By Aijaz Alam
Is AI replacing jobs?
95.7% of US firms that used AI said it left their employment unchanged over the last six months, in the Census Business Trends and Outlook Survey AI supplement (17 Nov 2025 to 8 Feb 2026). Only 2.0% said AI cut staff. The strain shows in hiring: Stanford payroll data put young workers in AI-exposed jobs 19% behind peers (June 2026).
- 95.7%AI-using US firms where AI did not change total employmentCensus BTOS AI supplement, 17 Nov 2025 to 8 Feb 2026
- 2.0%AI-using US firms where AI decreased total employmentCensus BTOS AI supplement, 17 Nov 2025 to 8 Feb 2026
- 19%Employment gap for US workers aged 22 to 25 in the most AI-exposed jobsStanford Digital Economy Lab, ADP payroll data to June 2026
- 39%Survey respondents expecting AI to cut their organization's headcount next yearMcKinsey State of AI, May-Jun 2026
- 49%Firms expecting no noticeable AI effect on staffing in the next 12 monthsSt. Louis Fed district survey, May 2026
- 62%Average wage premium in job ads that ask for AI skillsPwC Global AI Jobs Barometer, 2025 job ads
The numbers
The full breakdown
How AI affected total employment at US firms that used it (17 Nov 2025 to 8 Feb 2026)
| Effect of AI on total employment, last six months | Share of AI-using firms |
|---|---|
| No change | 95.7% |
| Increased | 2.3% |
| Decreased | 2.0% |
Census BTOS AI supplement, six biweekly panels pooled, released 18 Jun 2026. Base: firms that used AI in at least one business function, about 17.9% of US employer firms. Firms report their own view; this is not a payroll count.
Effect of AI on total employment at AI-using US firms, by sector (17 Nov 2025 to 8 Feb 2026)
| Sector (NAICS) | No change | Decreased | Increased |
|---|---|---|---|
| Information | 91.7% | 3.8% | 4.5% |
| Professional services | 94.3% | 2.4% | 3.3% |
| Firms in several sectors | 94.8% | 2.6% | 2.6% |
| Finance | 95.0% | 1.5% | 3.5% |
| Transportation | 95.1% | 3.2% | 1.7% |
| Wholesale | 95.2% | 2.1% | 2.8% |
| Retail | 95.4% | 2.7% | 1.9% |
| Administrative and support | 95.8% | 2.3% | 1.9% |
| Health care | 96.1% | 1.6% | 2.2% |
| Education | 96.2% | 1.7% | 2.0% |
| Real estate | 96.5% | 1.3% | 2.2% |
| Accommodation and food services | 96.7% | 2.3% | 1.0% |
| Manufacturing | 97.1% | 1.1% | 1.7% |
| Arts and entertainment | 97.1% | 0.9% | 2.0% |
| Other services | 97.1% | 1.7% | 1.2% |
| Construction | 97.4% | 1.1% | 1.5% |
| All AI-using firms | 95.7% | 2.0% | 2.3% |
Census AI supplement estimates file, question 6. Left out: agriculture, mining, utilities and management of companies, where Census suppresses the increased and decreased shares (their no-change shares are 96.7%, 97.7%, 97.8% and 92.1%).
What US firms that used AI did with it and for it, last six months (17 Nov 2025 to 8 Feb 2026)
| Measure (AI-using firms) | Share |
|---|---|
| Used AI to supplement or enhance a task an employee does | 43.7% |
| Used AI to introduce a new task not done by an employee before | 10.6% |
| Used AI to perform a task an employee used to do | 10.1% |
| None of these | 51.5% |
| Of firms where AI did employee tasks: a small number of tasks | 70.9% |
| Of firms where AI did employee tasks: a large number of tasks | 7.1% |
| Trained current staff to use AI | 15.0% |
| Developed new workflows | 15.4% |
| Hired staff trained in AI | 1.3% |
| Made no changes to use AI | 64.3% |
Census AI supplement estimates file, questions 3, 4 and 7. Firms could pick more than one answer. 22.0% of firms where AI did employee tasks said it was a moderate number of tasks.
What firms and managers expect AI to do to staffing over the next year
| Survey | Expectation | Share |
|---|---|---|
| St. Louis Fed, Eighth District firms (May 2026) | No noticeable effect on staffing in the next 12 months | 49% |
| St. Louis Fed, Eighth District firms (May 2026) | Skill shifts rather than headcount changes | About 18% |
| St. Louis Fed, Eighth District firms (May 2026) | Slight reductions in staffing needs | Nearly 20% |
| McKinsey, global respondents (May-Jun 2026) | AI will decrease the organization's headcount over the next year | 39% |
| McKinsey, global respondents (May-Jun 2026) | Little or no change in headcount | 43% |
| McKinsey, global respondents (2025 survey) | AI will decrease headcount over the next year | 32% |
Expectations, not measured job changes. In McKinsey's 2026 survey only 14% of AI users reported an actual AI-related decline in headcount.
US workers aged 22 to 25 in the most AI-exposed jobs, Stanford payroll study
| Measure | Value | When |
|---|---|---|
| Employment gap against peers in less-exposed jobs (revised measure) | 19% | Data to Jun 2026 |
| Same gap at the first data vintage (revised measure) | 15% | Data to Jul 2025 |
| Relative decline in the original paper, with firm controls | 13% | Aug 2025 paper |
| Change in employment, two most exposed fifths of jobs, Nov 2022 to Jun 2026 | About -11% | Data to Jun 2026 |
| Change in employment, three least exposed fifths of jobs, Nov 2022 to Jun 2026 | About +10% | Data to Jun 2026 |
Brynjolfsson, Chandar and Chen, "Canaries in the Coal Mine?", revised 12 Aug 2026; original Aug 2025. ADP payroll data; descriptive, not causal. The authors find no comparable gap for experienced workers and say the change comes mainly through less hiring.
Other measures of AI and jobs, by source
| Source | Who is counted | Period | Finding |
|---|---|---|---|
| Census CES working paper CES-WP-26-25 | US employer firms, weighted by employment | Nov 2025-Jan 2026 | 32% of US employment is at firms that use AI, though only 18% of firms do |
| PwC Global AI Jobs Barometer | More than one billion job ads, 27 countries and territories | 2025 | 62% average wage premium in ads asking for AI skills, up from 57% a year earlier |
Why sources differ
Same question, different answers
| Number | When | Who measured | Why it differs |
|---|---|---|---|
| 95.7% | Nov 2025-Feb 2026 | US Census Bureau, BTOS AI supplement | Share of AI-using firms that saw no change in total employment over the past six months, as the firm reports it |
| 19% | Jun 2026 | Stanford Digital Economy Lab | Payroll-based employment gap for workers aged 22 to 25 in the most AI-exposed jobs, against peers in less-exposed jobs |
| 39% | May-Jun 2026 | McKinsey State of AI | Respondents, mostly at larger organizations, who expect AI to cut headcount next year; an expectation, not a count |
| 49% | May 2026 | St. Louis Fed district survey | Firms in one Fed district expecting no noticeable staffing effect in the next 12 months |
| 32% | Nov 2025-Jan 2026 | Census CES working paper | Share of US employment at firms that use AI; it measures exposure, not jobs lost |
Headlines about AI and jobs swing between "no effect" and "mass layoffs" because the sources measure different things.
The Census figure counts firms, and it asks about the past. Each firm counts once, whatever its size, and the firm itself judges whether AI changed its total employment. Almost all say it did not. A few large firms cutting staff would barely move this number, even if many workers were affected.
The Stanford study counts workers, using payroll records. It looks at one group, workers aged 22 to 25 in the jobs most exposed to AI, and compares them with similar young workers in less-exposed jobs. The gap there is real and growing, but it comes mostly from fewer hires, not from layoffs, and the authors say it is not proof that AI caused it.
McKinsey and the St. Louis Fed ask about the future. Expectations run ahead of what has happened: 39% of McKinsey respondents expect headcount cuts, but only 14% of AI users reported an actual AI-related decline.
For a fair summary, say that firms report almost no job losses so far, while entry-level hiring in exposed jobs has slowed.
What it means
How to use these numbers
AI is not yet replacing jobs at most US companies, at least by the companies' own account. Among firms that used AI, about 96 in 100 said it left their headcount unchanged, and slightly more said it added jobs than cut them. Most firms use AI to help staff do a task, not to take the task over.
The pressure is showing up at the entry door instead. Young workers in the jobs most exposed to AI, especially those built on written, textbook knowledge, are being hired less than peers in other jobs. If you are starting a career, the safer path is work that leans on experience, judgement and contact with people, and learning to use AI tools yourself: job ads that ask for AI skills pay a premium of about 62%.
For employers, the data suggest AI is mostly a productivity tool for now. Watch the sector rows: information firms report the most change in both directions.
For journalists, keep past effects and forecasts apart, and say whether a figure counts firms or workers.
Method and sources
How it was measured
Method
- US Census Bureau, BTOS AI supplement
- Business Trends and Outlook Survey of US employer firms; six biweekly panels pooled, 17 Nov 2025 to 8 Feb 2026; released 18 Jun 2026; ai_supplement_estimates.csv
- Definition
- Base for the employment question is firms that used AI in at least one business function in the last six months. Total employment is all paid staff, as the firm reports it.
- Census CES working paper
- CES-WP-26-25, Bonney, Breaux, Dinlersoz, Foster, Haltiwanger and Pande, April 2026; BTOS AI supplement responses weighted by employment
- Stanford Digital Economy Lab
- "Canaries in the Coal Mine?", Brynjolfsson, Chandar and Chen, revised 12 Aug 2026; ADP payroll records for millions of US workers through June 2026
- St. Louis Fed
- Quarterly business survey of the Eighth Federal Reserve District, special AI questions in May 2026; sample size not published
- McKinsey
- The State of AI 2026; 1,719 online respondents in 97 nations, 4 May to 8 Jun 2026
- PwC
- 2026 Global AI Jobs Barometer, press release of 15 Jun 2026; more than one billion job ads in 27 countries and territories; advertised pay, not earnings
- Checked at source
- 23 Oct 2026 (repo stats); 31 Oct 2026, the Census estimates file, the CES abstract, the Stanford post and the St. Louis Fed post
- Last changed
- 31 Oct 2026
- Data files
- CSV · JSON (stable links)
Sources
- Business Trends and Outlook Survey, US Census Bureau Official data
- Center for Economic Studies working papers, US Census Bureau Study
- Stanford Digital Economy Lab, Stanford University Study
- On the Economy blog, Federal Reserve Bank of St. Louis Survey
- The State of AI (McKinsey Global Survey), McKinsey & Company Survey
- PwC research (Global AI Jobs Barometer, AI agent survey), PwC Research
Cite this page
APA
MLA
HTML link
FAQ
More questions people ask
Which industries report the most job cuts from AI?
Among AI-using US firms, information (3.8%), transportation (3.2%) and retail (2.7%) were most likely to say AI decreased their employment, against 2.0% overall (Census, Nov 2025 to Feb 2026). Information firms were also the most likely to say AI increased it (4.5%).
Are young workers losing jobs to AI?
Fewer are being hired into exposed jobs. Stanford payroll data put employment of 22-to-25-year-olds in the most AI-exposed jobs 19% below peers in June 2026, up from 15% a year earlier, mainly through less hiring. The authors say the pattern is not proof that AI caused it.
Do firms replace staff tasks with AI?
Some do, but on a small scale. 10.1% of AI-using firms used AI to do a task an employee used to do, and 70.9% of those said it was a small number of tasks. 43.7% used AI to help staff with a task instead.
Do AI skills pay more?
In job ads, yes. PwC found an average wage premium of 62% in 2025 ads that ask for AI skills, up from 57% a year earlier. This is advertised pay, not what workers earn.
Every number on this page was checked at its original source (US Census Bureau, Stanford University, Federal Reserve Bank of St. Louis, McKinsey & Company, PwC). Free to cite with a link to this page.