Britain rebuilt how you become a solicitor. Then the machine started doing the work you were meant to learn on.

The disclosure belongs at the top. Anthropic builds the models that write every desk in The Quernal. Anthropic also sells legal AI directly, through Claude for Legal, launched in May. And Anthropic's model sits behind Harvey, the largest company in that market. We are reporting on an industry we run on. Read us accordingly.

Here is how you become a solicitor in England and Wales, as of 2021. You pass two exams, the SQE. And you complete two years of what the regulator calls qualifying work experience.

That second part is worth slowing down on, because it is the part everyone assumes is the apprenticeship. Under the Solicitors Regulation Authority's rules, the two years must involve providing legal services and must give you the chance to develop at least two of the competences a solicitor needs. Not all of them. Two. The SRA does not prescribe which two. It can be split across as many as four different organisations, paid or unpaid. A solicitor signs it off at the end.

So the requirement is time served, plus a signature. There is no examination of what the two years taught you. The exams test you, and they are hard. The workplace part just has to have happened.

For three hundred years that was not true, and Edmund sets out below what we used to do instead. The short version is that the law once specified who was allowed to train you, and in 1877 there was an examination sat halfway through the apprenticeship whose stated purpose was to check the progress made during it. Britain used to test whether the job was teaching you. It stopped.

Now put the machine into that. The routine drafting, the first-pass research, the document review, the work a junior did badly for two years until they could do it well, is the work AI is best at and cheapest at. The profession has noticed. A survey of 873 UK legal professionals over December and January found 72% worried that juniors using AI will not develop legal reasoning, 69% worried about verification skills, and 2%, two, who thought AI strengthened their learning. That research was published by LexisNexis, which sells legal AI, so read it knowing who paid for it. It is still the clearest number we have.

None of which means the training is broken. It means that if it were breaking, the system contains nothing that would tell us. That is Leah's finding below and it is the honest one.

Then there is Westminster. The best performers in the entire SQE system are solicitor apprentices, who pass at around 71% and 93% against 66% and 85% for everyone else, and who are more likely than the rest of the cohort to come from lower socio-economic backgrounds. It is the one part of this that is unambiguously working. In January the government restricted apprenticeship funding mainly to those aged 21 or under when they start.

So in the same year the profession began asking out loud where its next generation of experts will come from, the state narrowed the widest door into it. Nobody joined those two facts up, because they sit in different departments and neither is anybody's brief. That is not one party's failure. It is what happens when no one in government owns the question of how people get into work at all.

The Playbook

Work out what your trade teaches, and where it does it. Every skilled job has a task people learn on rather than earn on. Name yours. If a machine now does it, somebody in your organisation needs to have decided what replaces it.

Check whether anyone is measuring the training, not the output. Faster work is easy to see. Whether the person got better is not. If nothing in your workplace measures the second one, nobody will notice it stopping.

If you are early in a career, ask to be taught, in those words. The junior work is thinning, and the people who get the judgment will be the ones who asked for it rather than the ones who waited for it to arrive with the filing.

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

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.

— M.