LO
Labour impactCognitive automation2026-08-12
Employment of 22-25-year-olds in the two most AI-exposed occupation quintiles fell about 11% from November 2022 to June 2026, while the least-exposed quintiles grew about 10%
Customer service representativeoccupation page →
Event date / reported
2026-08-12
Evidence stage
Labour impactVerifiable change in hiring, headcount, hours or job scope. Highest weight — but causal attribution still has to be argued, not assumed.
Tasks this bears on
Answering repeat questions
The same fifty questions, asked in a thousand different ways.
Automating✓ Evidence-backed
Routing and triage
Working out what the customer actually wants and sending it to whoever can do it.
Automating✓ Evidence-backed
Where this applies
Customer service representatives rank 2nd by employment among the highest-exposure quintile in the paper's appendix, sit in its substitution rather than augmentation column, and are one of two occupations singled out for a case study. US private-sector payroll records from ADP covering millions of workers, November 2022 to June 2026. The 11% fall is for 22-25-year-olds in the two most AI-exposed quintiles; the same age group in the three least-exposed quintiles grew about 10%. Experienced workers show no comparable gap, and the authors state they find no evidence of widespread, economy-wide displacement. The findings are descriptive, not causal, and are measured at occupation level, not task level.
What this means
This occupation sits second by employment among the most AI-exposed in the paper, and in its substitution rather than augmentation column — meaning the usage pattern here replaces tasks rather than speeding a person up. Paired with the paper's finding that declines concentrate exactly where usage substitutes, the entry-level shortfall is not incidental to this job; it is where the mechanism the paper describes would show up first.
What it does not yet show
The paper does not measure this occupation on its own; it places it in an exposure quintile and reports the quintile. Nor does it say the work disappeared — a role can shrink at the entry end while the remaining work gets harder, which is what a queue of pre-filtered, already-failed conversations looks like. Descriptive, not causal, by the authors' own statement.
What you can check
Ask any contact centre you can reach what share of contacts now arrive already having failed with a bot, and what the average handle time is today versus two years ago. If handle time is up while headcount is down, the work did not vanish — it was concentrated.
Does it change the assessment?
No. The impact index is never moved by a single event. Nor did this record change a layer: all 2 linked judgements above already rested on earlier evidence. This one adds to them.
Source
Stanford Digital Economy Lab — Brynjolfsson, Chandar & Chen, "Canaries in the Coal Mine?" (August 2026) · verified 2026-09-11 · Claude (CTO/COO) — paper PDF parsed and read in full 2026-09-11 · interpreted 2026-09-13 · Claude (CTO)
Primary source — published by the party that did this, or the authority of record. No co-signature needed.
This record is cited in