Defining pattern
This week's clearest AI effects were the hardest to see in the usual numbers. Federal Reserve economists found that workers in AI-exposed fields are finding and switching jobs less often, a shift that never shows up in layoff counts. Half of reported AI time savings went to 8% of workers, and where AI use is nearly universal, two-thirds of workers say their employer expects more. McKinsey projects that Black and Latino workers' automated hours are more likely to be eliminated than enhanced. Closer to home, Rochester lost about 1,500 education and health jobs while the same sector grew statewide. Washington reset the federal contractor test and proposed tougher recruiting rules for green card sponsorship, New York got its first Workforce Pell program, and Reuters journalists authorized a strike partly over who earns the AI licensing revenue from their work. The common thread: the effects are real, quiet, and uneven.
1. AI Isn't Showing Up as Mass Layoffs. Fed Economists Say It's Showing Up as Fewer Job Moves.
On November 17, the San Francisco Fed's Technology & Growth Center posted new research by Federal Reserve Board economists Hie Joo Ahn and Nicholas A. Carollo. They combined Current Population Survey and JOLTS data with AI exposure measures from OpenAI and AI adoption measures from Lightcast to track how workers move in and out of jobs. Since large language models were introduced, workers who are both highly exposed to AI and at employers that have adopted it have seen "larger declines in job-finding and job-switching rates" than other groups. At the same time, switching between tasks inside their current jobs rose noticeably.
The authors also find that AI's earlier positive effects on nominal wage growth and hours worked weakened considerably after LLMs arrived, and that differences in job-finding prospects have widened markedly since 2023, mostly because of AI factors. They suggest the adjustment has happened mainly through task reorganization inside firms and weaker demand for the most exposed workers, not mass layoffs. They estimate the natural rate of unemployment has risen by "about 0.2 percentage point since the introduction of LLMs," with "considerable uncertainty."
- Workers who are both highly exposed and at AI-adopting employers saw the largest drops in job-finding and job-switching.
- Task-switching inside current jobs rose, a sign that work is being reorganized in place rather than cut.
- The estimated 0.2-point rise in the natural rate of unemployment comes with "considerable uncertainty."
This is a working paper, and the public summary doesn't report the underlying rate figures, so treat it as a strong signal rather than a final measurement. What it adds is a mechanism: displacement that arrives as stalled mobility instead of pink slips, which layoff counts and WARN notices can't see. Changing jobs is how many people get raises and move up, so a narrowing door matters even when layoffs stay low. For career centers, it suggests that people coming from AI-exposed fields may need more time and more help repositioning, even in a low-layoff market.
Impact: Emerging.
2. Closer to Home: Rochester Lost About 1,500 Education and Health Jobs While the Sector Grew Statewide
Rochester metro private-sector employment declined by 2,000, or 0.4%, to 445,000 in the year ending October 2026, according to preliminary New York State Department of Labor estimates released November 19. Education and health services lost about 1,500 jobs, the largest decline. Information lost about 400 and financial activities about 300. Mining, logging and construction gained about 500, the only major gain cited. Statewide, the picture ran the other way: private-sector jobs rose 11,300 (0.1%) over the year, and private education and health services grew 0.9%.
- Education and health services accounted for about 1,500 of Rochester's 2,000 net private-sector jobs lost.
- Statewide, private education and health services grew 0.9% over the same period.
- Mining, logging and construction added about 500 jobs locally, the only major gain.
Health care is often treated as the safe bet in workforce planning, so this number deserves attention. But the category combines private schools and colleges with hospitals, clinics and social assistance, and it's a net figure, so it can't show which occupations lost ground, how many people were laid off, or whether AI played any role. These are preliminary estimates and may be revised. For job seekers and career advisors, the practical step is to check openings by job title, employer and shift before steering toward, or away from, the sector.
Impact: Immediate.
3. Black and Latino Workers' Automated Hours Are More Likely to Be Eliminated Than Enhanced, McKinsey Projects
The McKinsey Institute for Economic Mobility published a report on November 20 mapping projected automation onto U.S. occupations and demographic groups. Of automation-adopted work hours, 40% for Black workers and 41% for Latino workers are projected to be displaced rather than augmented, compared with 37% for White workers and 36% for Asian workers. Latino workers account for 43% of hand packers and packagers, about 2.3 times their share of the overall workforce. Lower-wage workers are 7.6 times as likely as higher-wage workers to need to move into a new occupation, and workers without a bachelor's degree are 1.8 times as likely as those with one.
The report draws on McKinsey Global Institute's base case, in which automation could reduce labor demand by 36 million jobs over the next decade while growth across existing and new occupations adds the equivalent of roughly 41 million. The gap is in who can reach that growth: only about 10% of workers in the lowest wage quintile have a direct pathway into growing work, compared with almost 40% in the highest.
- Latino workers hold 43% of hand packer and packager jobs, 2.3 times their workforce share.
- Lower-wage workers are 7.6 times as likely to need a new occupation; workers without a bachelor's degree, 1.8 times.
- About 10% of the lowest-wage workers have a direct path into growing work, vs. almost 40% at the top.
These are projections from a firm that sells AI consulting, and the model assumes people stay in today's occupations, so they describe risk, not outcomes. What's useful is the distinction it draws: the gap is less about whether work gets automated and more about whether that work is eliminated or enhanced. For workforce organizations serving Spanish-speaking, hourly and no-degree job seekers, the takeaway is pathways into growing occupations, not AI literacy alone.
Impact: Long-term (projections through 2035).
4. Half of Reported AI Time Savings Go to 8% of Workers
Northeastern University released "AI and Time at Work" on November 18, based on its AI & Society Survey of 38,375 adults, including 20,275 workers, fielded June 21 to August 11, 2026. Thirty-three percent of workers use AI at least monthly and say it saves them at least an hour a week. "Half of all the hours saved go to 8% of workers," the report finds; put another way, 9% of workers save eight or more hours a week and account for more than half of all hours saved. The share saving at least an hour was 59% for graduate-degree holders, 40% for college graduates, 20% for high school graduates and 14% for people who didn't finish high school.
The gaps by industry and age were just as wide: 74% in information and technology, 26% in health care and social assistance and 20% in hospitality or service; 39% for ages 25 to 44, 28% for 45 to 64 and 19% for 65 and older. Hispanic and White workers each reported 33%. Where almost everyone at a workplace uses AI, 66% of workers say their employer has expected more work since AI became available; where only a few people do, 25% say so. A separate Gallup survey of 21,648 employees, released November 16, found 32% use AI frequently (49% in remote-capable jobs vs. 22% in on-site jobs), and 29% in AI-adopting organizations say expectations for daily output have gone up.
- Graduate-degree holders were nearly three times as likely as high school graduates to save an hour a week (59% vs. 20%).
- Where AI use is near-universal, 66% of workers say employers expect more; where it's rare, 25%.
- Hispanic and White workers reported the same rate of time savings, 33%.
Both surveys measure what workers report, not timed work, and their questions differ, so the numbers shouldn't be combined. Still, they point the same way: the gains are concentrated among degree holders and desk workers, and some of the time saved may be going back to the employer as higher expectations. We touched on expectations last week with Payscale; what's new here is how lopsided the distribution is. For job seekers, a concrete example of time saved on a real task is more convincing than a general claim of AI skill. For employers, frontline and on-site staff are where the gap is widest.
Impact: Immediate.
5. Labor Department Finalizes a New Federal Test for Independent Contractors
The U.S. Department of Labor announced a final rule on November 19 that rescinds its January 2024 independent contractor regulation and replaces it with an "economic reality" analysis substantially similar to the 2021 framework. Two core factors carry the most weight: the nature and degree of control over the work, and the worker's opportunity for profit or loss based on initiative or investment. Three other factors usually carry less weight: the skill required, the permanence of the working relationship, and whether the work is part of an integrated unit of production. When the two core factors point the same way, the others are unlikely to outweigh them.
The analysis applies under the Fair Labor Standards Act, the Family and Medical Leave Act and the Migrant and Seasonal Agricultural Worker Protection Act. DOL says it received more than 16,000 comments and included 14 fact-specific examples, among them delivery workers and truck drivers. The rule takes effect 60 days after publication in the Federal Register. DOL says it doesn't change the analysis under other federal laws, including tax law, and doesn't preempt state or local laws.
- Control over the work and opportunity for profit or loss are now the two core factors.
- The test covers minimum wage and overtime, FMLA leave, and protections for migrant and seasonal farmworkers.
- State and local standards still apply on their own terms, and tax classification is a separate question.
This is the latest swing in a rule that has changed with each administration, and legal challenges are possible. For workers, the label on a contract doesn't settle status; who sets the schedule, assigns the work and pays the business expenses matters more. For career advisors, a contractor offer deserves a closer look at the day-to-day arrangement and the state standard, since a worker can be treated differently under federal and state law. How many workers will actually change classification is unknown.
Impact: Emerging (effective 60 days after Federal Register publication).
6. New York's First Workforce Pell Program Goes to SUNY Ulster
The U.S. Department of Education approved SUNY Ulster's Certified Clinical Medical Assistant program as New York's first Workforce Pell program, as reported November 19, and SUNY announced the designation November 20. Workforce Pell, which launched July 1, lets eligible students use Pell Grants for short programs that run at least 8 but fewer than 15 weeks (150 to 599 clock hours). According to The College Investor, New York is the eighth state with an approved program; the first national approval came August 4. SUNY has said additional programs are under consideration.
Eligibility works like a traditional Pell Grant: students file the FAFSA and must show financial need, and people whose highest degree is a bachelor's can still qualify. Each Workforce Pell term counts toward the 12-semester lifetime Pell limit, and approved programs must maintain 70% completion and 70% job placement rates. The College Investor estimates the maximum award at about $4,310 for a 14-week program. SUNY Ulster hasn't posted the program's price.
- Workforce Pell covers programs of 8 to under 15 weeks, or 150 to 599 clock hours.
- Each Workforce Pell term counts against the 12-semester lifetime Pell limit.
- Approved programs must keep 70% completion and 70% job placement rates.
One approved program statewide, nearly five months after launch, and none in the Finger Lakes that we could find. Read alongside story 2, it's a reminder that a credential's value depends on local demand: an approved medical assistant program in Ulster County says nothing about openings in Rochester. For job seekers, using Pell on a short program draws down eligibility you might want later for a degree, so ask financial aid about remaining cost and longer-term plans first. For WIOA providers, it's a possible co-funding source once local programs are approved.
Impact: Immediate.
7. New Jersey Apprentices in Nontraditional Fields Saw Big Earnings Gains. Completion Is the Catch.
Rutgers University's Heldrich Center released an evaluation on November 18 of New Jersey's Growing Apprenticeship in Nontraditional Sectors (GAINS) program, using state administrative records, federal apprenticeship data and unemployment insurance wage records from 2019 through 2025. Among completers, employment rose from 64.7% to 80.4%, and median annual earnings rose from $19,495 to $92,595, an increase of 375%. Women completers' median earnings rose 411.4%, to $92,137. Women made up 79.6% of new GAINS apprentices in program year 2024, compared with about 15% in New Jersey's other registered apprenticeship programs.
Completion is the limiting factor. In program year 2025, 47% of women and about 39% of men completed, according to Rutgers, and MyCentralJersey reported an overall completion rate of 44.8% among the apprentices examined. The researchers called the results promising and recommended further study of longer-term employment stability, career progression and wage growth.
- Completers' median annual earnings rose from $19,495 to $92,595.
- Women were 79.6% of new GAINS apprentices, vs. about 15% in other New Jersey apprenticeships.
- Fewer than half completed: 47% of women and about 39% of men in program year 2025.
The before-and-after numbers describe completers only, and the published summary doesn't describe a comparison group, so this isn't proof the program caused the gains. What it does show is that apprenticeship can reach people traditional programs rarely do, especially women. For workforce organizations building apprenticeships, the supports that keep people enrolled may matter as much as recruitment.
Impact: Immediate.
8. Reuters Journalists Authorize a Strike Over Pay, AI Oversight and AI Licensing Revenue
Members of the Reuters Guild, representing nearly 300 media workers, concluded a strike authorization vote on November 16. The NewsGuild reported nearly 96% turnout with 94% voting yes; People's World reported the figures the other way around (94% turnout, 96% in favor). Members have worked under a contract that expired in December 2025, after more than a year of bargaining. According to the union, management has rejected human oversight of AI-generated content, refused compensation for workers whose jobs may be eliminated by AI, and declined to share AI licensing revenue derived from journalists' work. The Guild also says management's pay proposal would amount to an annual pay cut with inflation near 4%.
- Disputed terms include human oversight of AI-generated content and compensation if AI eliminates jobs.
- The Guild wants journalists to share in AI licensing revenue earned from their work.
- The contract expired in December 2025, after more than a year of bargaining.
The vote authorizes a strike; it isn't a walkout, and management's position here comes through the union's account. What stands out is the licensing question. The dispute isn't only about whether AI replaces jobs, but about who benefits when existing work earns new AI revenue. That question is likely to reach other fields whose work feeds AI systems.
Impact: Immediate.
9. Young Workers in AI-Exposed Jobs Keep Losing Ground, and the Youngest Are Slipping Even in Low-Exposure Jobs (Recurring Theme)
The ADP Research and Stanford Digital Economy Lab Canaries Dashboard, updated November 18 with private payroll data, found that employment among workers ages 22 to 25 fell 4.2% from October 2025 to October 2026 in highly AI-exposed occupations and 1.3% in the least-exposed. For ages 26 to 30, employment fell 3.4% in highly exposed occupations and rose 0.7% in the least-exposed. Across all ages, highly exposed occupations shrank 0.8% and the least-exposed grew 0.3%. In a November 15 update, Stanford HAI said declines have also deepened in occupations where AI supports rather than replaces workers, including marketing and IT managers, and noted these patterns don't establish causality.
A separate Indeed and Glassdoor outlook for 2027, published November 18, found postings rising in AI-exposed fields such as software development, banking and engineering. Senior-level postings were up 19% since January 2022, while entry- and mid-level postings each fell about 4%, according to HR Dive's summary of the report.
- Workers ages 22 to 25 lost ground even in the least-exposed occupations (down 1.3%).
- Stanford HAI says declines deepened in roles where AI supports workers, such as marketing and IT managers.
- Indeed: senior postings up 19% since January 2022; entry- and mid-level postings down about 4% each.
We've covered the entry-level squeeze often, most recently with Fed Governor Barr last week. What's new is a fresh month of payroll data showing the youngest workers losing ground even in low-exposure jobs, which suggests AI is widening a gap rather than causing all of it. For young job seekers, a "safer" field still needs a check on whether employers there hire beginners. For employers, entry-level hiring is worth tracking separately from total headcount.
Impact: Immediate.
10. Labor Department Proposes Tougher Recruiting Rules Before Employers Can Sponsor Green Cards
DOL published a proposed rule in the Federal Register on November 17 that would overhaul Permanent Labor Certification (PERM), the labor market test most employers must pass before sponsoring a worker for a green card. Job advertisements, including the State Workforce Agency job order, would run for at least 90 calendar days, and print-based options would be replaced with electronic recruitment such as employer websites, job search sites and social media. DOL would also stop administering Schedule A, which exempts certain occupations from the labor market test, except as permitted by Congress.
Employers that laid off U.S. workers in the same or a similar occupation at the same location within the prior two years would be barred from requesting certification, and employers with later layoffs would have to consider laid-off U.S. workers and document those offers. Third-party staffing firms could not place sponsored workers with employers conducting layoffs. Comments are due January 19, 2027.
- Job ads, including the state job order, would run at least 90 calendar days.
- Employers with layoffs in the same occupation and location in the past two years couldn't seek certification.
- Ending Schedule A would remove a long-standing exemption from the labor market test for certain occupations.
This is a proposal, and DOL's announcement doesn't quantify effects on domestic hiring, employer costs or processing times. Schedule A has historically covered occupations such as registered nurses and physical therapists, so health care employers that recruit internationally would likely feel this first. Last week IBM ended sponsorship for most consulting roles; together, these point to a narrower path for sponsored workers. For American Job Centers, longer state job orders could keep these positions visible to local job seekers for longer.
Impact: Emerging (comments due Jan. 19, 2027).
11. The Numbers: Continuing Claims Jump 70,000, Mostly in Michigan
Initial unemployment claims rose 1,000 to a seasonally adjusted 203,000 for the week ending November 14, DOL reported November 19; the prior week was revised up from 200,000 to 202,000. Continuing claims rose 70,000 to 1,763,000 for the week ending November 7, after the prior week was revised down to 1,693,000. Reuters reported that was the highest level since the end of August. Both measures remain below year-earlier levels of 222,000 and 1,949,000. Michigan accounted for most of the increase: its unadjusted continuing claims rose 39,343, to 69,421, a week after falling 14,771. Michigan's initial claims also rose 6,095.
- Initial claims: 203,000; continuing claims: 1,763,000.
- Michigan's continuing claims more than doubled in one week, to 69,421.
- Both measures remain below where they were a year ago.
A one-week jump concentrated in one state, right after a sharp drop there, may reflect something specific to Michigan; we couldn't confirm the cause. Reuters also noted that the median duration of unemployment reached 11.9 weeks in October, the longest since November 2021 (earlier October data). Read alongside story 1, these numbers fit a market where layoffs stay low but the people already out of work take longer to land. These are advance figures, subject to revision.
Impact: Immediate.
Bottom line
For workers and job seekers, this week's evidence says AI's effects are more likely to show up as fewer job moves, uneven gains and rising expectations than as layoffs. If you work in an AI-exposed field, plan for a longer search when you decide to move. If you're weighing training, check local demand by job title, since Rochester's education and health numbers ran opposite to the state's.
For employers and workforce organizations, the usual measures, layoffs and weekly claims, miss much of what's happening. It's worth tracking hiring and mobility by role, who actually gets AI's time savings, and whether training reaches frontline staff. The new contractor rule and the proposed PERM changes are also worth reviewing with anyone who advises on contract offers or sponsored hiring; PERM comments are due January 19.
