video
play-sharp-fill

A widely shared Medium essay by PomeroySays, “AI is coming for these jobs. Is yours on the list?”, laid out the version of this story most people are carrying around in their heads: five kinds of jobs AI is coming for, and ten jobs to run toward instead. It is a clean, readable list. It is also built on numbers that have already been replaced.

So we did the unglamorous thing. We took the essay’s lists to the primary sources — the U.S. Bureau of Labor Statistics’ newest Employment Projections (2024–34, released August 2025), the World Economic Forum’s Future of Jobs Report 2025, Microsoft Research’s 2025 study of 200,000 real AI conversations, Stanford’s 2025 payroll-data study of young workers, and the Anthropic Economic Index — and we checked every claim, job by job.

FACT: Most of the essay’s instincts were right. The jobs it says are shrinking really are shrinking, and all ten of the jobs it tells you to run toward are still on the government’s fastest-growing list. But two of its “at-risk” picks are actually projected to grow, its growth numbers are a full projection cycle out of date, and it missed some of the biggest declines in the entire federal dataset. Here is the corrected version — with the receipts shown, including what we removed and why.

The Jobs Actually Shrinking

Data entry and office administration. This was the essay’s first category, and the data backs it harder than the essay did.

FACT: The fastest-declining occupation in America is word processors and typists, projected by the BLS to shrink 36.1% by 2034. Data entry keyers are right behind at −25.9%, a loss of about 36,700 jobs. Secretaries and administrative assistants decline more slowly (−1.6%) but from an enormous base, shedding roughly 30,800 positions. The World Economic Forum’s employer survey independently puts data entry clerks among the fastest-declining roles on Earth. Verdict: this category is not hype. It is the steepest cliff in the federal data.

Customer service and telemarketing. Also real — with one correction to the popular picture.

FACT: The BLS projects customer service representatives to decline 5.5% through 2034. That sounds mild until you see the base: about 2.8 million Americans do this work, so a 5.5% slide is roughly 153,700 jobs. Telemarketers fall much faster, −22.1%. Microsoft’s study of actual Copilot usage ranks customer service among the occupations whose daily tasks overlap most with what AI already does well, and Stanford’s researchers found customer service is one of the fields where hiring of 22-to-25-year-olds has fallen hardest since 2022.

COMMENTARY: But this is a slope, not a cliff — and the industry’s flagship experiment shows why. In February 2024, the fintech company Klarna announced its AI assistant was doing the work of 700 customer service agents. By May 2025, its CEO admitted to Bloomberg that cost had driven the rollout too far, quality had dropped, and the company was hiring humans again, settling into a hybrid model. The honest reading: routine support work is genuinely being automated, and the companies that tried to automate all of it at once hit a quality wall. Both halves of that sentence matter.

Customer service and data entry desks giving way to automation

Finance and accounting — the routine end. The essay named junior bookkeepers, and on that one it was right.

FACT: Bookkeeping, accounting, and auditing clerks are projected to decline 5.8%, about 94,300 jobs, as software absorbs the routine ledger work. (Two other finance picks from the essay did not survive the fact-check — they are covered in the removals section below.)

Content and creative work. Here the essay’s framing fails the data, and the truth is more interesting.

FACT: The BLS does not project these occupations to shrink at all. Writers and authors are projected to grow 3.6%, graphic designers 2.1%, editors 0.6%, technical writers 0.9%. What is true is narrower and sharper: Microsoft ranks writers and authors fifth in the nation for AI task overlap, and Stanford’s data shows the pain landing on entry-level hiring. The work is being transformed and the first rung of the ladder is being squeezed — but “these jobs are disappearing” is not what the numbers say, and we are not going to print it just because it is the scarier headline.

Junior coding. Same pattern, bigger scale.

FACT: Software development is one of the great growth occupations of the decade — the BLS projects +15.8%, roughly 267,700 additional developers by 2034. And yet Stanford’s payroll study found that employment for software developers aged 22 to 25 fell nearly 20% from its late-2022 peak, part of a roughly 13% relative decline for young workers across AI-exposed occupations. Experienced developers in the same study were stable or growing.

COMMENTARY: Put those two facts together and you get the single most important sentence in this whole story: so far, AI is not mainly deleting occupations — it is narrowing the door into them. The risk is real, but it is aimed at the first job, not the profession.

What We Took Off the List

This is the part most viral lists never show you. Three of the essay’s picks did not survive contact with the current data:

Basic tax preparers — removed. The essay listed them as at-risk. The BLS projects tax preparers to grow 4.5% through 2034. Whatever AI does to tax software, the federal forecast for the human occupation points up, not down. Off the list.

Entry-level financial analysts — removed as a declining occupation. Financial and investment analysts are projected to grow 5.7% (+21,100 jobs). Entry-level analysts absolutely feel AI pressure on their daily tasks — but a pressured occupation and a shrinking occupation are different claims, and only one of them is supported.

Every growth number — updated. The essay’s safe-list figures came from the BLS 2023–33 projections. The government has since published the 2024–34 cycle, and every number moved. The section below uses the new ones.

Two more we checked and deliberately left off the risk list, because checking cuts both ways: paralegals, a favorite of “AI eats the professions” headlines, are projected essentially flat (+0.2%) with about 39,300 openings a year; and proofreaders, already reduced to a 12,000-person occupation by earlier technology waves, are projected at −0.6% — a rounding error, not a collapse.

What We Added to the Risk List

The essay missed several of the largest declines in the federal data. These belong on any honest version of the list:

FACT: Cashiers are projected to decline 9.9% — about 313,600 jobs, the largest numeric loss of any occupation we checked. The WEF’s employers likewise put cashiers and ticket clerks in the largest absolute decline group worldwide.

FACT: Bank tellers fall 12.9% (−44,900 jobs), and the WEF ranks bank tellers among the fastest-declining roles globally. The ATM famously did not kill the teller; mobile banking plus automation is doing what the ATM didn’t.

FACT: Postal service clerks make the WEF’s fastest-declining list — though here the two sources argue, and we will show you the argument rather than pick a winner quietly: the BLS projects only −3.5% for the U.S. occupation. Global employer expectations and the U.S. federal forecast are telling different stories about the same job.

FACT: Insurance underwriters are a slow squeeze, −2.6% — included for completeness, and labeled as what it is: gradual, not dramatic.

And one addition of a different kind. Interpreters and translators are, by Microsoft’s measurement, the single most AI-exposed occupation in America: roughly 98% of their work activities overlap with tasks AI demonstrably performs. The essay buried them inside “technical translators.” They deserve their own line — with the twist that makes this whole subject humbler than the headlines: the BLS still projects the occupation to grow 1.7%. The most exposed job in the country is, so far, still growing. Exposure is not the same thing as elimination, and any list that pretends otherwise is selling you something.

The Growth List, With Fresh Numbers

All ten of the essay’s “jobs to go for” survived the fact-check — the picks were good. The numbers were stale. Here they are from the current BLS 2024–34 projections, with May 2024 median wages:

  • Wind turbine service technicians: +49.9% (+6,800 jobs, $62,580 median) — still the fastest-growing occupation in America.
  • Solar photovoltaic installers: +42.1% (+12,000, $51,860).
  • Nurse practitioners: +40.1% (+128,400, $129,210) — the rare pick that is both fast-growing and huge.
  • Data scientists: +33.5% (+82,500, $112,590) — with an honest caveat below.
  • Information security analysts: +28.5% (+52,100, $124,910).
  • Medical and health services managers: +23.2% (+142,900, $117,960) — the biggest numeric gain on the list.
  • Physical therapist assistants: +22.0% (+24,500, $65,510).
  • Operations research analysts: +21.5% (+24,100, $91,290).
  • Physician assistants: +20.4% (+33,200, $133,260) — the growth rate was trimmed the most of the ten in the new cycle, from the essay’s 28.5%.
  • Computer and information research scientists: +19.7% (+7,900, $140,910) — and one detail only a real check would catch: this occupation’s median wage actually ticked down between wage reports, even as demand grows.

FACT: Two footnotes the original list didn’t have. Actuaries (+21.8%) have climbed into this tier and now outrank two of the essay’s picks. And the largest numeric job gain in the entire economy belongs to home health and personal care aides: +739,800 jobs (+17.0%) — unglamorous, physically present, human work that no chatbot is applying for.

Wind turbine technician, solar installer, and nurse practitioner — the growth side

COMMENTARY: The caveat on data scientists is the one to sit with. The occupation appears in Microsoft’s top-40 for AI task applicability even while the BLS projects a third more of them. Both are true. The growth jobs are not jobs AI never touches — they are jobs where demand is growing faster than automation eats the tasks, and where the human remainder (judgment, presence, responsibility) is the valuable part. There is no bunker on this list. There are only better and worse positions.

So What Do You Actually Do?

The essay’s core advice was sound, and the research since has only sharpened it. In our own words:

Stop competing with the machine on the machine’s terms. Speed-typing, first-draft generating, form-filling, and summarizing are now cheap. If your value at work is mostly those tasks, the Stanford numbers say the pressure arrives at hiring time first — so move your value toward the parts of the job that carry judgment, trust, and physical presence.

Learn to run the tools, visibly. Anthropic’s Economic Index found that in observed real-world use, AI is a collaborator more often than a replacement — 57% of usage was people working with it (iterating, checking, learning) versus 43% hands-off automation. The workers best positioned in every dataset we read are the ones already fluent in directing the tools.

If you are early-career, treat the narrowed door as the problem to solve. Portfolios, certifications, internships, and demonstrated work beat a bare credential when employers can automate the tasks juniors used to learn on. The growth list above is a map of where doors are widening instead — five of the top ten are in health care, and two need no bachelor’s degree at all.

If you run a business, keep a second income skill warm. That was the essay’s closing advice and it stands: backup income and a real understanding of how your business makes money are the two hedges that show up in every serious version of this conversation.

The Argument

Here is what the fact-check actually leaves standing. The viral version of this story — whole professions deleted, a neat list of safe harbors — is wrong in both directions. The BLS data shows a labor market being re-sorted, not emptied: the steepest declines are concentrated in routine clerical work (word processors −36.1%, data entry −25.9%, telemarketers −22.1%), the losses are largest where the base is largest (cashiers −313,600 jobs), and the fastest growth is in work that is physical, licensed, or responsible for other people.

The most defensible one-line summary of all five sources is Stanford’s pattern, not the headline version: AI’s first measurable effect is on the entry-level door. Occupations mostly persist; the first rung gets harder to reach, the routine middle gets squeezed, and the people who direct the tools, carry the judgment, or show up in person do fine. Microsoft’s researchers themselves warn that their exposure scores do not mean AI “can fully perform any single occupation” — and the most exposed occupation in their entire study, translators, is still projected to grow.

So: is your job on the list? If your job is mostly routine information handling, the honest answer is that parts of it are, and the time to move your value is while the slope is still a slope. If your job involves showing up, taking responsibility, or caring for people, the data says demand for you is rising. Either way, the correct response to this list is not panic and it is not comfort. It is a plan.

Frequently Asked Questions

Which job is shrinking the fastest?

FACT: Word processors and typists, projected by the BLS to decline 36.1% from 2024 to 2034 — the steepest percentage decline of any U.S. occupation. Data entry keyers are second among major clerical roles at −25.9%.

Is AI going to take my job, yes or no?

COMMENTARY: The evidence supports a more precise answer: AI is taking tasks, and occupations built mostly from routine tasks are shrinking. Anthropic’s usage data shows AI acting as a collaborator in 57% of observed use, and only about 4% of occupations see AI used across three-quarters or more of their tasks. Whole-occupation replacement is the rare case so far; task-level change is the common one.

Why did you remove tax preparers from the risk list?

FACT: Because the current BLS projections say the occupation grows 4.5% through 2034. A fact-check that only removes claims it dislikes isn’t a fact-check. The same standard removed “entry-level financial analysts” (the occupation grows 5.7%) and kept bookkeeping clerks on the list (they decline 5.8%).

Are the growth-list jobs actually safe from AI?

COMMENTARY: No job on the list is untouched — data scientists are a top-40 AI-exposure occupation in Microsoft’s study while also being the fourth-fastest-growing job in the BLS data. “Safe” means demand growing faster than the tasks automate, plus a human core — physical presence, licensure, judgment, care — that stays valuable. That is a strong position, not a guarantee.

I’m in one of the declining occupations. What’s the realistic move?

COMMENTARY: Move toward the adjacent work that is growing: the customer service rep who becomes the escalation and quality specialist (the role Klarna ended up rehiring for), the bookkeeper who becomes the person who reviews and advises rather than keys in, the junior developer who builds a portfolio of shipped work. Every dataset we read points the same direction: the premium moves to judgment, verification, and client trust.

What’s the single biggest opportunity number in the data?

FACT: Home health and personal care aides add the most jobs of any occupation in America — about 739,800 by 2034. For a high-wage version of the same pattern, nurse practitioners combine +40.1% growth, 128,400 new jobs, and a $129,210 median wage.

Sources: U.S. Bureau of Labor Statistics, Employment Projections 2024–34 (Aug 2025); World Economic Forum, Future of Jobs Report 2025; Microsoft Research, “Working with AI” (2025); Stanford Digital Economy Lab, “Canaries in the Coal Mine?” (2025); Anthropic Economic Index (2025); Klarna/Bloomberg reporting (2024–25). Original essay: PomeroySays, Medium, Sep 2026.

Want Jackson’s top local stories in one place? Browse the businesses shaping the area in the NeighborClick Local directory — JOIN the FREE Private Insider List for local stories, business spotlights, community updates and even if you don’t have business you’ll be an Insider.

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *

Sign In

Register

Reset Password

Please enter your username or email address, you will receive a link to create a new password via email.