Legora’s first university partners are a test of developing capable lawyers

Legora is expanding its legal AI training programme to UK law schools and signing up its first university partners, Non-Billable reports. The move puts Legora’s technology inside legal education before those students enter practice. It is a concrete expansion of training infrastructure, but the report does not identify the universities, describe the curriculum, state how many people will take part, or report any result from the programme. The immediate news is therefore the partnership and the expansion, not a measured improvement in lawyer capability.

The move supports the direction that matters: firms gain a durable advantage by developing capable lawyers faster, rather than merely reducing the number of beginners they need. Training law students in legal AI can reach the point at which lawyers learn to direct, check, and own machine-produced work before those skills are embedded in a firm. LawNext podcast reports that Alex Su said demand for lawyers with real judgment would grow as AI improved, and that Latitude had seen a 52 percent jump in new law firm and corporate clients in the first half of the year. That account is a company-reported commercial signal, not a result from Legora’s university programme. It does, however, give the training question practical stakes: if AI increases the volume of work, the scarce capability may be judgment rather than basic production.

The stronger reading remains unproved. Legora’s expansion shows that a legal AI company is putting training into UK law schools; it does not show that the programme produces capable lawyers faster than conventional training, or that those lawyers perform better once they reach practice. Jordan Furlong says law firms will not train associates after AI and will instead train partners. That is the central objection, because a university programme could become an early channel for capability while firms reserve judgment-heavy development for senior lawyers. But the announcement supplies no comparison between participants and nonparticipants, no assessment of legal work, and no time-to-capability measure. It therefore exemplifies the proposition without establishing it. For the stronger claim to hold, training would have to be connected to observable improvement in the work of people who pass through it, not just to wider access to an AI tool.

The next confirming fact is not another university partnership but a published result from the programme. Legora would need to identify the participating universities, explain what trainees are taught, and report a measure that can be compared with ordinary legal training: how quickly participants reach reliable performance, how well they identify and correct model errors, or how much judgment-intensive work they can perform under supervision. A follow-on account showing participants entering firms with stronger capabilities would support the faster-development thesis. A programme described only through partner names, enrolment, or tool access would not. The same test applies to the senior-lawyer objection: if firms still have to rebuild judgment training after hiring these graduates, the programme has expanded AI education without solving the development problem.