Somewhere in Companies House sits the most honest account of automation in Britain, and it is an annual report
Every argument about what AI and robotics will do to work runs into the same wall. Nobody can agree on the numbers, because almost nobody publishes any. Surveys ask employers what they intend to do. Consultancies model what might happen by 2035. Politicians promise retraining into jobs they do not name. What almost never appears is the thing you would want most: a measured, audited account of machines absorbing human work, in Britain, with the cost attached.
It exists. Ocado Retail Limited, the online grocer owned jointly by Marks & Spencer and Ocado Group, filed its annual report for the 70 weeks to 6 April 2025 last autumn. Deloitte audited it. It was signed on 2 September 2025, and the trade press reported the top lines within days: revenue up, earnings up, more customers. Then everyone moved on. Nobody, as far as we can find, read it as what it also is: the closest thing Britain has to a public record of substitution, updated on a statutory schedule, free to anyone with an internet connection.
Here is what it records. Ocado runs two labour-heavy operations. Picking groceries in warehouses, which it has automated. And driving them to your door, which it has not. On the 52-week comparison the company itself publishes (the statutory period is an awkward 70 weeks, because the firm was aligning its year-end with M&S), orders grew 15.2 per cent. The costs of the automated warehouses, which the document says primarily comprise labour and consumables, grew 3.7 per cent. The costs of delivery, still human, grew 22.7 per cent. And the company names the cause of the first in its own words: network reorganisation and "increased adoption of on grid robotic pick".
The company does not draw the comparison between those two lines. We do, and today's edition is that reading, held tightly to the documents. The pattern did not stop when the audit closed. In the half year to 31 May 2026, reported on 16 July, warehouse costs rose 1.7 per cent against 12.8 per cent more orders, while delivery costs rose 14.9 per cent.
One caution before anyone reaches for a national conclusion, and it stands over every piece today. Ocado Retail is one company holding roughly a seventh of online grocery, and online grocery is itself about a seventh of British grocery shopping (Nielsen, reported in the July interims). Call it under two per cent of the nation's food shop. It cannot carry a claim about British warehousing, and we do not ask it to. What it can carry is this: the ratio between human labour and output in an automated British warehouse is published, audited and moving, and the company sells the machinery to other grocers while naming, in its own materials, the work it removes.
This is what we mean when we say the ground is the primary. Not a leak, not a forecast, not a minister's speech: a filing, free to read. The lag between what the machines are already doing and what the state still believes about work is this paper's whole argument, and here is a piece of that lag with an audit trail. We have looked for any sign that the people planning Britain's labour market read these documents the way you are about to. If we find one, we will report it.
What to watch from here, since Saturdays in this paper look forward. Whether the delivery line bends as routing software squeezes the one leg still human. Whether the fees Ocado Retail still pays for Hatfield, the closed and least automated site, £15.6m in the latest half year alone, run their course. And what the numbers show when Asda's Ocado-powered operation goes live next year, because the machinery in these accounts is about to run a second British supermarket.
— M.
*This note is mine: the view, and the call to run it. It begins as a draft, drawn from work the AI and I have researched and argued out together, the same way every desk in this paper is made, and I answer for every line because I read every line. Those desks run on models built by Anthropic, one of the labs sitting on the very scoreboard we report, so we tell you plainly: we cover this from inside it.*
The paper's thesis, with an audit trail: the distance between what machines are already doing to work and what the state believes is measurable, and here one end of it is published under audit on a statutory schedule.
Held at thesis level, not as a sourced absence: the letter states the paper's standing argument about the lag and records, honestly, that we searched for state engagement with these filings and will report any found. No specific ministerial absence is asserted.