insight
AI and the junior lawyer pipeline
Automate the grunt work entirely and you erase the route that produces the partners who supervise the AI. Training is not a side effect of junior work — it is the product.
Every generation of legal technology promises to remove “drudgery.” Some of that promise is real and overdue. Document assembly, first-pass review, and routine research scaffolding can be faster with machines. The risk is not speed. The risk is deleting the apprenticeship that sat inside the grunt work.
Grunt work was never only grunt work
First-year lawyers do not review a bundle solely to produce a summary. They learn how facts fail, how authorities conflict, and how a supervisor thinks when the file is messy. Remove the task without replacing the pedagogy and you do not get efficiency — you get a thinner profession.
Firms that treat juniors as interchangeable prompt operators will discover, a decade later, that they have fewer people who can tell when the model is confidently wrong.
Supervision cannot be imported
Senior lawyers are not born knowing how to supervise. They acquire judgment by being supervised — by having their drafts returned with red ink, by sitting in the room when a risk call is made, by watching a partner refuse a shortcut.
If AI absorbs the drafts that used to generate that feedback loop, the firm must invent a new loop on purpose. “The model drafted it and a partner glanced” is not a training system. It is a liability transfer.
Keep the formative work visible
A healthier pattern:
- let machines take the mechanical first pass
- keep juniors responsible for the contested middle — where judgment lives
- require seniors to leave a trail of why, not only approved
That pattern preserves throughput without pretending that expertise arrives by download. It also produces better audit trails: the junior’s reasoning and the supervisor’s dissent are both part of the record.
Build tools that teach, not only tools that finish
Product teams will keep shipping features that complete tasks. Firms should prefer tools that expose intermediate work: sources, assumptions, open questions, and the places where a human must still decide. Completion without visibility trains nobody — and eventually leaves nobody competent to supervise the completer.
The junior pipeline is not a cost centre to automate away. It is how the profession renews the people who will be asked, years from now, who signed off the work.