The Office for National Statistics (ONS), the United Kingdom's statistical agency, published a labour market transformation update on 11 August confirming that its Transformed Labour Force Survey (TLFS) will not become the headline measure of British employment before November 2027, with the readiness decision itself pushed back to July 2027. 1 The TLFS is the online-first replacement for a survey whose response rates collapsed after the pandemic, and it has been the agency's route back to reliable labour data for three years.
Two admissions inside the document carry more weight than the delay. The ONS records that industrial and occupational coding under the new survey remain problematic, and it discloses a May 2026 collection round in which under-resourced telephone interviewing left a noticeable effect on average-hours estimates. Occupational coding is the machinery that assigns a person to a job category, and every study elsewhere in this briefing depends on exactly that assignment before it can say anything about AI at all.
The agency reported UK vacancies rising to 712,000 last month, the first quarterly increase in the series , and it reported unemployment at 4.9% in June with no AI attribution layer attached . Neither number can be decomposed by exposure, and neither could be if a minister demanded it tomorrow. Korea reached its finding through a pension register, Mexico through a national survey matched to an index, and Britain currently has neither route open to it.
Each year without occupational coding is a year of AI-era labour history that cannot be reconstructed later, because the underlying microdata will have been collected under a classification nobody trusts. The evidence that would settle the British version of this argument is not being contested; it is not being gathered.
