We said the number did not exist. It does. It is in a document about interest rates.

On 2 July I wrote that Britain's entry-level market was shrinking fast, that AI was the obvious culprit, and that the government's own data would not pin it on AI. That was true when I wrote it and I would write it again.

It needs an update.

Last Friday the Bank of England published the intelligence its Agents gather from British firms, and inside it is a box saying that automation is reducing demand for entry-level and junior roles, that several professional services contacts report reduced graduate recruitment, and that some of them are concerned about what that does to their talent pipelines.

I want to be exact about what has and has not changed.

What has not changed: there is still no statistic. This is not the Office for National Statistics counting jobs displaced by AI, because the Office for National Statistics does not count that. It is business intelligence, which means it is what firms told somebody. Tomás sets out below why that is weaker than it feels, and he is right about most of it.

What has changed is that the sentence now exists in an official British document, written by an institution with no reason to overstate it, naming the roles and naming the tasks that have gone.

So I was wrong in one specific way, and I would rather say so than let it pass. On 10 July this paper said we had spent a year arguing about a number nobody collects. Somebody was collecting something. They were not collecting it for us.

That last part is the thread, and it is why this is a piece rather than a correction filed quietly.

The Agents were not sent out to discover what AI is doing to young people's careers. They were sent out to find what is happening to demand, costs and pay, so that a committee can decide where to set Bank Rate. The finding about entry-level work is a by-product. It surfaces because it affects wages, and wages affect prices, and prices are what the Bank is required by law to care about.

Which means the evidence exists at exactly the resolution monetary policy needs and no finer. It gives you a direction. It cannot tell you how many, or where, or in which sectors, or whether this quarter is worse than the last. Those are the questions a labour market institution would ask, and Britain does not have one asking them. Yesterday this paper reported that when AI functions were moved to the centre of government last week, three were named and the unit set up in January to monitor what AI does to jobs was not among them.

I am not going to call that deliberate. It is more ordinary than that, and worse. Nobody decided not to measure this. The measuring is not anybody's job, so the only place it appears is where somebody needed it for something else.

So here is the test, and it is the same kind I have set before, because I do not think a bigger one is warranted. When any British body publishes a measure of AI-related displacement on purpose, with a number, a method and a repeat date, we will report it here and say the gap has closed.

Until then the honest position is the one the Bank has arrived at by accident: something is happening to the first rung, firms say so out loud when asked, and nobody in this country can tell you how much.