The Role Decomposition Doctrine: How Anthropic's Just-Released Numbers Reveal Why 88,000 Workers Lost Their Jobs This Year, And The 5-Question Audit Every Founder Should Run On Every Role Before The Next Hire Or Fire

TL;DR

• Anthropic's June 26, 2026 Economic Index survey of 9,700 verified Claude users found that more than 35% expect AI to do most or nearly all of their work tasks within 12 months (Anthropic).
• US companies have already cited AI in 87,714 job cuts year-to-date through May, more than the 54,836 attributed to AI in all of 2025 (TheStreet).
• AI was the named driver in 40% of the 97,000 May 2026 cuts, the highest May since 2020 (Yahoo Finance).
• Oracle alone shed 21,000 roles tied to an AI-driven reshape, Elastic cut about 7% of its workforce, and a Brynjolfsson-led payroll study found a 13% relative decline in employment for 22-25 year olds in AI-exposed jobs while older workers held steady (Forbes Jon Markman).
• On June 25, 2026, RAISE US launched with more than $500 million already committed and Amazon, Anthropic, Microsoft, and the OpenAI Foundation as anchor partners (Rockefeller Foundation).
• Most founders are still hiring and firing off an org chart instead of the actual task atoms inside each role.
• Run The Role Decomposition Doctrine 5-question audit before the next hire, fire, or promotion.

The hook

Look at this number.

35%.

That is the share of working professionals who told Anthropic, with their actual usage data attached, that AI will be able to do most or nearly all of their job within 12 months (Anthropic).

Not assist.

Not help with email.

Most.

Of their job.

Now look at the other number.

87,714.

That is how many US workers have been cut from payrolls this year with AI named as the reason, surpassing all of 2025 before the calendar even hit July (TheStreet).

If you are running a company right now, those two numbers live in the same building.

Inside that building is your team.

What just happened in the labor market?

On June 26, 2026, Anthropic published the sixth edition of its Economic Index (Anthropic).

This one was different.

For the first time the report paired survey responses from approximately 9,700 verified Claude users with their actual usage data, so you can stop arguing about hypothetical AI capability and start reading what real users say about their own jobs (Build Fast with AI).

The headline finding: more than 35% expect AI to do most or nearly all of their work tasks within 12 months (Rediff).

About 6 in 10 respondents put themselves in a higher AI task-share band next year than they sit in today (New Indian Express).

And in roughly 49% of US occupations, at least one in four tasks is already being run through Claude, a number that has barely moved in the last three reports, suggesting the breadth of AI task coverage has stabilized (Build Fast with AI).

Then the same week the payroll data arrived.

Challenger, Gray & Christmas reported 97,000 announced US job cuts in May 2026, the highest May since the pandemic in 2020, with AI named as the primary reason for 40% of them (Yahoo Finance).

AI-attributed cuts hit 38,579 in May, the largest single month on record since Challenger started tracking the reason in 2023 (TheStreet).

Oracle alone announced about 21,000 cuts as part of an AI-driven business reshape (Forbes).

Elastic cut about 7%, just under 300 roles, while telling staff it still plans to grow total headcount in customer-facing sales (Metaintro).

The pattern is not "shrink the whole company."

The pattern is "decompose roles and re-stack the org around the work AI can already do."

Then on June 25, 2026, the rest of the response showed up.

RAISE US, co-founded by former Commerce Secretary Gina Raimondo and former Indiana Governor Eric Holcomb, launched as a national nonprofit dedicated to AI workforce transition, with more than $500 million already secured against a $1 billion target (Rockefeller Foundation).

The anchor partners include Amazon, Anthropic, Microsoft, the OpenAI Foundation, Bank of America, IBM, Cisco, Autodesk, General Motors, Eli Lilly, and the Stephen A. Schwarzman Foundation (Politico).

State partnerships are live in Arkansas, Connecticut, Maryland, and Utah (Business Insider).

Boston Consulting Group estimated that roughly half of all US jobs will be reshaped by AI over the next few years, with up to 25 million eliminated in the next five (Fortune).

That is the macro. Here is the part that matters to you as a founder.

Why are the cuts happening at companies that have not seen the productivity payoff yet?

A Brynjolfsson-led study using payroll data from millions of employees found a 13% relative decline in employment among 22-to-25-year-olds in jobs most exposed to AI, while employment for older workers in the same roles held steady or even rose (Forbes Jon Markman).

PwC's 2026 outlook found a 35% rise in entry-level roles requiring senior-level skills since 2019, while traditional entry-level postings dropped 10% (Forbes Jon Markman).

Translation.

The work formerly called "entry-level" is being done by AI.

The work formerly called "senior" is being done by humans plus AI.

The hire you used to make at 23, with a 12-month ramp, is collapsing into one seat that demands judgment, taste, and AI fluency on day one.

One LinkedIn analysis flagged that roughly 80% of companies that cut jobs for AI in 2026 have not yet shown the productivity payoff (LinkedIn analysis).

They cut first and restructured later, and now they are scrambling to figure out which task atoms moved and which they accidentally orphaned.

You do not have to make the same mistake.

What is The Role Decomposition Doctrine?

This is the framework.

The Role Decomposition Doctrine: before you hire, fire, promote, or restructure any seat on your team, decompose the role into its task atoms, tag each atom Human-Only, AI-Assistive, or AI-Native, then re-stack the role around the work that humans uniquely create.

Five questions.

You can run them in one afternoon.

Question 1: Task Atom Map. What are the 15 to 25 discrete things this role actually does this month?

Not job description language.

Real tasks.

"Drafts the Monday sales recap email." "Reconciles Stripe and QuickBooks." "Builds the weekly ad creative brief." "Sits in the 11am client call and types the action items in Notion."

Get specific.

If you cannot name 15 task atoms inside the role, you do not know the role well enough to make a hire or fire call.

Question 2: Decomposition Tag. Which atoms are Human-Only, AI-Assistive, or AI-Native?

Human-Only: judgment under high stakes, client relationships, creative origination, leadership, anything that fails without a human in the loop.

AI-Assistive: AI drafts, a human reviews and ships. This is most knowledge work in 2026.

AI-Native: AI runs the task on a recurring trigger with zero human touch except an exception alert. This is the category that exploded in the last 12 months.

Most founders are shocked to find that 30% to 60% of the tasks in a typical white-collar role tag AI-Assistive or AI-Native (Anthropic).

Question 3: Hours Reclaimed. How many weekly hours come back if you push every AI-Assistive and AI-Native task into a system?

Multiply the task atom count by the hours-per-week estimate.

You will see a number that looks like 8, 12, or 18 hours per role per week.

Run that across five seats and you have a half-time hire's worth of capacity sitting inside your existing payroll.

That is the lever Oracle pulled when it took 21,000 seats off the org chart this year (Forbes).

You can pull a smaller, smarter version of the same lever without firing anyone.

Question 4: Re-Stacking. Can the role be re-stacked at higher impact instead of cut?

This is where most founders rush.

Cutting the seat is the easy answer.

Re-stacking the seat is the high-impact answer.

Take the reclaimed hours and push the human into work that compounds: relationships, taste, strategy, more output volume, more accounts, more client time.

Elastic is the public version of this. The company cut about 7% in some functions while telling staff it still plans to grow customer-facing sales headcount this year (Metaintro).

Question 5: Hire or Fire Trigger. Is the decision you are about to make based on org chart inertia or on the decomposed task map?

If you cannot point to specific task atoms and explain which were absorbed by AI, which were re-stacked into the role, and which justify a brand new seat, you are making the decision on inertia.

Run the audit first.

Then make the call.

How do you actually do this in a small business this week?

Three moves.

Move one: pick the role on your team that you have been quietly worried about and run a 30-minute task atom map. Write out the 15 to 25 actual tasks that seat did this month.

Move two: tag every task Human-Only, AI-Assistive, or AI-Native. Be honest. If you would not bet your own money on the AI handling the task to your standard with a review pass, it is not AI-Native yet.

Move three: sit down with the person in the seat and walk through it together. Ask them which AI-Assistive tasks they would hand to a workflow tomorrow if you funded it. You will be stunned how often the human is already two steps ahead of you.

Decompose first. Then decide.

What does this mean for your next hire?

The 22-year-old hire with AI fluency on day one looks completely different in 2026 than the 22-year-old you hired in 2022.

PwC data shows entry-level postings demanding senior-level skills are up 35% since 2019 (Forbes Jon Markman).

That is what your next job description should look like. Senior-level taste and judgment. AI-native execution. One seat doing what used to take two or three.

If you cannot find that hire on the open market, build one. Take a high-potential person you already have and give them the AI toolkit, the workflow training, and the runway to grow into that seat.

Anthropic's data shows workers with at least 15 years of experience estimate AI can handle about 10 percentage points fewer of their tasks than first-year workers (New Indian Express).

Senior workers are not delusional. They are doing tasks that survive automation better. That is the kind of work you want the surviving seats focused on.

What does the doctrine say about workforce retraining?

RAISE US launched with $500 million committed and Amazon, Anthropic, Microsoft, and the OpenAI Foundation anchoring it, because the companies that built the wave know they cannot leave the workforce holding the bag (Politico).

You are not RAISE US. You do not need a $1 billion fund.

You need a single line in your operating plan: every team member gets two paid hours a week to learn AI workflows inside their role.

That is the smallest viable version of internal retraining, and Gallup data shows tech workers who skip AI face roughly triple the layoff risk of peers who use it (updater.ge).

The data is too consistent to ignore.

FAQ

Will AI replace my whole team?

No, but it will reshape which seats earn their cost. The Anthropic data shows about 35% of workers expect AI to do most of their tasks within a year, and roughly half of US occupations already run at least 25% of tasks through Claude (Anthropic). That is not a wipeout, it is a re-stack. Run the doctrine before you make headcount changes.

How do I tell if a task is AI-Native, AI-Assistive, or Human-Only?

Ask one question: if this task runs without a human in the loop today, does the business lose money or trust? If yes, it is at most AI-Assistive. If no and the AI output is already meeting your standard, it is AI-Native. Everything that requires judgment, relationships, or original creative is Human-Only (Forbes Jon Markman).

Should I fire someone because AI can do their job?

Probably not as your first move. Companies that cut first and re-stacked later are showing the productivity payoff problem cited in the LinkedIn analysis of 2026 cuts (LinkedIn analysis). Re-stack the role first. If the person cannot grow into the higher-impact version of the seat, then make the call.

How much of a normal knowledge-work role is AI-Native right now?

For most roles, 10% to 25% of task atoms tag AI-Native and another 30% to 50% tag AI-Assistive with proper workflow setup (Anthropic). The exact mix depends on the role, the industry, and how mature your AI tooling is. Run the map and find out.

Where does the $500 million RAISE US effort actually go?

RAISE US is funding employer-government partnerships in Arkansas, Connecticut, Maryland, and Utah, with founding partners that include Amazon, Anthropic, Microsoft, OpenAI Foundation, Bank of America, IBM, Cisco, Autodesk, General Motors, Eli Lilly, and the Stephen A. Schwarzman Foundation (Rockefeller Foundation, Business Insider). The money targets retraining for workers transitioning out of AI-exposed roles.

The bottom line

35% of working professionals just told Anthropic, with usage data attached, that AI will do most of their work within a year (Anthropic).

87,714 US workers have already been cut from payrolls this year with AI named as the reason (TheStreet).

Your team sits somewhere in the middle of those two numbers.

The founders who win the next 12 months are the ones who decompose the work before they restructure the team.

The Role Decomposition Doctrine is the audit.

Five questions. One afternoon. Every seat.

Run it.

Then run it again every quarter.

If you want help running it on your specific team and roles, book an AI Implementation Session and we will walk it together.

If you would rather start with a system that already replaces AI-Native task atoms inside your business, the 8 Figure AI Toolkit includes First AI Hire, the AI Executive Assistant that handles inbox, scheduling, follow-up, and CRM updates so you can re-stack the human seats toward work that compounds.

Either way, do not let June 2026 pass without running the audit.

The team you take into July is the team you take into the second half of 2026.

Make it count.