roles Growing
  • Last-mile logistics & delivery operations
  • Experience design & store innovation
  • AI-powered personalization managers
  • Supply-chain orchestration & resilience leads
Where the Wings Go · Issue 18

The wing is at the assortment and category call, close to the customer and close to the shelf. Demand planning, promotional analysis and reporting are the clerical contraction. What widens is the person who can override the forecast for reasons the forecast cannot see: a local event, a competitor move, a product failing for a reason nobody has coded yet. This is the clearest case of a wing that sits low in the structure, at store and category level, well below where most operating models put decision rights. Companies that keep that authority at head office get the shape wrong even when they get the technology right.

Talent Strategist's Outlook: The physical-digital split will define retail workforce strategy. Expect continued contraction in back-office and content roles alongside strong demand for physical fulfillment and experience-driven positions. Agentic commerce remains the wild card.

Public Services & Government

Government is the slowest-adopting sector but faces enormous transformation potential. OpenAI reports Pennsylvania state workers save 95 minutes/day and teachers save 6 hours/week. The US Department of Labor issued an AI literacy mandate in February 2026 for all workers. But public sector adoption is constrained by procurement cycles (18–36 months), extreme risk aversion, and intensive regulatory oversight.

Only 17% of firms overall use AI (Fed Reserve), and government adoption is even lower. Constituent services, benefits administration, fraud detection, and policy analysis are all highly automatable — but the political and institutional barriers to change are the highest of any sector. The opportunity cost of inaction is enormous: hundreds of billions in operational inefficiency preserved by institutional inertia.

Impact Assessment
Displacement Risk
3/10
New Role Creation
4/10
AI Adoption Speed
2/10
Reskilling Urgency
5/10
Agency Read — Where the Work Wants to Sit

H3 Highly automatable on paper, but accountability and citizen trust pull the desired level up — and procurement and governance gate the capability. Partnership, governed, is the realistic destination.

Green-light · eligibility processing · records · tier-1 inquiriesRed-light · policy judgment · public accountability
Role Impact Map
Roles at Risk
  • Claims processors & eligibility reviewers
  • Data entry & records management
  • Routine compliance & audit checkers
  • Constituent inquiry handling (tier 1)
Roles Growing
  • AI governance & public accountability specialists
  • Digital service designers
  • Algorithmic impact assessment officers
  • AI-enabled policy analysts
Where the Wings Go · Issue 18

The wing is at eligibility determination, appeals and casework, the points where a rule meets a citizen whose circumstances the rule did not anticipate. Form processing, routing and status inquiry are the shrinking band, and the sector is automating them faster than its reputation suggests. Two structural differences are worth naming. The waist is thicker here than anywhere else, because coordination layers are often required by statute rather than by function and cannot simply be absorbed. And the detachable peak band runs into procurement rules that were not written with orchestration in mind.

Talent Strategist's Outlook: The gap between government AI potential and adoption speed is the largest of any sector. Early movers (like Pennsylvania) are demonstrating extraordinary time savings. The challenge is scaling these successes across institutional barriers. Expect 3–5 year lag behind private sector.

Travel & Hospitality

High-touch industry where human interaction is the product. AI agents are transforming booking, dynamic pricing, and concierge services. The OECD finding that jobs involving physical tasks and human emotions are least affected directly benefits this sector. WEF data shows 52% of professionals now view trades and hospitality as less vulnerable than white-collar work.

Risk concentrates in the intermediary layer: traditional travel agents, call center support, administrative booking, and revenue management analysts. But the experiential core — hotels, restaurants, curated travel — retains a strong moat. Deloitte projects 36% of managers will manage digital agents within 5 years, which in hospitality means AI concierge and booking systems overseen by human experience curators.

Impact Assessment
Displacement Risk
4/10
New Role Creation
4/10
AI Adoption Speed
4/10
Reskilling Urgency
5/10
Agency Read — Where the Work Wants to Sit

H3–H4 Hospitality is a high-touch product: the booking and revenue-analytics layer automates well, but the guest relationship is the value and sits firmly in partnership-and-above.

Green-light · booking · revenue-management analytics · tier-1 supportRed-light · guest experience · concierge relationships
Role Impact Map
Roles at Risk
  • Booking agents & reservation managers
  • Revenue management analysts
  • Call center & email support
  • Back-office admin & scheduling
Roles Growing
  • Experience curators & brand storytellers
  • AI-augmented concierge leads
  • Personalization & loyalty designers
  • Sustainability & wellness managers
Where the Wings Go · Issue 18

The wing is the moment of recovery: the delayed flight, the failed booking, the guest whose stay has gone wrong. Reservation handling, itinerary changes and standard inquiries are the ordinary case and they automate almost completely. What is left is unusually high-stakes and decides whether the customer comes back, which means this sector concentrates more differentiating value in fewer human interactions than any other in this set. The orchestration band is thin and the remaining execution band stays large for physical reasons. The detachable peak band has been native to the sector for decades.

Talent Strategist's Outlook: The human touch is the product in hospitality. AI will transform the operational backbone but amplify — not replace — the guest experience layer. Focus reskilling on blending AI fluency with emotional intelligence.

Global & Regulatory

The Global View

Issue 18 makes the regulatory column the lead story, because on 1 October the United States acquires its first statutory AI-attribution field. Connecticut SB 5 requires every WARN-standard mass-layoff filing to state whether the cuts are related to the employer's use of AI or other technological change; California's Executive Order N-6-26 instead directs agencies to measure first and publish a dashboard; the bipartisan GAAIA discussion draft would add an equivalent federal disclosure where AI is a "substantial factor." None of the three defines the term. Also live: California's AI Transparency Act, operative since 2 August 2026, with further California bills on automated decision systems and AI-driven mass layoffs pending; Illinois SB 315 and Colorado SB 26-189 from 1 January 2027; and the Colorado AI Act itself enjoined pending preliminary-injunction proceedings. The EU AI Act continues to treat recruitment and worker management as high-risk. Internationally, the OECD Employment Outlook 2026 supplies the most important finding of the cycle for anyone thinking about shape. Adjustment to trade and technology shocks runs through people falling into joblessness and through opportunities for new entrants. It does not run through displaced workers moving into the roles that get created, and the scarring lasts. That is rotation failing at national scale, and it is the strongest warning available that a shape change left to the market does not move people. It replaces them.

70%

North America

Click for full regional analysis →
4%

Europe

Productivity up, jobs stable →
1:3.6

Asia-Pacific

Talent hunger, fear paradox →
26%

Emerging Markets

The exposure gradient →
70%

North America — The Buyout Quarter Hits Home

North America's tech sector is now in its largest concentrated workforce displacement wave in over a decade. Meta's 8,000 cuts (May 20), Microsoft's 8,750 buyouts (first ever), and continued Oracle restructuring stack on top of 186,000+ tech jobs cut YTD. The funding mechanism is now visible: $725B in 2026 Big Tech capex (up 77% YoY) — more than the entire global oil and gas industry spends on exploration. Goldman Sachs: AI suppressing ~16,000 U.S. jobs per month, concentrated in Gen Z white-collar entry roles. BLS March jobs report: 4.3% unemployment, +178K nonfarm payrolls, but information sector lost 3K and financial services lost 15K. Newmark: office-using employment essentially flat through 2030 — first time outside a recession since 1944.

4%

Europe — Productivity Up, Jobs Stable

CEPR study of 12,000+ firms found AI boosts productivity ~4% with no job losses — but gains depend on training investment (5.9× multiplier). EU trails US in AI patents. The EU AI Act regulates development but doesn't address displacement. Carnegie Endowment calls for a dedicated EU labor transition framework. The European Globalisation Adjustment Fund — just €35M/year — is widely seen as inadequate.

1:3.6

Asia-Pacific — The Offshore Repricing Hub

The capital reallocation story has a cross-border dimension. Bloomberg analysis finds roughly half of "AI-attributed" U.S. cuts result in the same roles being rehired offshore at lower wages — turning the Asia-Pacific region into the labor repricing destination of the buyout quarter. India absorbed 12,000 of Oracle's 30,000 March 31 cuts (40%) — Bengaluru, Hyderabad, NetSuite IDC. Most severe AI talent shortage globally (1:3.6 ratio). BCG: countries with highest AI usage (63% Middle East, 48% India) also report highest job-loss fear. Singapore tops Stanford's AI adoption rankings at 61%. China has nearly closed the model performance gap with the U.S. The same forces that fund $725B in U.S. capex are creating an offshore engineering boom — at U.S. job cost.

26%

Emerging Markets — The Exposure Gradient

IMF data shows AI affects ~26% of jobs in low-income countries vs. 60% in advanced economies — but with less capacity to adapt. Geopolitical export controls could slow AI rollout. WEF projects farmworkers as top absolute growth category. The risk: a widening digital divide where AI benefits concentrate in high-income nations.

← PreviousIndustry Pulse
10 of 13
All sections
Next →Voices
The Human Element

Thought Leader Spectrum

The researchers, executives and thinkers shaping the debate, scored by disruption outlook. Issue 18 adds four. Three speak to the attribution problem: the Atlanta Fed's Salomé Baslandze, lead author of the CFO study behind this issue; Revelio Labs' Lisa Simon, who argued AI was partly a cover story before August made the case for her; and Oxford Economics' Ben May, on why a retrenchment gets relabeled as a strategy. The fourth is Stephen Wroblewski, who writes this publication and is listed here so readers can score the author on the same scale as everyone else. Challenger, the Budget Lab and IBM's LaMoreaux are all refreshed for this issue.

Research Pioneers
SB
Salomé Baslandze
Federal Reserve Bank of Atlanta & CEPR, lead author on the CFO Survey AI study
Nuanced
The productivity comes from growth motives, not cost motives — and the workforce effect is composition, not level
Lead author on the Atlanta/Richmond Fed and Duke survey of roughly 750 corporate executives, the most directly decision-relevant AI evidence published this year. Three findings anchor Issue 18. Executives report AI productivity gains around three times what their own revenue and employment numbers imply. The gains correlate with innovation and demand motives — developing products, serving customers — while cost-reduction motives show imprecise and frequently negative coefficients. And near-term employment effects are small in aggregate while occupational composition shifts sharply: routine clerical down 2.19 points by 2028, skilled-technical up 1.35, with large firms shedding and smaller firms adding. The paper is a working paper without peer review, and its core method is executive self-attribution — a reason code — which is why this issue leans on its relationships rather than its levels.
NBER Working Paper 34984
Disruption
40/100
Added: Issue 18 — September 2026
GH
Geoffrey Hinton
Nobel Laureate, "Godfather of AI"
Alarmed
AI will "replace many, many jobs" in 2026
Capabilities doubling every 7 months. Recommends plumbing and nursing as AI-resistant careers.
Disruption
92/100
BBC Interview
Last updated: Issue 7 — March 2026
EB
Erik Brynjolfsson
Stanford Digital Economy Lab Director
Cautious
"We are flying blind into one of the most consequential periods in world history"
Canaries lead author. July 2026 refresh: the 22–25 divergence in exposed occupations has widened ~0.5pp/month and shows no reversal. Notably rigorous about his own limits — published a February update conceding the timing only becomes significant from 2024 and framing the dashboard as correlational.
Disruption
78/100
Stanford Canaries Paper
Last updated: Issue 15 — July 2026
NR
Nela Richardson
Chief Economist, ADP Research
Cautious
Co-leads the Canaries Dashboard — and insists on what it doesn't prove
Brings ADP's payroll microdata (25,000 firms, 4.6M matched workers, 730+ occupations) to the entry-level question. Her contribution to the July 2026 cut is the discipline around interpretation: the Gen Z gender gap is attributed primarily to occupational sorting, not AI. A useful corrective in a field prone to overclaiming.
Disruption
64/100
Canaries Dashboard
Last updated: Issue 15 — July 2026
DA
Dario Amodei
Anthropic CEO
Alarmed
AI is a "general labor substitute for humans"
Calls for "surgical intervention" by the state. Predicted 50% of entry-level white-collar jobs wiped out within 5 years. Anthropic valued at ~$380B.
Disruption
88/100
Fox/Axios Interviews (March 2026)
Last updated: Issue 9 — March 2026
SR
Stuart Russell
UC Berkeley, Human Compatible
Alarmed
AI safety is existential; displacement is a symptom
Argues the alignment problem and economic disruption are fundamentally linked. Among the most alarmed voices in the field.
Disruption
98/100
Human Compatible
Last updated: Issue 6 — February 2026
DY
Diyi Yang
Stanford SALT Lab — WORKBank Lead PI
Nuanced
Ask the workers — they want partnership, not replacement
Yang's team built WORKBank by auditing 844 tasks across 104 occupations with 1,500 workers and 52 experts — the first large-scale measure of desired human agency. The finding reframes the debate: workers welcome automation on 46% of tasks, but the most-desired state is Equal Partnership (H3), and on 47.5% of tasks they want more human agency than experts deem necessary. The value is in augmentation design, not subtraction.
Disruption
45/100
Stanford WORKBank
Last updated: Issue 14 — July 2026
Industry & Market Voices
LS
Lisa Simon
Revelio Labs — Chief Economist
Cautious
AI has been partly a front and an excuse for cuts firms wanted to make anyway
One of the few labor economists with firm-level workforce records who has consistently pushed back on the attribution itself rather than the technology. Her position — that some share of AI-cited layoffs are corrections to overhiring dressed in a more flattering explanation — was a minority read when she made it and looks considerably stronger after August, when AI dropped from first to fourth among stated reasons with no change in the underlying technology. She is also a co-author on the Ramp/Revelio work showing high-intensity AI spenders growing entry-level headcount, which makes her unusually well positioned: the same researcher arguing both that the attribution is unreliable and that the heaviest adopters are hiring.
Compiled AI-layoff attributions and economist commentary
Disruption
48/100
Added: Issue 18 — September 2026
BM
Ben May
Oxford Economics — Director of Global Macro Research
Cautious
Some firms are dressing up layoffs as a good-news story rather than a bad one
May has argued through 2026 that technological change is being cited in place of less flattering explanations — post-pandemic overhiring, demand softness, rate pressure — because it reframes a retrenchment as a strategy. The claim is difficult to test directly and he does not overstate it. What makes it load-bearing for this issue is the timing: a reason that can move four places in a single month is behaving the way an explanation behaves, not the way a cause does. From 1 October that same explanation becomes a field on a state filing, with no statutory definition behind it.
Compiled AI-layoff attributions and economist commentary
Disruption
45/100
Added: Issue 18 — September 2026
FM
Fabien Curto Millet
Google Chief Economist
Optimistic
"The most profoundly transformative technology" — but displacement is premature
Challenged Stanford's Canaries paper directly. Argues timing tracks rate hikes, not ChatGPT. Sees "micro multinationals" and new work categories emerging.
Disruption
40/100
LSE Business Review Paper
Last updated: Issue 8 — March 2026
JH
Jensen Huang
Nvidia CEO
Pragmatic
Companies cutting jobs to AI are short of imagination, not short of people
Huang has become the loudest industry voice against the displacement narrative, and he argues it from two directions. On the mechanism, at GTC he told Jim Cramer that companies announcing cuts and crediting his chips are out of imagination, because firms with ideas do more with more while firms without them leave the extra capability unused. That is the Fed CFO regression in plain language: growth motives produce the measured productivity, cost motives do not. On the attribution, he told Channel NewsAsia in May that blaming AI for layoffs is lazy, since the tools only became broadly useful recently and cuts made a year or two earlier cannot honestly be traced to them. He is also, separately, reported to be working on more than $750B in fresh AI infrastructure deals, reviving the question of whether vendor financing is inflating apparent demand. Note the interest. Huang sells the compute, and a world where AI expands ambition is a bigger market than one where it trims payrolls. His timeline objection still holds on its own terms, and it converges with a Federal Reserve survey that has no stake in the answer. Treat him as corroboration, not as authority.
Disruption
60/100
GTC 2026 / Cramer Interview
Last updated: Issue 18 — September 2026
MS
Mustafa Suleyman
Microsoft AI CEO
Alarmed
"Most white-collar tasks will be fully automated within 12–18 months"
The most aggressive timeline from any major tech leader. Named accounting, legal, marketing, and project management as vulnerable. Envisions "professional-grade AGI" and "billions of digital minds." Yet also says AI should operate "in a subordinate way to us" — a tension between disruption forecasts and safety rhetoric.
Disruption
90/100
Fortune / FT Interview (Feb 2026)
Last updated: Issue 10 — April 2026
SA
Sam Altman
OpenAI CEO
Pragmatic
"There's some AI washing where people are blaming AI for layoffs they would otherwise do"
At BlackRock's U.S. Infrastructure Summit and the India AI Impact Summit (early 2026), the OpenAI CEO acknowledged what this publication has called expectation-driven cuts: "I don't know what the exact percentage is, but there's some AI washing… and then there's some real displacement by AI of different kinds of jobs." When the most prominent AI builder publicly admits the rationale is partially fake, the rhetorical cover for capital-reallocation-as-AI-strategy is gone. Joins Cognizant CAO Babak Hodjat in calling out the gap between AI rhetoric and AI reality in layoff announcements.
Disruption
65/100
Tom's Hardware / India AI Summit
Last updated: Issue 11 — May 2026
CR
Citrini Research
Top Finance Substack + Alap Shah (LOTUS)
Alarmed
"Ghost GDP" — output that never reaches household wallets
Scenario that spooked markets and divided Wall Street. Co-author Shah called for AI tax. Moved software and fintech stocks.
Disruption
95/100
The 2028 Global Intelligence Crisis
Last updated: Issue 8 — March 2026
FF
Frank Flight / Citadel
Citadel Securities Macro Strategy
Optimistic
"Rising productivity expands the consumption frontier"
Systematic rebuttal. Keynes was right on productivity, wrong on labor, because human wants are elastic.
Disruption
30/100
Fortune: Citadel Rebuttal
Last updated: Issue 8 — March 2026
HP
Helen Poitevin
Distinguished VP Analyst, Gartner
Cautious
"Reductions create budget room — they do not create return"
Poitevin's Gartner analysis is the balance-sheet half of the Issue 13 thesis. Across deployers, the firms cutting deepest showed no return premium. Cutting frees budget. It does not on its own create value, because value comes from recomposing the work. Read alongside Stanford's worker-desire data, the two point the same way: return comes from augmentation toward partnership rather than from subtraction.
Disruption
50/100
Gartner — Future of Work
Last updated: Issue 14 — July 2026
TC
Tracey Countryman
Global Lead, Supply Chain & Engineering, Accenture
Nuanced
Turn the talent shortage into strength — redesign the work, don't just fill it
Countryman frames the 1.1-million-role supply-chain gap as a design problem, not a hiring problem. With demand outrunning the labor force, the organizations that win won't be the ones that recruit hardest — they'll be the ones that redesign roles around human–machine teaming, sequence reskilling by urgency, and pair the CSCO's work definition with the CHRO's skills plan. The throughline of her team's model is the same as this publication's: technology adoption is inevitable; workforce redesign is the choice.
Disruption
42/100
Accenture — Workforce of the Future
Last updated: Issue 14 — July 2026
AB
Asher Bantock
Head of Research, SignalFire
Nuanced
Engineering didn't crater — it's the most resilient function
Bantock's team tracks hiring rather than the noisier layoff signal. Their read on the twelve big-tech "Majors": total hiring is down roughly 25% versus 2019, but engineering only ~11% — and engineers now make up 55% of new hires, up from 46% in 2019. If AI were truly substituting for engineering talent, that hiring would be the first to fall; it isn't. The counter-narrative tracks the redesign thesis: roles are being repositioned and made more demanding, not cleanly deleted.
Disruption
38/100
SignalFire — State of Talent
Last updated: Issue 14 — July 2026
NL
Nickle LaMoreaux
Chief Human Resources Officer, IBM
Pragmatic
Rebuilt entry-level work around what AI cannot do, then tripled the intake
The clearest counter-model to cutting the base. After AI handled 94% of routine HR requests and failed on the 6% that involved exceptions and ethical calls, IBM tripled US entry-level hiring and pointed new people at the 6%. Read through the shape argument in this issue, that is the base being redeployed rather than removed: the same entry band, doing orchestration and exception work instead of routine processing. LaMoreaux's question is the one every CHRO should be able to answer: if you stop investing in entry-level hires, what happens in three to five years?
Disruption
48/100
TechCrunch / Bloomberg
Last updated: Issue 16 — August 2026
AC
Andy Challenger
Chief Revenue Officer, Challenger, Gray & Christmas
Nuanced
AI was the top stated reason for five months, then fell to fourth in one
Challenger's monthly report is the most-cited source on why US employers say they are cutting, which makes its August print the most important single data release of this cycle. AI fell to the fourth-most-cited reason at 3,462 cuts, its lowest month since December 2025, ending a run that started in March. Restructuring took the lead at 16,173. AI still leads the year at 116,175, about 22% of all cuts. Challenger's own read of the month is that employers are now planning to add workers, with 46% of those plans coming from manufacturing. He has been careful throughout not to treat the stated reason as a verified cause, which is the right posture and one this issue adopts.
Disruption
62/100
Challenger H1 2026 report
Last updated: Issue 16 — August 2026
LA
Leopold Aschenbrenner
Founder, Situational Awareness LP
Alarmed
The capacity thesis, expressed at maximum conviction — and maximum leverage
Built a fund on the proposition that increasingly capable AI would drive enormous demand for semiconductors, memory, data centers and power, paired with shorts against software firms judged vulnerable. Up 439% net through June 30, 2026 and near $45B in assets; forced to sell the entire public book to Citadel weeks later after margin calls, ending near $10B. Included here not as a market call but because his thesis is the one many workforce plans quietly assume.
Disruption
95/100
CNBC
Last updated: Issue 16 — August 2026
DG
Dan Gilbert NEW
Founder & Global CEO, Brainlabs
Optimistic
"We didn't do that by cutting junior roles to save money on tools"
Grew entry-level intake from 19 in October 2023 to 64 in April 2026 — a 237% increase — at a 1,000-person media agency. His account is the cleanest description of a teaching institution in the market: an academy running since the company started, always training entry-level hires on whatever skill mattered most at the time, now used to teach people to work alongside AI from day one rather than bolting it on after they learned the old way. Note the sequence — the institution came first, the AI curriculum second.
Disruption
30/100
Forbes
Last updated: Issue 17 — August 2026
BO
Brian Ong NEW
VP of Recruiting, Google
Pragmatic
Entry-level hiring holds where the work is redefined around judgment
Speaks to what large-scale technology recruiting now selects for: candidates who can reason with and validate AI output rather than produce work the tools already produce. Sits alongside Google’s wider position that a degree or prior industry experience should not be the only route in — career certificates requiring neither, apprenticeships for graduates and career changers, and reported pilots of technical assessments in which candidates may use approved AI assistants. Testing people on work they must do without the tool they will always have measures the wrong thing.
Disruption
45/100
Forbes
Last updated: Issue 17 — August 2026
YB
The Budget Lab at Yale NEW
Economy-wide labor market analysis
Optimistic
No economy-wide disruption, and a warning about what gets attributed to AI
The strongest available check on the entire displacement thesis, and Issue 18 agrees with more of it than any previous issue. The Budget Lab finds the US occupational mix shifting no faster than it did for the PC or the internet, and no relationship between an occupation's AI exposure and its employment or unemployment duration. Executive director Martha Gimbel has become the most prominent critic of what is now called AI washing, arguing the macroeconomic effect is not visible however you cut the data. Where we part company is on the inference rather than the finding: a steady national mix can sit on top of heavy movement inside firms, because the national number averages companies moving in opposite directions.
Disruption
22/100
Budget Lab analysis
Last updated: Issue 17 — August 2026
Ethics, Policy & Future Thinkers
SW
Stephen Wroblewski
Accenture — Managing Director, Talent & Workforce Reinvention; author of this publication
Nuanced
Capacity freed is not headcount reduced — it is an enterprise asset, and what you do with it is a choice
Works at the task level rather than the job level, on the argument that two people with the same title can sit on opposite sides of the same technology. Across more than a hundred client engagements the recurring finding is that AI frees capacity long before it removes work, and that what happens next is a leadership decision rather than a technical outcome. His position is that freed capacity belongs to the enterprise rather than to the function that freed it, and should be rotated, redeployed or released against that company's own value equation — its particular mix of growth, quality, pace, risk, cost and customer intimacy — instead of against a benchmark. Issue 18 extends this into organizational shape: the pyramid recomposes rather than shrinks, the base is redeployed upward into orchestration, and the wide bands of judgment and proximity that form above it are the part of a workforce a competitor cannot buy. Writes and builds in the open, including the diagnostic tooling behind the analysis, and publishes the counter-evidence alongside the argument.
The Great Recomposition · LinkedIn
Disruption
58/100
Added: Issue 18 — September 2026
AM
Andrew McAfee
MIT — Principal Research Scientist
Cautious
The "apprenticeship ladder" is what's actually being lost
Names the mechanism behind the entry-level data: the pathway through which new hires gradually accumulate the judgment senior work requires. Firms that cut junior cohorts without rebuilding it save money over two years and deplete the partner pipeline they need over a decade — a cost that never appears in the quarter that causes it.
Disruption
70/100
Entry-Level Pipeline Analysis
Last updated: Issue 15 — July 2026
EM
Ethan Mollick
Wharton, Co-Intelligence
Nuanced
"No one knows anything" about AI jobs impact
Entry-level jobs a "huge concern." Advocates using AI to learn, not just produce. Perfectly captures the Issue 8 zeitgeist.
Disruption
50/100
One Useful Thing
Last updated: Issue 8 — March 2026
JS
Jeffrey Sonnenfeld & Steven Tian
Yale Chief Executive Leadership Institute
Alarmed
"AI won't kill your job — it will kill the path to your first one"
Yale CELI's April 29 piece named the long-term cost more sharply than anyone yet: agentic AI is not creating a layoff event you can see; it is steadily narrowing the entry-level openings that train future seniors. The biggest impact will be invisible in earnings calls — and devastating for the workforce pipeline five years out. Quote of the quarter for the Issue 11 ladder-collapse thesis.
Disruption
80/100
Fortune / Yale CELI
Last updated: Issue 11 — May 2026
YL
Yann LeCun
Turing Award, AMI Labs
Optimistic
Current AI can't do what others claim
Hardest skeptic among frontier researchers. Challenges whether LLMs represent real intelligence.
Disruption
35/100
Public Statements
Last updated: Issue 6 — February 2026
FL
Fei-Fei Li
Stanford HAI, World Labs CEO
Pragmatic
Human-centered AI is the path to shared prosperity
Advocates for AI that augments human capabilities. Emphasizes the importance of diversity in AI development.
Disruption
55/100
Stanford HAI
Last updated: Issue 6 — February 2026
AK
Andrej Karpathy
Former Tesla AI / OpenAI; Independent Researcher
Cautious
Job-level AI scoring is the wrong frame — occupation-level analysis misses the task-level reality
Published BLS occupation scoring research at karpathy.ai/jobs, systematically rating 700+ occupations for AI exposure. Key insight: scoring at the job level produces misleading conclusions because every job is a bundle of tasks with different exposure profiles. Reinforces the task-level thesis at the core of this publication. One of the few frontier AI researchers doing granular, occupation-by-occupation empirical work rather than top-down forecasting.
BLS Occupation Scoring Research
Disruption
72/100
Last updated: Issue 9 — March 2026
← PreviousGlobal & Regulatory
11 of 13
All sections
Next →Implications
What To Do

Strategic Implications

Role by role. Issue 18 adds one question that sits above everything else on this list. Can you draw your organization's shape, not its org chart but its shape, and point at the bands where judgment concentrates? If the answer is a functional hierarchy, you have described reporting lines. The shape is the part a competitor cannot buy, and no disclosure field will ever ask you about it.

1. Draw the Shape Before You File the Reason ISSUE 18 LEAD

From 1 October, a Connecticut mass-layoff filing has to say whether AI was involved, against no legal definition of the term. More states and a federal draft follow. Do not treat this as a compliance task. The work that protects you is the same work that creates value: name your bands and where people sit in them. Four answers, this quarter. Where are your wings, meaning the layers where closeness and judgment create value a competitor with identical technology could not reproduce, named as bands of people rather than as functions. Is your entry level being redeployed or removed? Redeployment means the same people doing orchestration instead of execution, with job architecture and pay bands updated to match. Removal means the band thins and nothing takes its place. On a headcount report those look identical. On a capability plan they are opposites. Who is handling the hard cases the machines send back, and is their time funded as real work or running on goodwill? And who decides where freed capacity goes? If that call sits with the function that freed it, you get release by default, because release is the only pathway visible in their budget. Freed capacity is an enterprise asset and the pathway decision belongs above the function. The Fed CFO evidence is clear about which direction pays: growth and product motives correlate with real productivity gains, cost-reduction motives do not.

2. Rebuild the On-Ramp Before You Need It ISSUE 15 LEAD

The Canaries data shows the 22–25 cohort in the most AI-exposed occupations running ~13% below peers and the gap widening ~0.5pp a month; entry roles across law, consulting and banking are down ~35% since 2023. Treat this as a production-capacity problem, not a cost line. Concretely: sort entry-level tasks on two axes — automation exposure and learning value — and defend the high-learning quadrant by redesigning it, not preserving it. Apply the four tests: AI literacy inside onboarding (not tool access); a written account of how juniors supervise and validate machine output; promotion on demonstrated judgment rather than tenure; and senior expectations genuinely redesigned rather than handed to a smaller team. Then measure entry-level openings as a ratio of total openings, quarter over quarter, with the seriousness you give senior attrition. The causal debate is unresolved — the move isn't.

3. Redesign the Role — the Lever That Scales the Roles You Have ISSUE 14 LEAD

Accenture's task-level model settles the strategy question: a 1.1-million-role gap hiring can't close, and a pharma case where full automation netted ~6% against a 33% target. Neither reflex works on its own — so the move is to redesign each high-impact role around what it becomes once systems absorb the routine: oversight, exception-handling, judgment. Run the three plays together — foresight (model demand at task level), redesign (the CSCO defines the work, the CHRO equips it), and continuous skills tied to real mobility. Deploy to where worker desire and capability overlap — the same agency map from Issue 13 — so the redesigned role is also the one people actually want to keep.

4. Start with Task & Skill Architecture REINFORCED

MIT CSAIL tested 11,500 tasks: AI succeeds at 47% (legal) to 73% (maintenance admin). HBS/Srinivasan: automatable tasks ↓17%, augmentation tasks ↑22%. Now Stanford adds the worker layer at the same unit — 844 tasks, not 104 job titles. Decompose roles into tasks and map capability and desire against each. Organizations still planning at the job level — headcount up or down — are deciding on the wrong unit of analysis. Get to the task to get to the strategy.

5. Design Toward H3 — the Augmented-Human Sweet Spot NEW

Equal Partnership (H3) is the most-desired agency level in 47 of 104 occupations, and 69.4% of workers want AI specifically to free time for higher-value work. Translate that into operating-model terms: most work should land in Augmented Human, not full automation or untouched manual. Build the human-AI interaction deliberately — handoffs, review points, escalation — rather than defaulting to "AI does it" or "leave it alone." This is where worker desire and Gartner's returns evidence overlap.

6. Stop the Pipeline Choke — and Track the Invisible Cut DEFINING

Oracle cut 30,000 to fund infrastructure; IBM tripled entry-level hiring and redesigned every junior role. Yale's Sonnenfeld: "AI won't kill your job — it will kill the path to your first one." Track entry-level openings the way you track senior departures. If junior posting volume falls while compute capex rises, your future leadership bench is being silently mortgaged — invisible in any earnings call, devastating five years out.

7. Distinguish the Three Cuts — Honestly, Before the Law Does UPDATED

Three different things travel under "AI restructuring": capability replacement (AI genuinely doing the work), capital reallocation (salaries cut to fund capex), and reputation-managed restructuring (buyouts replacing layoffs for the same outcome). The new tell: NBER finds 90% of execs say AI had zero impact at their own firm — even while headlining it externally. That gap is now meeting a paper trail: Oracle named AI in a June 22 SEC filing, the federal Great American AI Act would require WARN notices to disclose AI as a "substantial factor," and California's AI Transparency Act becomes operative August 2 (Colorado's original AI Act was repealed and replaced by SB 26-189, effective January 1, 2027 — see the correction in What's New). You will soon have to name which system actually triggered a cut. Audit your portfolio now.

8. Close the Deployment & Adoption Gap REINFORCED

MIT's 5× capability gap, the SF Fed's productivity paradox, and Gallup (half of U.S. workers use AI yearly or never) all say the same thing: deployment, not model quality, is the bottleneck. You cannot claim AI-driven savings when half your workforce hasn't adopted the tool — and Gartner shows the deepest cutters got no return premium. Measure actual adoption before cutting headcount; the firms that close this gap first capture the value the cutters never see.

9. Plan for Agents, Not Just Copilots

The shift from AI-as-tool to AI-as-teammate is accelerating. Microsoft: 82% plan agents within 18 months. Gartner: 32M jobs/year reshaped by 2028. Your workforce model needs a human-agent ratio and a governance layer, not just a headcount — and WORKBank's skill-shift data says the human side of that ratio is increasingly about coordination, coaching, and judgment exercised alongside agents.

10. Invest in Adaptive Capacity & Reskilling REINFORCED

The CEPR finding — training investment delivers a 5.9× AI-productivity multiplier — makes the business case plainly: investing in people is the highest-return AI strategy. Brookings shows adaptive capacity varies enormously; workers with transferable skills and networks absorb disruption, those without don't. As AI absorbs routine information tasks, deliberately build the coordination and judgment skills that rise in value — the Keep/Build/Retire move, applied at task level.

11. Calibrate to Geography, Demography & Sector — Act Enterprise-Wide

Impact lands unevenly. Tufts: Information at 18% displacement risk, Finance 16%, physical labor under 1%. It concentrates in white-collar metros and disproportionately affects women (PwC) and young workers (Goldman). Calibrate the agency-gap analysis to your sector and workforce — but build reskilling, governance, and disclosure-readiness enterprise-wide. And design strategies that hold under both the gradual-evidence and rapid-expectation scenarios.

← PreviousVoices
12 of 13
All sections
Next →Archive
Archive

The Series So Far

Eighteen issues, twelve months, 52+ institutions. The arc is cumulative: each issue positions itself against the one before it. The full hub is at the index.

Issue 17
The Teaching Problem
Issue 16
The Leveraged Bet
Issue 15
The Missing Rung
Issue 14
The Redesign Mandate
Issue 13
The Spectrum
Issue 12
The ROI Reckoning
Issue 11
The Mechanism
Issue 10
Scale Without Strategy
Issue 09
Expectation vs Evidence
Issue 08
Markets vs Research
Issue 07
The Synthesis
Issue 06
Foundations

Every issue in one place: open the hub →

← PreviousImplications
13 of 13
All sections
↑ Back toCover

The Great Recomposition — Issue 18 — A Multi-Institutional Analysis of AI's Impact on the Future of Talent and the Workforce

Including perspective from Stephen Wroblewski and his ongoing work across industries and functions as a practitioner and thought leader in Talent & Workforce Consulting at Accenture.

This publication synthesizes research from 52+ institutions including MIT, MIT CSAIL, Stanford HAI, Stanford SALT Lab, Stanford Digital Economy Lab, Oxford, Harvard Business School, Yale CELI, The Budget Lab at Yale, Wharton, Tufts Digital Planet, Harvard/Brookings, McKinsey, Accenture, BCG, Deloitte, PwC, WEF, IMF, OECD, Anthropic, OpenAI, Google DeepMind, Goldman Sachs, Microsoft/LinkedIn, IBM, ServiceNow, Cognizant, Gartner, Federal Reserve Bank of Atlanta, Federal Reserve Bank of Richmond, Federal Reserve System, Duke/NBER, CEPR, PIIE, ADP Research, Mercer, Citrini Research, Citadel Securities, Burning Glass Institute, Gallup, Bloomberg, SignalFire, Challenger Gray & Christmas, Indeed Hiring Lab, Randstad, NACE, Handshake, Ramp/Revelio Labs, Robert Half, Strada Institute for the Future of Work, Orgvue, Forrester, the US Department of Labor, the Connecticut Department of Labor, Epstein Becker Green and Newmark Research.

Research synthesis spanning February 2024 through September 11, 2026. Market and capital-markets reporting is included for structural context only; nothing in this publication is investment advice or a market view.

Last updated: Friday, September 11, 2026

Contact: Stephen Wroblewski | stephen.m.wroblewski@accenture.com

Back to Top
60-Second Read What's New Stephen's Take The Reason Code The New Pyramid The Debate Signal Map Industry Pulse Global & Regulatory Voices Implications Archive
Sections
0
Skip to Content
The Great Recomposition
Issue 17
Issue 16
Issue 15
Issue 14
Issue 13
Issue 12
Issue 11
Home
Issue 10
Issue 9
The Great Recomposition
Issue 17
Issue 16
Issue 15
Issue 14
Issue 13
Issue 12
Issue 11
Home
Issue 10
Issue 9
Issue 17
Issue 16
Issue 15
Issue 14
Issue 13
Issue 12
Issue 11
Home
Issue 10
Issue 9