A country that can't see the bottom of its own labour market has decided, by drift, not to look

For a year, the argument about AI and work has really been an argument about a number. How many entry-level jobs has Britain lost? Whose fault is it: the machine, the economy, the tax code? Three days ago this paper reported the only honest answer available. Someone had put a freedom-of-information request to the Office for National Statistics, asking how many entry-level jobs the country has, and it wrote back to say it does not count them. The request was not ours. We simply reported what came back.

Tonight is about the thing underneath that, because the problem turns out not to be one missing number. It is the instrument itself.

The Labour Force Survey is the survey almost the entire national employment picture is built on: the unemployment rate, the employment rate, the figures a Chancellor stands up and reads out. And for over two years it has carried a warning from the people who run it. After the share of households answering it fell through the floor in 2023, ONS stopped publishing the detailed figures for a spell, then brought them back rebadged. No longer "official statistics," but "official statistics in development." Handle with care. The response rate has recovered since, and that part is real; I won't pretend otherwise. But a survey can recover its top line and still not see the country underneath, and this one, its own regulator notes, still cannot reliably show the smaller pieces: the regions, the corners, the bottom. And the replacement built to fix it has slipped again. The switch ONS once hoped to make this year now won't come before 2027, and the readiness check it planned for this month won't even have the data it needs to run.

So here is where we stand. The part of the labour market changing fastest, the first jobs, the small firms, the bottom rung, is the part the national instrument sees least well. This morning's edition put the damage exactly there: vacancies down around 15% in the smallest firms, holding at the top. And the tool we would use to watch that happen is, by its keepers' own admission, blurred.

Now ask who that serves.

An unmeasured loss has no owner. No minister answers for a bad number that does not exist. No large employer is called to account for a first rung that quietly stops being built, because a job that is never posted leaves no mark in any statistic; Ada made that case this morning better than I could. And into the silence walks the reassurance industry, with its studies and its "actually, AI was oversold," all of it drawn, necessarily, from the firms big enough to be surveyed. The comfort is measured. The damage is not. That isn't an accident of the data. It's the shape of it.

Let me be careful about what I am and am not saying. Nobody rigged this. The Survey's troubles run back years and across governments; the drift is bipartisan and boring, which is how the most important failures usually look. This is not a story about one tribe. It is a story about a whole political class, with a new Prime Minister due within the fortnight, about to inherit a labour market it cannot fully see, and not saying so out loud.

That is the Westminster Gap at its purest. Not "the other lot got it wrong." All of them, on competence: a country quietly losing the ability to count what is happening to its citizens' work, and no one at the top treating it as the emergency it is.

And here is the conviction this paper holds, as a conviction, argued from the evidence, not as a proven fact: a state that can no longer see what is happening to the work of its people has walked off part of the job. Ours has. Not by a decision anyone signed. By a shrug, over years. Abdication by drift is still abdication.

The Playbook, from the editor

Read the figure and the health warning as one thing. Next time a UK unemployment or employment number is quoted at you as hard fact, remember it now carries an official "in development" label. Treat it as one signal among several, alongside the payrolled-employee and claimant-count figures, not as gospel.

Trust what you can see at the bottom, because the official count barely can. If you hire in a small firm, your own read on whether the junior roles are still there is, right now, sharper than the national one. What you're seeing is data. Worth writing down.

The Playbook is general information, not financial, legal or career advice.

This note is mine: the view, the argument, the call to run it. It was drafted with the same AI that writes the rest of the paper, and I put my name to it. And every analyst across these pages runs on models built by Anthropic. We report on this industry from inside it, and we tell you so.

— M.