ADP released its May 2026 National Employment Report at 8:15 ET on Wednesday, and the headline came in better than the consensus had any right to expect. Private payrolls rose by 122,000, above the +99K to +117K range of analyst estimates and well above April’s reading, which was revised down to +105,000 from an initial +109,000. ADP chief economist Nela Richardson called it “more broad-based in May than we’ve seen in the last few years.”
Read the sector breakdown and the broad-based framing is true, but it has a hole shaped like the information sector. Education and health services led with +57,000. Trade, transportation and utilities added +36,000. Professional and business services added +11,000. Leisure and hospitality and construction each added +8,000. Financial activities added +7,000. Other services added +4,000. Manufacturing added +3,000. Eight of ten sectors hired. Two cut. Natural resources and mining lost -3,000. The other one — information services — lost -9,000, and that is the line worth dwelling on.
The sector that won’t stop bleeding
Information is the BLS taxonomy for telecom, publishing, broadcasting, data processing, software, motion pictures — i.e. the white-collar tier with the highest AI exposure in the entire industry classification system. ADP’s May print is the third consecutive monthly loss for the sector and the largest of 2026 so far. Cumulative information-sector losses in ADP’s data now sit at roughly -24,000 jobs year-to-date, in a year where total private payrolls have averaged barely above +100,000 a month.
Put differently: the entire economy added 122,000 jobs last month while the AI-exposed sector subtracted 9,000. The math means information services accounted for net -7% of the economy’s job creation in May — a sector that employs roughly 1.9% of the workforce dragging on a headline that is otherwise broad-based. Richardson’s “broad-based” line is correct everywhere except in the place where AI productivity claims are most concentrated.
What’s hiring instead
Education and health services continue to absorb almost half of all net job creation in the US — they led April’s print at +61,000 and led May at +57,000. These are the sectors where AI rollouts are slow, regulated, and labour-intensive at the floor level. Trade, transportation and utilities — warehouse and last-mile — rebounded after two soft months. Professional and business services finally turned positive (+11,000) after April’s -8,000 print. Manufacturing added a modest +3,000 even as goods-producing employment broadly trended sideways.
The size breakdown is even more revealing of what is and isn’t working. Establishments with fewer than 50 employees added +67,000 jobs — more than half the total. Large firms (500+) added +40,000. Medium firms — the 50-to-499 band — added just +17,000. Richardson described this as “softness in the middle” last month and the same phrase fits now. Big enough to deploy AI tooling at scale, small enough to be nimble — but the middle, where a Salesforce or a Cisco or a Meta lives, is the band that’s cutting headcount while hiring AI specialists on net-negative replacement ratios.
Pay growth says workers aren’t getting more leverage either
Pay for job-stayers held at +4.4% year-over-year, unchanged from April. Pay for job-changers ticked down to +6.5% from +6.6%. The spread between stayers and switchers — 2.1 points in May — is the narrowest it has been since 2023. The “switcher premium” is the price employers will pay to pry someone out of an existing seat. When it compresses, it means employers don’t need to pay up to fill openings, which is consistent with yesterday’s JOLTS print: postings are inflating without translating into offers, and offers are not having to compete for talent.
What this print means together with the JOLTS one
Tuesday: 7.62M April openings, 5.12M April hires, the lowest hires-rate non-recession reading since 2014. Wednesday: 122K May payrolls, sectorally broad except for the AI-exposed one, with pay growth flat and the switcher premium narrowing. The pair is consistent with the structural reading of the labour market the Indeed Hiring Lab and Stanford AI Index have been pushing: postings are budgeted, hires are not extended, and the sector with the highest AI productivity-pitch is the one shedding jobs net every month.
Challenger, Gray & Christmas has been calling AI the number-one cited reason for job cuts for two months running, and their May print (due in the next week) will fold in the same Meta, Walmart, Salesforce, Intuit, and Oracle announcements that BLS hasn’t formally caught yet. Friday’s BLS nonfarm payrolls report — the official number that often diverges from ADP — is expected at +85,000, well below ADP’s +122K and below April’s +115K. If BLS prints a sub-100K number with the information sector again negative, the “information sector is the AI sector” reading stops being a cute pattern and starts being a stylised fact.
The headline is fine. The footnote is the story.
122K is fine. 122K is, in fact, above consensus. The CNBC ticker chyron and the Fox Business chart both said “broad-based hiring continues” and they were not lying. But ADP itself flagged information services as “a possible impact from artificial intelligence growth” in its release-day notes. ADP is the largest private payroll processor in the country, and when it puts AI as the reason on the line item that is bleeding three months running, that is the same company that built the report telling you to look at the footnote.
The footnote is the story. The headline is broad-based hiring everywhere AI hasn’t shown up yet.
