We asked the government which measure tells it what AI is doing to British jobs. The answer names a voluntary agreement in which firms choose what to share

On 24 September we asked the government three questions. Which measure does it rely on to track AI's effect on UK employment, and how often is it produced. Whether the occupational breakdown removed from the HR1 redundancy form would be restored. And whether the January assessment of AI and the labour market would be repeated.

The reply came on Tuesday evening from the Department for Science, Innovation and Technology.

The answer to the second question

"There are currently no plans to restore the occupational breakdown to HR1 forms."

Nothing was named as a replacement. That is the clearest sentence in the reply and it closes something this paper has been asking about since 8 September.

The answer to the first question

No single measure was named. The government said it "is monitoring a range of indicators using data from government and private sector sources", and pointed to two things.

The first is the AI Adoption Insights agreement, signed at the AI Summit in June. We read it.

Thirty-seven organisations have signed, among them Accenture, Amazon Web Services, BT, EDF, HSBC, LinkedIn, Microsoft, National Grid, NatWest, Rolls Royce, Sage, Salesforce, Sky, Vodafone, WPP and the British Chambers of Commerce.

They commit to sharing aggregated data on AI adoption with the AI Economics Institute. The document says this "may include use-cases from our adoption journeys, activities we are undertaking to support our workforce, and information about how we are redesigning business processes to embed AI within our operations."

Then, at the foot of the page, a note: "Data will be aggregated and anonymised, and there is no expectation to share sensitive or proprietary information. Firms remain in control of what they choose to provide."

So the terms are that firms decide what goes in, nothing sensitive is expected, and what arrives is aggregated and anonymised. There is no mention of redundancies anywhere in the agreement.

The second thing is the AI Economics Institute itself, which the department says "will build and analyse the evidence base on AI's economic impacts". On when any of it becomes public, the reply says the institute "will publish selected analysis externally where it supports open debate", and that the regularity "will increase as new policy questions emerge".

Selected, where it supports open debate, at a frequency that will increase. That is the answer to how often.

The answer to the third question

There wasn't one. We asked whether the January assessment would be repeated. The reply states that it was the most recent published assessment, links to it, and moves to describing the institute.

We checked the page. Published 28 January 2026 by the department with the AI Security Institute, described as "an initial, high-level assessment". Its update history has one entry, the day it came out. Eight months, no second edition, and no commitment to one.

Two instruments, side by side

One is compulsory. An employer making twenty or more people redundant at a site must file an HR1 in advance, naming a reason, with prosecution behind it. One of the reasons is new technology. The count is not published and we have been told to request it under the Freedom of Information Act. The part of that form that recorded what kind of work people did has been removed and is not coming back.

The other is voluntary. Thirty-seven firms have agreed to send aggregated, anonymised material of their own choosing to a new institute, which will publish selected analysis when it supports open debate.

The government pointed us at the second one.

On the thread

Logged 24 September: AI at the heart of Britain's G20 presidency, standards promised.

Added Tuesday: the Prime Minister said it himself in Liverpool. His commitment on AI was about Britain's standing abroad. It was to "a new global code to capture its benefits whilst being clear-eyed about its risks".

Added Tuesday evening: we asked which instrument shows where those benefits land in Britain. The government named a voluntary agreement in which firms remain in control of what they choose to provide.

Still open: whether anything the British state runs is being changed to show where the gain goes. On the evidence of this week, the movement is the other way.