AI and Entry-Level Tech Jobs: Skills to Build in 2026
AI and Entry-Level Tech Jobs: Skills to Build in 2026
Updated September 26, 2026. Every figure below was read that day from the Bureau of Labor Statistics, the National Center for Education Statistics, the Federal Reserve Board and O*NET, all linked at the end. Where a number is a projection, it says so; where it is a measurement, it says so; and the one federal paper that measures rather than forecasts is flagged as such. Ratios and per-year figures are arithmetic done here from those sources.
Two BLS tables, one revision, and why the date on the projection matters
The Employment Projections program rolled to a new base period on August 27, 2026: projections now run 2025–2035, replacing the 2024–2034 cycle. That is not housekeeping. Software developers were projected to grow about 15.8% and add roughly 267,700 jobs under the old cycle. Under the new one the same occupation is projected at 10% and 185,400 jobs. Nothing about the work changed between those two publications; the model's read of it did, by about 82,000 jobs.
So the first practical skill is reading the vintage stamp on any number anyone quotes at you about tech careers. O*NET, the companion database most career pages pull from, was still showing the older 2024–2034 figures for software developers when checked for this article. Two federal sources, both current, disagreeing by five percentage points, because one has not refreshed yet.
The eight occupations, sorted by the direction they are actually moving
All figures are the 2025–2035 projection cycle published August 27, 2026, with May 2025 median wages. For scale: all occupations together are projected to grow 3.5%, adding 5.9 million jobs and taking total employment from 170.3 million to 176.2 million, against a median annual wage of $50,980 across every job in the economy. BLS notes the previous decade grew 10.9%, so 3.5% is the slow baseline everything below is measured against.
| Occupation | 2025–35 change | Jobs | Median wage | Typical entry education |
|---|---|---|---|---|
| Data scientists | +35% | +95,400 | $120,230 | Bachelor's |
| Computer and information research scientists | +22% | +8,400 | $140,300 | Master's |
| Information security analysts | +21% | +40,600 | $129,180 | Bachelor's |
| Software developers (with QA analysts and testers) | +10% | +185,400 | $134,040 | Bachelor's |
| Computer network architects | +8% | +14,100 | $134,050 | Bachelor's |
| Web developers (with digital designers) | +5% | +11,300 | $99,520 | High school to bachelor's |
| All occupations, for comparison | +3.5% | +5,900,000 | $50,980 | — |
| Computer support specialists | −3% | ~48,700 annual openings despite decline | $62,890 | Associate's to some college |
| Computer programmers | −7% | −8,100 | $100,390 | Bachelor's |
The computer and mathematical occupations group as a whole is projected at +7.3%, which BLS calls the fifth-fastest-growing major group. That single number hides the spread in the table: the distance between data scientists and computer programmers is 42 percentage points inside one major group.
Two lines deserve a second read. Computer support specialists are projected to shrink and still generate about 48,700 openings a year, because replacement demand in a base of 903,100 workers swamps the net change — a declining occupation can still be the easiest one to get hired into. And computer programmers, at 110,800 workers, is now a small occupation; the separate OEWS survey counts 92,230 in the specific programmer code against 1,687,890 software developers. "Programmer" as a job title has been absorbed, not eliminated.
The supply side nobody puts in the same table
Projections describe demand. The other half of an entry-level outcome is how many people arrive at the same door. NCES counts bachelor's degrees conferred in computer and information sciences at 108,503 in 2021–22, up from 71,416 five years earlier and 47,406 ten years earlier — a 129% increase over the decade, which NCES describes as more than doubling. No more recent year was published at the time of writing.
Put that against the projection table and one ratio falls out. Software developers are projected to add 185,400 jobs over ten years, which is about 18,540 net new positions a year, against roughly 108,500 computing graduates arriving each year. Net growth is not the same as hiring — replacement demand adds far more openings than net change — but the ratio is the reason the phrase "entry-level" is doing so much work in this conversation. The bottleneck is not whether the occupation grows. It is that the entering cohort is roughly six times the annual net growth of the largest occupation in the field.
The current labour market, measured rather than projected
For August 2026, BLS reports overall unemployment at 4.1%. The information sector's unemployment rate was 5.7% not seasonally adjusted; professional and business services was 3.8%. Information-sector payrolls fell by 23,000 in the month, with losses in computing infrastructure, data processing and web hosting (−8,000), publishing (−7,000) and broadcasting (−5,000).
The New York Fed's tracker for recent college graduates put their unemployment rate at 5.6% and their underemployment rate at 42% as of the second quarter of 2026. Underemployment — working in a job that does not typically require a degree — is the statistic that matches what the phrase "I can't get an entry-level job" usually means, and it is far larger than the unemployment rate.
The one federal estimate that measures the AI effect instead of forecasting it
A Federal Reserve Board working paper published March 20, 2026 uses Current Population Survey data linked to O*NET occupational classifications in an event-study design. Its finding: annual employment growth among "coders" decelerated by roughly 3 to 3.9 percentage points after November 2022. Pre-period growth ran about 4.8% a year across all industries and about 6.1% in coder-intensive industries; the latter flattens to near zero afterwards. Cumulated over about three years, the authors put the gap at roughly 500,000 jobs relative to the counterfactual trend.
The paper's own caveats matter as much as the number. This is reduced hiring relative to trend, not measured job destruction, and the design does not follow workers who moved into other roles. It also cannot separate AI from the interest-rate cycle and the post-2022 correction in tech headcount. It is, however, the only backwards-looking federal estimate available; GAO's most recent workforce-automation work predates the large-language-model era entirely, and BLS's own AI commentary is projection, not measurement.
Turning the table into a decision about the next twelve months
The useful conversion is not "which occupation grows fastest" — that is a headline, and data scientists at +35% of a 275,600 base add fewer jobs than software developers at +10% of a 1,905,400 base. The useful conversion is jobs added per year against wage, which is what a two-year training decision is actually trading.
| Track | Net jobs added per year (projection ÷ 10) | Median wage | What the number is sensitive to |
|---|---|---|---|
| Software developer | ~18,540 | $134,040 | Revised down ~82,000 jobs between two consecutive BLS cycles |
| Data scientist | ~9,540 | $120,230 | Small base; a 35% rate on 275,600 is fewer jobs than 10% on 1.9 million |
| Information security analyst | ~4,060 | $129,180 | Base of 192,900; steady growth, no cycle revision flagged |
| Computer network architect | ~1,410 | $134,050 | Small absolute flow; wage close to developer |
| Support specialist | Negative net, ~48,700 openings/yr | $62,890 | Openings driven by turnover, not growth |
| Computer programmer | ~−810 | $100,390 | Title being absorbed into developer classification |
Read the last column first. The developer line has the biggest annual flow and the biggest measured forecasting error; the security line has a smaller flow and no revision attached to it. That is a trade-off between size and confidence, and it is the actual shape of the decision.
Tip: Information security analysts are the one line in the table with above-20% projected growth and no cycle revision attached. If that is the track, a Security+ study guide is the cheapest way to find out within a month whether the material holds your attention. (These are Amazon Associate links — we may earn a small commission on qualifying purchases.)
What the projections cannot tell you, stated plainly
Three limits are worth carrying around. First, the 2025–2035 cycle is one year old at most and was revised materially from its predecessor; treat any single occupation's rate as a range, not a point. Second, no federal source publishes an occupation-level measurement of how much of a specific job AI now does — the Fed paper measures employment growth, not task share, and NIST-style task studies do not exist for these occupations. Third, the degree data stops at 2021–22, so the supply ratio above uses a four-year-old numerator against a 2026 projection.
There is also a disagreement to keep in view rather than resolve: BLS projects computer and mathematical occupations growing at double the all-occupation rate over ten years, while the Fed paper finds coder employment growth already flat over the last three. Those are not contradictory — one is a decade forecast and one is a three-year measurement — but anyone citing only one of them is telling half the story. If you have to be wrong in one direction, being wrong by over-preparing for a slower market is cheaper than the reverse.
Where this lands if the question is what to learn next
On the published numbers alone, the defensible reading is that the field is not shrinking and the entry door is narrowing anyway, for a reason that has nothing to do with AI: a graduating cohort of roughly 108,500 a year against net growth in the tens of thousands. That makes the differentiator whatever separates a candidate from the same cohort, and the table above points at two places where the cohort is thin — the security line, where growth is 21% and the entering pipeline is not a dedicated degree for most people, and the support line, where 48,700 openings a year exist in an occupation students are told to avoid because it is declining.
The counter-argument deserves space: median wages in support ($62,890) are less than half the developer median ($134,040), so the easier door opens onto a lower floor, and the projection says the floor is sinking. Anyone taking that route should treat it as a two-year entry point with a named exit, not a destination.
Five checks before committing to a track
- Find the projection's base period on the BLS page itself. If it says 2024–2034, it has been superseded by the August 27, 2026 release.
- Read the jobs column, not the percent column. A 35% rate on a 275,600 base is smaller in people than a 10% rate on 1.9 million.
- Check whether the occupation's openings come from growth or from turnover — BLS prints annual openings separately, and a declining occupation with 48,700 openings is a real door.
- Compare the median wage against $50,980, the all-occupation median, before treating any of these as a low-paying option.
- Assume the entering cohort is about 108,500 a year and ask what, specifically, takes you out of it. "Knows Python" does not.
For related coverage, see our look at what Nvidia's quarter says about where the money is going and our comparison of the AI tools most useful for everyday work.
Where each figure came from
- BLS Employment Projections, 2025–2035 — released August 27, 2026. All-occupation totals and the major-group rate.
- BLS Occupational Outlook Handbook, computer and IT occupations — per-occupation projections, job counts, median wages and entry education.
- BLS Occupational Employment and Wage Statistics, May 2025 — employment and median wage by occupation code; the all-occupation median of $50,980.
- BLS Employment Situation, August 2026 — overall unemployment and information-sector payroll change.
- BLS Employment Situation, Table A-14 — unemployment by industry, August 2026.
- Federal Reserve Bank of New York, labour market for recent college graduates — 5.6% unemployment, 42% underemployment, 2026 Q2.
- NCES Digest of Education Statistics, table 325.35 — bachelor's degrees in computer and information sciences by year.
- Federal Reserve Board, Finance and Economics Discussion Series, March 20, 2026 — measured deceleration in coder employment growth.
- O*NET, software developers — Bright Outlook designation and the older projection figures still displayed.
Employment projections are estimates, not promises, and BLS revises them on a published schedule. This article is general information about public labour-market data, not career or financial advice for any individual situation.
Comments
Post a Comment