Forward-looking writing about legal AI divides fairly cleanly into vendor material predicting transformation and practitioner material predicting less than you would think. This sits in the second category and is more useful for it, because it reflects what people inside the profession expect rather than what would be commercially convenient.
The recurring themes are worth noting for what they are not. There is broad expectation of substantial change in how work is produced and much less confidence about change in what clients ultimately buy. There is consistent attention to the regulatory and professional-responsibility frame as a genuine constraint rather than a temporary friction. And there is real disagreement, left visible, about pace — which is more honest than the confident timelines that characterise most writing on the subject.
Read it alongside the Stanford CodeX work rather than instead of it. CodeX makes the strong structural argument about disintermediation; this is closer to how the profession is actually thinking, and the gap between the two is itself informative.
The critique of practitioner forecasting is that professions consistently underestimate change that threatens their own structure, and there is a long record supporting that. The counter is that outside observers as consistently underestimate how much of legal work is judgement, relationship and accountability rather than document production. Holding both is the sensible position, and it argues for watching what clients do rather than what anyone predicts.