Customer service employment 50% below May 2023
Will the latest BLS OEWS release show customer service representative employment (SOC 43-4051) 50% or more below the May 2023 level?
Chance by Q4 2031
3%
Cumulative probability
- Latest forecast
| Year | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| 2026 | <1% | |||
| 2027 | <1% | <1% | <1% | <1% |
| 2028 | <1% | <1% | <1% | <1% |
| 2029 | <1% | <1% | <1% | <1% |
| 2030 | <1% | <1% | <1% | <1% |
| 2031 | 1% | 3% | 3% | 3% |
Model reasoning
The government's count of customer service jobs fell about 5% a year to 2.6 million in 2025. Reaching the threshold needs roughly another 45% drop in the yearly estimates published in 2028–2031. The survey pools three years of data, so it lags reality, which makes such a fall even harder to show in time. That would be unprecedented and far beyond even aggressive industry forecasts. The development that would most change this is evidence of sudden collapse, such as call-center employment falling 20% or more a year.
Where things stand
I downloaded the BLS national OEWS files for 2010–2025 (oesm10nat.zip … oesm25nat.zip, linked from https://www.bls.gov/oes/tables.htm). They give these national figures for Customer Service Representatives (43-4051):
- May 2023: 2,858,710 (the baseline)
- May 2024: 2,725,930 (down 4.6%)
- May 2025: 2,595,750 (down 4.8%). This is the latest release, published May 15, 2026 (USDL-26-0725). It came out late because of the shutdown.
The threshold is 1,429,355 or below. From the May 2025 level, that means a further fall of 44.9%.
Which releases count. Under the observation-period rule, the May reference period has to end after 2026-09-16. So the May 2026 estimates, due in spring 2027, do not count. The qualifying releases are:
- May 2027 estimates, about April 2028
- May 2028 estimates, about spring 2029
- May 2029 estimates, about spring 2030
- May 2030 estimates, about spring 2031
Releases normally come in late March to early April, so each one most likely lands in the second quarter. From now until roughly April 2028, YES is close to impossible.
Annual decline the published series would need, starting from May 2025:
| Hit the threshold by | Required decline per year |
|---|---|
| May 2027 | 25.8% |
| May 2028 | 18.0% |
| May 2029 | 13.9% |
| May 2030 | 11.2% |
For comparison, the series has actually been falling about 4.7% a year.
The estimate lags real employment
OEWS combines six survey panels collected over three years. It uses a model-based method that leans toward recent donors, and it scales to current employment counts from unemployment-insurance records (OEWS technical notes; BLS Monthly Labor Review 2019 article on the model-based method). Most customer service jobs are spread across industries: call-center firms are only about 8% of the total, and insurance, banking, retail, grocery and hospitals account for much of the rest. A drop inside those industries therefore shows up in the published count with a lag of roughly 1 to 1.5 years.
I built a simple lag model with half-year panels and recency weights. To reach the threshold in the May 2030 estimates, actual employment would have to fall about 20% or more per year, sustained from early 2027. Hitting it by the May 2029 estimates needs about 35% or more per year. By May 2028, even a 50% annual decline starting in mid-2026 falls short.
Base rate and current evidence
- Base rate. In the national OEWS data for 2010–2025, I checked every 5-year window for detailed occupations with at least 500,000 jobs. None of the 605 windows shows a fall of 45% or more. The largest falls are about 41% (executive secretaries, partly a coding change) and 30% over two years (waiters during COVID).
- Call-center industry. BLS payroll series CEU6056142001 (telephone call centers) went from 449.2k in May 2022 to 398.8k, 351.7k, 325.4k, and 301.5k in May 2026, and was 303.3k in August 2026. That is a 7%-a-year decline, slower than the 11–12% seen in 2023–24.
- Household survey. The Current Population Survey counted 2.60M customer service representatives in 2025, against 2.714M in 2022.
- Global picture. Revelio Labs (Oct 6, 2026) puts global call-center headcount 5.3% below its December 2023 peak, with the decline driven by less hiring rather than layoffs.
- Forecasts. Forrester's July 2026 quantitative forecast, built on this same BLS series, expects a 13.5% fall to 2.24M by 2031 (No Jitter, Aug 3, 2026). Its headline "half of jobs by 2030" refers to work handled by AI; even taken literally, the lag means the published count would stay above the threshold. BLS projections for 2025–35 show a 5.3% decline. One self-described "bold" public forecaster treats a 30% fall by May 2032 as contrarian.
- Classification changes. The 2028 occupational classification revision is behind schedule: its second Federal Register notice was still "TBD" on the BLS page last modified Sept 30, 2026. Under the question's rules on successor occupations, a reclassification is very unlikely to create a YES.
Paths to YES and NO, and where I land
YES needs economy-wide, AI-driven displacement of the most exposed occupation, faster and deeper than any published forecast. Real employment would have to roughly halve between 2027 and 2030 despite the many in-person roles (retail, banks, hospitals, government). The decline would also have to start almost immediately, while data through August 2026 show only a moderate trend.
Strongest case against my low number. AI capability is improving fast. Turnover in these jobs is 30–100% a year, so a hiring freeze could shrink headcount quickly. Companies such as Microsoft, Uber and Verizon are already cutting, and prediction markets give meaningful odds to AI-driven economic shocks. I give this tail weight, but the reporting lag keeps the window narrow.
Probabilities. I put the chance that each estimate has crossed the threshold at about:
- May 2027: 0.05%
- May 2028: 0.15%
- May 2029: 0.7%
- May 2030: 2.7%
I spread each release's chance over the year it is published, putting about 20% of it at the March quarter-end and 95% by June. I kept a 0.1% floor for resolution oddities.
Resolution criteria
Question. By the last day of each quarter, has the latest BLS OEWS release shown national customer service representative employment (SOC 43-4051) at least 50% below its May 2023 level?
Definitions.
- Employment: the national, all-industry, all-ownership employment estimate for 43-4051, Customer Service Representatives, in the Bureau of Labor Statistics’ annual Occupational Employment and Wage Statistics (OEWS), National Occupational Employment and Wage Estimates. Use the employment count, labeled
TOT_EMPin the national downloadable table, not wages, employment share, job openings, or a projection. This measures jobs within OEWS coverage, not unique workers or full-time equivalents. - May 2023 level: fixed at 2,858,710, as published in the May 2023 occupation table. Later revisions to the baseline do not change the threshold.
- 50% or more below: a published employment estimate ≤1,429,355. Compare the published count directly, without rounding the percentage decline or using confidence intervals.
- Latest release: the regular national OEWS estimates for the most recent reference period published by BLS at the time of the qualifying publication. An official correction to that latest release counts on its correction-publication date; revisions only to superseded reference periods do not count. Research or experimental estimates do not count.
- Occupational successors: a renamed or renumbered occupation counts if BLS identifies it as covering the same occupation. If BLS splits 43-4051 into mutually exclusive categories that together exactly exhaust its former scope, sum those categories. A merger with other occupations does not count unless BLS separately publishes the employment of the former 43-4051 scope. Missing or suppressed counts are not zero.
Resolves YES if a qualifying BLS publication reports the latest-release employment count at or below 1,429,355. Apply the Resolution window and Observation period conventions: the estimate’s May reference period must end after the opening date, regardless of the older survey panels used to construct it. No finding that AI caused the decline is required. Ordinary changes in BLS estimation methods do not disqualify a published estimate.
Once a qualifying latest release meets the threshold, the question remains YES for later horizons under the Resolution window convention, even if employment subsequently rises or BLS later corrects the count upward.
Resolves NO at a horizon if no qualifying publication has met the threshold by that horizon. State, metropolitan-area, industry-only, private-sector-only, Census/CPS, and Employment Projections figures do not count. Discontinuation of the series or inability to obtain a count for the specified occupational scope does not itself establish YES.
Resolving source. BLS only: the annual national OEWS tables and downloadable national files linked from OEWS Tables, the corresponding BLS occupational data pages, and official BLS corrections. If formats disagree, use an explicit BLS correction first, then the national downloadable table. This is a narrower resolving-source rule than Public confirmation (A2). Use the qualifying release or correction’s publication date under Resolution date (A3), not its reference month.