Low displacement
95.4 by Dec 2029
Relative recruitment recovers 10%. Employers continue opening human roles alongside AI.
Sep 2027 checkpoint 89.3
Definitions frozen September 2026Download frozen definitions ↓
United States / Daily · updated weekly
Is white-collar recruitment falling behind?
Demand for six white-collar job families, relative to the overall jobs market. A falling line means these careers are losing ground even after allowing for a broad hiring slowdown.
Source: Indeed Hiring Lab ↗ · Latest observation 28 Aug 2026
Latest reading86.728 Aug 2026
Relative recruitment · 1 Feb 2020 = 100
ObservedLowMediumHigh
Solid line: published history. Dashed lines: illustrative paths, not estimated probabilities.Horizon: December 2029
Reading the line: Lower values are consistent with greater displacement pressure.
Download history ↓White-collar basket88.6Source index · Feb 2020 = 100
All US postings102.1Source index · Feb 2020 = 100
Tracking against three potential scenarios fixed in September 2026.
High / medium / low refers to displacementMedium displacement
78.0 by Dec 2029
Relative recruitment falls another 10%. Entry and career moves become harder even if the wider market holds up.
Sep 2027 checkpoint 84.1
High displacement
60.7 by Dec 2029
Relative recruitment falls another 30%. A large, persistent loss of opportunities would support the hiring-squeeze scenario.
Sep 2027 checkpoint 78.8
What would change the evidence?
A sustained relative decline supports a recruitment squeeze. A recovery while adoption grows would weaken it. Require corroboration from actual hiring and employment before interpreting this as displacement.
What this cannot tell us
Postings are intentions, not completed hires or jobs lost. The benchmark includes exposed work. Offshoring, remote work and the post-pandemic correction can create the same pattern.
Frozen definition & source notes Version 1.0 · 6 September 2026
Equal-weight mean of total-posting indices for Software Development, Accounting, Marketing, Administrative Assistance, Customer Service and Banking & Finance, divided by the all-US total-posting index, × 100. Daily seasonally adjusted seven-day trailing indices; 1 February 2020 = 100. All daily observations are retained. The six families are a frozen watchlist, not a census of white-collar work. An equal-weight index is not the share of all advertisements.
Worked Excel example ↓ · Source inputs for Excel ↓
Indeed Hiring Lab data: CC BY 4.0. This dashboard transforms the published indices; Indeed does not endorse these scenarios.
Historical data use the provider’s current revised methodology, including its November 2024 seasonal-adjustment change.
Scenario rule: a straight-line path from the frozen 28 Aug 2026 observation (86.7) to December 2029. Endpoint changes are +10%, -10%, -30% for low, medium and high displacement. These are judgement-based, falsifiable scenario assumptions, not fitted forecasts or a no-AI counterfactual.
The metric definition, source snapshot and forecast anchors are saved together. Future releases should be appended as observations without moving these original lines. Source revisions should be reported separately. Review sustained changes, not a single monthly surprise.
September 2027 review: Compare the first published September 2027 value (Q3 for Canadian metrics) with the frozen path at its observation date. Report signed errors against all three paths, and the nearest path by absolute error. Require persistence for at least four successive quarters before describing structural displacement. Report revised-data results separately. No scenario fit establishes AI causation.
Source downloads: indeed-us-sectors.csv ↗indeed-us-all.csv ↗indeed-LICENSE.txt ↗
Displayed history starts in 2015 where available; original source files retain the longer published record. Missing periods are not interpolated. All plotted history is downloaded data or a disclosed calculation from it.
Blog
About me
Peter Donaghy is a Belfast based accountant and data analyst.
This is my own independent work. The views expressed are my own and are not endorsed by, or representative of, any current or former employer.
AI disclosure: GPT-6 Astra was used to define the metrics, source data and create the dashboard. I do not use AI for blog posts.