AI Labor Tracker / Foundations

The job thermometer: which work is cooling and which is heating

Postings, payrolls and pay all point the same way: the screen-based jobs AI handles easily are cooling, and the work of building and wiring the AI economy is running hot.

If you could take the temperature of every job in America, where would the reading be hottest? Right now, in the jobs you need boots for. The coldest readings are in jobs that live entirely on a screen, and the latest postings, payroll and pay data say the gap is wider than most people think.

First, the good part. AI is making a lot of work faster and cheaper, and the overall labor market is holding up.

Employers added 29,000 jobs in September and unemployment sat at 4.2%, according to the Bureau of Labor Statistics. That is a cool market, not a collapsing one.

But averages hide the story. The question this publication keeps asking is where the work is going, and who is left holding the jobs it leaves. So we built a thermometer, occupation by occupation, from three sources: what employers are advertising, how many people actually hold each job, and what those jobs pay.

How the thermometer works

The first reading is job postings. Indeed’s postings index stood at 103.8 on October 16, almost 4% above its February 2020 level, and 68% of occupational sectors had more postings than a year earlier, up from 49% at the start of September, according to Indeed Hiring Lab. Postings tell you where employers want to hire next.

The second reading is employment. The Bureau of Labor Statistics’ Occupational Employment and Wage Statistics survey counts workers in hundreds of occupations as of May each year. We compared May 2022, the last count before ChatGPT launched that November, with May 2025, the latest release, using the national data files.

The third reading is pay. If employers are fighting over workers, wages rise faster than average. If not, they lag.

One caution before the numbers. The survey counts payroll jobs, not the self-employed, and BLS warns that occupation definitions and methods shift between years. Treat these as temperature readings, not lab results.

The cold end

Computer programmers are the coldest job on our board. Their numbers fell from 132,740 in May 2022 to 92,230 in May 2025, a drop of 30.5%, according to BLS data. Over the same three years, total employment grew 5.1%.

Programmers are not alone. Web developers fell 20.8%. Writers and authors fell 11.2%.

Customer service representatives, one of the biggest occupations in the country, fell 9.9%, a loss of about 284,000 jobs. Graphic designers fell 6.6%.

The same pattern shows up in smaller occupations. Technical writers went from 48,620 to 45,500.

Software quality assurance testers went from 196,420 to 186,740. Word processors and typists went from 41,990 to 35,010.

Some of the slide began before most firms had AI tools, and a three-year comparison cannot prove cause. So I would not pin all of it on AI. But the direction matches what other trackers see.

Revelio Labs found global call-center headcount 5.3% below its December 2023 peak, after more than a decade of growth, with declines in eight straight quarters (Revelio Labs). Employers cited AI in 120,136 announced job cuts so far this year, about 21% of all cuts, making it the leading reason year-to-date, according to Challenger, Gray & Christmas.

The official projections agree. BLS expects customer service representatives to shrink 5.5% between 2024 and 2034, with medical transcriptionists, claims adjusters, legal secretaries and procurement clerks all expected to decline as well (BLS). Total employment is projected to grow 3.1%.

The hot end

Now the warm readings. Electricians grew 9.7% over the same three years, to 757,220. Electrical power-line installers also grew 9.7%.

HVAC mechanics and installers grew 9.3%. Plumbers and pipefitters grew 8.9%. Construction laborers went from 1,012,780 to 1,096,780.

Pay tells the same story in several trades. The median wage for all occupations rose 10.1% between May 2022 and May 2025. For HVAC mechanics it rose 18.7%, to $61,010.

For power-line installers it rose 15.8%, to $95,320. Electricians were the exception, with pay up 4.9%, to $63,190, even as their numbers grew.

The forecasts are strong too. BLS projects electricians to grow 9% over the decade to 2035, HVAC mechanics 11% and power-line installers 10%, all “much faster than the average for all occupations,” with about 72,700 openings for electricians every year (BLS; BLS; BLS).

None of these jobs need a four-year degree. Most need an apprenticeship and a license. That matters to me, because the good jobs without a degree are where the dignity of work gets proved or lost in this transition.

The postings data are even hotter. Indeed’s mechanical engineering index jumped 44.4% from a year earlier, and electrical engineering rose 33.4%, according to Indeed’s sector data. Construction postings are up 6.1%, and installation and maintenance 3.5%, against 3.1% for all postings.

A lot of that heat comes from one place. Data center job postings have grown 20% in three months and are now 147% above their June 2024 level, according to Indeed. About a quarter of data center openings are for installation and maintenance workers, and the 10 largest tech firms account for 71% of data center postings this year (Indeed).

It is the old gold-rush pattern. The people selling picks and shovels did fine.

In this rush, the picks and shovels are transformers, cooling systems and the people who install them. We look at one of those jobs in our piece on the data center electrician.

The lukewarm middle

Not every job fits neatly at one end. Some of the occupations people assume AI will hit first are holding steady, and a few are growing.

Interpreters and translators, a favorite example of AI exposure, were essentially flat, at 52,060 in May 2025 against 52,160 three years earlier. BLS still projects 2% growth for them from 2025 to 2035, slower than average (BLS). Paralegals grew from 345,240 to 392,880.

Market research analysts grew from 798,620 to 899,580. All three spend their days on language and documents, and all three held up better than the forecasts in many headlines.

My guess is that in these jobs AI is taking over tasks rather than whole jobs, at least so far. A paralegal who can check an AI summary against the actual file is worth more, not less. That guess needs better data than we have, and we will keep testing it.

The postings middle is softer. Indeed’s indices for administrative assistance, customer service and marketing all sit below their pre-pandemic levels and slightly below a year ago. Media and communications and data and analytics are up from last year but still far below 2020, at 67.6 and 62.4 on an index where February 2020 equals 100.

Pay is not giving workers much cover either. Posted wages rose 2.4% from a year earlier in September, and after inflation take-home pay was 0.1% lower, according to Indeed. Electrical engineering postings were an exception, with advertised pay up 5.6%.

The closing entry ramp

The fear I hear most is simple: if you are young and your work happens on a laptop, the door is closing.

There is evidence behind that fear. Stanford economists using ADP payroll data found that employment of workers aged 22 to 25 in the most AI-exposed occupations is now 19% below where it would be had it kept pace with their less-exposed peers, and experienced workers show no comparable gap (Brynjolfsson, Chandar and Chen). Recent college graduates had an unemployment rate of about 5.6% in the second quarter, and 42% were underemployed, according to the New York Fed.

That evidence deserves weight. But three facts in the data cut against the idea that screen work is finished.

First, software developers are growing. Their employment rose 10% between 2022 and 2025, to 1,687,890, with a median wage of $135,980. Indeed’s software development postings are up 23.8% from a year ago.

They are still about 19% below February 2020, so this is a rebound from a deep hole. And it is concentrated: 71% of the increase in software postings between May 2025 and May 2026 came from senior roles, and 37% from jobs with AI in the title (Indeed).

Second, the fastest-growing projections are technical. BLS expects data scientists to grow 33.5% over the decade, information security analysts 28.5% and actuaries 21.8% (BLS).

Third, the freelance market is paying for people who use the tools. Upwork found that freelancer earnings tied to skills that explicitly reference AI grew 109% in 2025, compared with 23% for all other skills, and AI video work grew 329% (Upwork).

So the honest reading is narrower than “AI kills office jobs.” Routine screen work is cooling. Entry-level screen work is cooling fastest.

Screen work that sits on top of AI, or that needs senior judgment, is still hiring. Here’s where I land: the biggest risk is not a sudden wave of layoffs. It is a slow closing of the entry ramp, where the junior tasks that used to train people get automated first.

If you hire, that is the decision in front of you right now: whether to backfill the junior role a tool can partly cover, or keep it as the place your next senior person learns the job.

What the pay data won’t tell you yet

One reading surprised me. Customer service jobs shrank by almost 10%, yet the median wage for the people still doing them rose from $37,780 to $44,770, faster than the national median. Fewer jobs, higher pay.

I can think of two explanations, and the data can’t separate them yet. Either the simple calls went to chatbots and the remaining human work is harder and better paid, or the lowest-paid jobs simply disappeared from the count.

Both could be true. Revelio’s tracking of former call-center workers is a warning here: only 10.8% of those who moved landed in higher-paying technical support, customer success, or software and data roles (Revelio Labs).

Nobody called all of this in advance, which is why we built a thermometer instead of a forecast.

We will update this board as new data land. The October jobs report is due November 6 (BLS). The machine side of the ledger lives in our Gross Intelligence Product, and the bigger frame is in what post-labor economics means.

What to do with the reading

Cheap knowledge and scarce creativity get their own piece in this series, but the thermometer already shows the split. The work cooling fastest is work where the answer can be looked up or generated. The work heating up needs hands, judgment, or both.

If you are early in your career and your job lives on a screen, get good at the tools now. Learn the AI tools in your field well enough that you are the person who checks the machine’s work, because that is where the senior hiring is.

If you are open to a different path, take a hard look at the electrical, HVAC and line trades. They pay well, they don’t need a degree, and the demand data are as strong as anything on this board.

If you run a business, keep hiring juniors even when a tool can do their first-year tasks. Someone has to become your senior people in 10 years. And if you make policy, fund apprenticeships the way we fund college, because that is where the heat is.

Sources

1. Indeed Hiring Lab, “US Labor Market Snapshot: October 2026,” Oct 22, 2026. https://hiringlab.indeed.com/2026/10/22/us-labor-market-snapshot-october-2026/
2. Indeed Hiring Lab, Job Postings Tracker data (sector indices, through Oct 23, 2026), GitHub. https://github.com/hiring-lab/job_postings_tracker
3. Indeed Hiring Lab, “Hiring for the Data Center Build-Out,” Jul 14, 2026. https://hiringlab.indeed.com/2026/07/14/hiring-for-the-data-center-build-out/
4. Indeed Hiring Lab, “AI and Job Postings: From Destruction to Creation?,” Jul 8, 2026. https://hiringlab.indeed.com/2026/07/08/ai-and-job-postings-from-destruction-to-creation/
5. U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics, national tables, May 2022 and May 2025. https://www.bls.gov/oes/tables.htm
6. U.S. Bureau of Labor Statistics, OEWS national data file, May 2025. https://www.bls.gov/oes/special-requests/oesm25nat.zip
7. U.S. Bureau of Labor Statistics, OEWS national data file, May 2022. https://www.bls.gov/oes/special-requests/oesm22nat.zip
8. U.S. Bureau of Labor Statistics, “Artificial intelligence, information technology, and employment, 2024–34,” The Economics Daily, Jul 16, 2026. https://www.bls.gov/opub/ted/2026/artificial-intelligence-information-technology-and-employment-2024-34.htm
9. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. https://www.bls.gov/ooh/construction-and-extraction/electricians.htm
10. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Heating, Air Conditioning, and Refrigeration Mechanics and Installers. https://www.bls.gov/ooh/installation-maintenance-and-repair/heating-air-conditioning-and-refrigeration-mechanics-and-installers.htm
11. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Electrical Power-Line Installers and Repairers. https://www.bls.gov/ooh/installation-maintenance-and-repair/line-installers-and-repairers.htm
12. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Interpreters and Translators. https://www.bls.gov/ooh/media-and-communication/interpreters-and-translators.htm
13. U.S. Bureau of Labor Statistics, “The Employment Situation, September 2026,” Oct 2, 2026. https://www.bls.gov/news.release/empsit.htm
14. Challenger, Gray & Christmas, “Challenger Report, September 2026,” Oct 1, 2026. https://www.challengergray.com/wp-content/uploads/2026/10/Challenger-Report-September-2026.pdf
15. Revelio Labs, “After a Decade of Growth, Global Call Center Employment Is Shrinking,” 2026. https://www.reveliolabs.com/news/ai-and-work/after-a-decade-of-growth-global-call-center-employment-is-shrinking
16. Brynjolfsson, E., Chandar, B., and Chen, R., “Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence,” Stanford Digital Economy Lab, revised Aug 12, 2026. https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/
17. Federal Reserve Bank of New York, “The Labor Market for Recent College Graduates,” updated Q2 2026. https://www.newyorkfed.org/research/college-labor-market
18. Upwork Research Institute, “In-Demand Skills 2026,” Feb 4, 2026. https://www.upwork.com/research/in-demand-skills-2026