When a Company Cuts Jobs, Blaming AI Is Free. This Strand Checks the Invoice.
Estimated reading time: 3 minutes
.jpg){kind=link}
In short. When an employer attributes job cuts to AI, the ledger records that attribution as a claim and this strand tests it against the available evidence: filings, investor statements, subsequent hiring, and whether the roles cut match the capability claimed. The same scepticism applies to governments crediting AI investment for job gains. The aggregate lives on the scoreboard; the case-by-case examination lives here.
There is a market incentive to blame the machines, and it runs in both directions. A company presenting cuts as an AI transformation is telling investors a story about efficiency and the future. A government presenting a subsidy as AI job creation is telling voters the same story in reverse. Neither claim is evidence, and the official statistics that would settle the question are aggregates that do not carry attribution at all [1]. The attribution gap is where this strand works.
The test applied to every entry
What was said, to whom, and what would make it true. For each AI-attributed cut the strand records: the exact claim and its audience (press release, investor call, filing); whether the cut roles plausibly match the claimed capability; what the company filed before and after [2]; and what it advertised for in the months following, because a company that cuts 300 roles citing AI and then rehires 200 humans for adjacent titles has told the ledger something its press release did not.
The verdict vocabulary is deliberately narrow. An entry ends as: attribution consistent with evidence, attribution unsupported, attribution contradicted, or evidence insufficient. Nothing here calls a named company dishonest; the record is laid out and the reader can do their own arithmetic.
Why this matters beyond the companies
The question of whether AI is actually reducing UK employment, and how fast, will be answered by exactly this kind of accounting or it will be answered by press release. The aggregate AI-attributed figure on the scoreboard only means something because each row under it survived the test above; the method is described in full in What the Ledger Counts.
Sources
- Office for National Statistics, employment and labour market statistics (aggregate redundancy and vacancy series; no attribution dimension). https://www.ons.gov.uk/employmentandlabourmarket
- The Gazette, insolvency notices (the statutory record against which company narratives are checked). https://www.thegazette.co.uk/insolvency