Treat Legora’s university training as unproven until it shows legal judgment

On September 29, Legora expanded its legal AI training programme to UK law schools and signed its first university partners, Non-Billable reports. The event puts a concrete institutional form around a question that otherwise remains mostly theoretical: whether firms can use AI to make capable lawyers faster, rather than simply remove entry-level work. Legora has not, in the report, published a measure of trainee performance, a curriculum, a completion rate or a comparison with conventional training. The news is therefore the launch of a training programme, not proof that the programme develops lawyers. That distinction matters because the advantage belongs to firms that grow lawyers faster, not merely to firms that hire fewer beginners.

The programme supports the direction of AI-native law because it treats lawyer development as part of the operating model rather than as a leftover from the old associate pyramid. Thomson Reuters Institute says law firms need to redesign new attorney onboarding for the AI era. LawNext podcast reports that, as AI improves, demand is growing for lawyers with the judgment to direct it, check it and own the work; it also reports that Latitude says new law firm and corporate clients rose by 52 percent in the first half of this year. Those points support the need for a different training path, but they do not show that Legora’s path produces better lawyers. The training proposition is promising precisely because it can be tested against capability instead of being credited for adoption alone.

The strongest objection is that AI may eliminate the work through which associates learn, leaving firms to train partners instead. A post on Bluesky says firms will not train associates after AI and will instead train partners. That reading would cut directly against the stronger interpretation of Legora’s move: a training programme could be a response to the loss of routine work, but it could also be only a way to teach students how to use a product. The launch establishes neither outcome. For it to prove faster lawyer development, Legora would need to show that participants acquire legal judgment, not just familiarity with AI tools, and that they do so faster or better than comparable trainees. Nothing reported so far establishes that result.

The next confirming fact is a published result from the programme: the participating universities, the work trainees complete, the skills assessed and a comparison with a conventional training route. A later account that reports only partner count, access to software or student enthusiasm would show reach, not lawyer development. A measured improvement in drafting, review, fact checking or legal research would support the claim that AI helps build capable lawyers faster; those are the skills Legora describes as parts of legal work in the surrounding discussion. A result showing that trainees become better at prompting without stronger legal judgment would cut against it. We will treat the first outcome measure, not the expansion announcement, as the decisive fact.