Last Updated on September 5, 2026 by Van Phillips
I sit with numbers for a living. Between watching futures tick on my screens and building out my own little corner of the internet, I’ve learned that the charts that move fastest ain’t always the ones that matter most. Sometimes the real story is sitting in a government PDF, released quiet on a Thursday morning, that most folks will scroll past on their way to check their portfolio.
That’s exactly what happened last week. The Bureau of Labor Statistics dropped their Employment Projections for 2025–2035, and this time they came with something new attached — a whole framework for measuring how exposed different jobs are to AI. Between the main release and a few companion deep-dives BLS has published on construction and food manufacturing, there’s a real, detailed map of where this economy is headed. If you make your living off what this economy needs — or you’re raising somebody who’s about to — you need to sit with this one.
Let me break it down the way I’d break it down for my people at the kitchen table.
The Headline: Growth, But Slower Growth
Here’s the top line. The U.S. economy is projected to add 5.9 million jobs between now and 2035, taking total employment from 170.3 million up to 176.2 million. That sounds like a lot — and it is — but context matters. That’s 3.5% growth over a decade. Compare that to the 10.9% growth the economy put up between 2015 and 2025, and you start to see the shape of what’s coming.
We’re not looking at a jobs collapse. We’re looking at a labor market that’s maturing — slower, more selective about where it puts its chips down.
Where the Jobs Are Actually Going
The growth isn’t spread evenly. Three sectors are doing the heavy lifting:
| Industry Sector | Projected Growth (2025–35) | New Jobs (thousands) |
|---|---|---|
| Utilities | +9.8% | 58.8 |
| Healthcare & social assistance (private) | +9.5% | 2,204.8 |
| Professional, scientific & technical services | +8.6% | 926.7 |
| Federal government | -3.4% | -98.5 |
| Retail trade | -0.2% | -27.5 |
Healthcare isn’t just growing fastest in percentage terms — it’s carrying the whole economy. That one sector alone is expected to account for about 37% of every new job created through 2035, driven by an aging population and rising rates of chronic conditions.
The Piece Nobody Was Expecting: BLS Is Now Grading Every Job’s AI Exposure
Alongside this year’s release, BLS quietly rolled out something brand new: an AI Exposure Categories dataset. They took every detailed occupation in the country and sorted it into one of four buckets — Low, Moderate, High, or Very High — based on how exposed that work is to AI, pulling from five different research sources to build the classification. I sat with the full dataset, and it’s the most interesting thing in this whole release.
Exposure doesn’t mean job loss. BLS is explicit about that. It means AI could be or has been used to assist or complete some of the work. What you do with that information is up to you.
Here’s what the aggregate numbers look like when I ran them:
| AI Exposure Category | 2025 Employment (millions) | Projected Growth 2025–35 | Avg. Median Wage |
|---|---|---|---|
| Very High | 53.2 | +2.0% | $83,285 |
| High | 42.1 | +3.1% | $89,570 |
| Moderate | 44.7 | +6.2% | $69,557 |
| Low | 30.3 | +2.5% | $58,130 |
Now look closer, because the pattern isn’t what you’d guess. “Very High” exposure occupations are actually the slowest-growing bucket in the whole classification. Meanwhile “Moderate” exposure jobs — the ones where AI assists but doesn’t dominate the task — are growing more than three times as fast. And “High” exposure jobs, on average, pay the most of any category, higher even than “Very High.” AI exposure and job security don’t move in a straight line together, and neither does AI exposure and pay.
Zoom into the “Very High” category and the split gets sharper. This bucket is where you’ll find:
| Occupation (Very High exposure) | 2025 Employment | Change 2025–35 | Median Wage |
|---|---|---|---|
| Word processors and typists | 40.4k | -34.4% | $49,280 |
| Data entry keyers | 131.8k | -25.5% | $41,340 |
| Office clerks, general | 2.6M | -6.0% | $45,010 |
| Bookkeeping, accounting & auditing clerks | 1.5M | -5.6% | $50,670 |
| Software developers | 1.7M | +10.2% | $135,980 |
| Accountants and auditors | 1.6M | +5.0% | $83,680 |
Same exposure classification, completely different outcomes. Word processors and data entry keyers are shrinking fast, while software developers and accountants — also flagged as “Very High” exposure — are growing. The difference isn’t whether AI touches the job. It’s whether the human judgment layered on top of that work is easy to strip away or hard to.
Compare that against the “Low” exposure column, which is dominated by physical, hands-on work: laborers and freight movers (2.9M jobs, +1.8%), janitors and cleaners (2.4M, +2.2%), construction laborers (1.5M, +7.3%). And sitting right at the top of “Moderate” exposure, by far the single largest occupation in the entire dataset: home health and personal care aides, 4.7 million strong, projected to grow 18.1%. That’s the aging-population story showing up again, in a different chart.
AI Is Playing Both Sides
AI shows up in this report as both a builder and a bulldozer. It’s creating the fastest-growing jobs in tech and energy, and it’s quietly hollowing out the categories that used to be considered safe.
On the growth side: computing infrastructure, data processing, and web hosting is projected to grow 25.1%, adding 120,400 jobs. Data scientists are up 34.6%. On the other side: office and administrative support occupations are projected to shrink 4.0%, losing 752,100 jobs — the steepest decline of any major occupational group in the entire report.
The Construction Boom Nobody’s Talking About
Here’s a piece I wasn’t expecting to find as compelling as I did. Construction employment hit an all-time high of 8.0 million jobs in 2023, finally surpassing the pre-recession peak of 7.7 million from 2006 — after taking over a decade to climb back from losing 2 million jobs in the Great Recession. And BLS’s deeper research on this sector points to three forces stacking on top of each other to keep it growing:
- Electricity capacity expansion — total U.S. electric power capacity is projected to grow 42.3% by 2033, with renewable sources alone up 141.8%. Somebody has to physically build all of that.
- AI data centers — demand from data centers could climb from 4% to as much as 9–12% of total U.S. electricity generation by 2030. One AI query uses roughly 10 times the electricity of a regular search.
- EV charging infrastructure — EV sales went from under 1% of new vehicle sales in 2013 to nearly 10% by 2023, and every one of those cars needs somewhere to plug in.
The occupations riding this wave are specific and worth knowing:
| Construction-Adjacent Occupation | Projected Growth | Median Wage |
|---|---|---|
| Wind turbine service technicians | +60.1% | $61,770 |
| Solar photovoltaic installers | +48.0% | $48,800 |
| Electricians | +10.8% | $61,590 |
| HVAC & refrigeration mechanics | +9.1% | $57,300 |
| Construction laborers | +8.2% | $45,300 |
| Electric power-line installers | +7.6% | $85,420 |
Every AI data center is, underneath the servers, a construction project — concrete, steel, electrical wiring, and cooling systems, built by people whose hands never touch a keyboard for a living.
What’s Cooking: Food and Beverage Manufacturing
One more corner of this report deserves its own light, because I think people sleep on manufacturing as a growth story. Food and beverage manufacturing is projected to add more new jobs than any other manufacturing sector — about 130,000 by 2034 — growing at 6.2%, twice the pace of the overall economy, even while total manufacturing employment barely moves.
The engine here is convenience and protein. Beverage manufacturing is the fastest-growing piece (+11.4%), riding demand for gut-health drinks, energy drinks, and enhanced waters. Meat processing is adding the second-most jobs of any food industry, pushed by a dietary shift toward higher protein intake. And “other food manufacturing” — your snacks, your prepared meals — is growing 8.4% because, plainly, Americans snack: more than 90% of adults eat at least one snack a day now.
But it’s not a free ride. Robot adoption in food manufacturing grew 227.5% over the 2010s, and it’s already showing up in the numbers — grain and oilseed milling, which has leaned hardest into automated precision equipment, is projected to grow just 0.5%, and seafood processing is actually projected to decline 2.5% even as its output keeps climbing. Same story as the AI exposure data: automation raises output, but it doesn’t raise headcount evenly across every corner of an industry.
The 10 Fastest-Growing Jobs in America
| Occupation | Projected Growth | New Jobs (thousands) |
|---|---|---|
| Nurse practitioners | +41.0% | 137.8 |
| Solar photovoltaic installers | +36.5% | 11.3 |
| Data scientists | +34.6% | 95.4 |
| Wind turbine service technicians | +29.5% | 3.5 |
| Medical and health services managers | +24.2% | 155.1 |
| Physical therapist assistants | +23.0% | 26.2 |
| Psychiatric technicians | +22.3% | 36.0 |
| Computer and information research scientists | +21.8% | 8.4 |
| Occupational therapy assistants | +21.5% | 11.2 |
| Ophthalmic medical technicians | +21.4% | 15.5 |
Seven out of ten of those roles are healthcare-adjacent. Every chart in this report keeps landing back on the same place.
What Your Education Actually Buys You in 2025
| Education Level | Median Weekly Earnings | Unemployment Rate |
|---|---|---|
| Doctoral degree | $2,307 | 1.8% |
| Professional degree | $2,294 | 1.9% |
| Master’s degree | $1,876 | 2.6% |
| Bachelor’s degree | $1,578 | 2.8% |
| Associate’s degree | $1,135 | 3.0% |
| Some college, no degree | $1,062 | 3.8% |
| High school diploma | $966 | 4.3% |
| Less than high school | $770 | 6.1% |
| All workers (average) | $1,268 | 3.4% |
Every rung up that education ladder buys you both more money and more job security at the same time. That’s the pattern the data’s shown for decades, and 2025 is no exception.
One honest caveat — the 2025 figures are 11-month averages excluding October, since data wasn’t collected that month due to the federal government shutdown.
The Workforce Is Getting Older, Point Blank
Back in 2005, workers 65 and older made up just 3.5% of the labor force. By 2025, that hit 6.9%. Projected to 2035: 8.2%. Meanwhile workers 16 to 24 keep shrinking as a share of the workforce, from 14.9% down to a projected 11.4%. That demographic engine is underneath nearly every healthcare number in this entire report.
My Takeaway
I read reports like this the same way I read a chart before I put on a trade — not for the headline, for the structure underneath it. And this year’s structure has an extra layer nobody had access to before: an actual, occupation-by-occupation read on AI exposure. That’s not a forecast, it’s a map — and I’d encourage anybody reading this to go look up their own job title in BLS’s classification and see where they land. Knowing the category is one thing. Knowing that the category alone doesn’t determine your outcome — that the humans who kept the judgment-heavy parts of their work intact are the ones still growing inside it — that’s the actual lesson sitting in this data.
I’ll be back with more of these breakdowns as it keeps coming. Stay sharp out here.
Full release and detailed tables available at BLS Employment Projections, with the AI exposure methodology at bls.gov/emp/publications/ai-exposure-categories.htm, and occupational deep-dives in the Occupational Outlook Handbook.
