The Talent Risk Index

What the Index rates

Large employers (initially those with public workforce action in the last 90 days; expanding to FTSE 100 and S&P 500) on two separate questions:

State — how much risk is there to jobs at this company right now? Workforce action of any kind. Where a company has cited AI as a reason, we show that as a flag with the source; we do not decide whether AI "really" caused it.

Conduct — when a company acts, how decently is it treating the people affected?

These are the Editor's opinions, formed from the evidence listed on each company's page. On every page, the facts (with sources) sit above a line; the Editor's assessment sits below it and introduces no new facts. They are not predictions of any individual's employment, and not financial advice.

This is the Editor's opinion based on the sources listed. — appears on every company page.

State

| State | Definition | |---|---| | Calm | No material AI-linked workforce action in the last two quarters and no precursor signals above threshold. | | Watch | Precursors present: one or more of — AI deployed into an operational function with stated efficiency goals; hiring freeze or materially slowed hiring; restructuring provision raised; executive language linking AI to cost or headcount; a peer in the same sector and function moving to Warning; verified-insider signals above threshold. | | Warning | Action announced or evidently imminent: cuts, voluntary exit programme or statutory consultation announced; restructuring charge booked against workforce; executives linking AI to headcount reduction on the record; or precursor signals at a level that in our record has preceded action. |

Movement between states is shown with the date and the evidence that moved it.

Conduct (assessed only in Watch or Warning)

| Grade | Definition | |---|---| | A | Voluntary before compulsory; terms above statutory and above sector norm; funded retraining or redeployment with published numbers; consultation beyond the legal minimum; AI's role stated plainly. | | B | Voluntary programme with substantive support (e.g. continued healthcare, extended vesting, outplacement); clear, timely communication. | | C | Statutory process; standard terms; AI's role acknowledged but not explained. | | D | Compulsory reductions framed as efficiency; functions targeted without redeployment; restructuring costs materially revised upward mid-year; criticism from recognised unions unanswered; selection criteria unexplained. | | E | Opaque or automated selection without meaningful human involvement; terms below sector norm; AI-use assessments tied to continued employment; retaliation or silencing. |

Conduct can improve. We re-grade when the evidence changes.

Evidence we use

In order of weight: company filings and results statements; company announcements and internal memos that are on the public record; regulatory notices (US WARN notices; SEC 8-K Item 2.05 and 10-Q restructuring disclosures; UK RNS announcements); executive statements reported by named outlets, quoted verbatim; recognised trade-union statements; reporting by named outlets. Every item is dated and linked on the company page. A Warning requires the company's own words, a filing or a formal notice, or two independent named outlets; reports that a company is "weighing" or "considering" action produce at most a Watch.

UK collective-redundancy (HR1) statistics are published only as national aggregates and are used as context, never as evidence about a named company.

Right of reply. Before any company is newly rated Watch or Warning, or given a Conduct grade of D or E, we send the facts we intend to publish to its press office with a 24-hour window. The response, or "no response by deadline", is published on the page.

Verified-insider signals (members verified by employer email domain, reporting anonymously in categories) are not yet shown. When they are, they will be shown to everyone at the same time, at least seven days after the reporting week, only as counts by category, only where at least 25 verified employees of that company have reported (and 10 for any breakdown by function), and never for employers with fewer than about 2,000 UK staff. Certain categories are never shown before the company itself has announced. They corroborate; they do not on their own move a State.

Package data (what leavers were offered) is not yet shown. When it is, it will appear only where at least 10 verified submissions exist for a company, bucketed by level and amount, never as individual records. We only ask people to share what they are free to share.

What we don't do

We don't rate individuals. We don't publish anything an insider wrote in their own words. We don't show information below the thresholds above. We don't take money from the companies we rate.

Corrections

Errors of fact are corrected within 24 hours of being established, marked on the page with the date and what changed. Companies may send a response to editor@talentrisk.com; responses are published in full.

Versions

v0.1 — 31 Oct 2026 — initial.