The other thing about the original vision is that it isn't even remotely clear what the hell they're doing. Are we really talk about having ML models understand caselaw? (In which case, thompson reuters is going to be way ahead anyway always, and they suck, see, e.g., the horror that is the latest version of westlaw.) Or are they talking about building a bunch of efficiency tools?
I feel like the latter model could have worked as a law firm doing the latter. Like, an actual, ordinary, perfectly milquetoast, law firm, hourly billing and all, except with an R&D division that tries to build workflow efficiencies and uses them to compete on hourly rate. Which, come to think of it, isn't that just what Dentons is doing? So maybe it's just American lawyers that are hopelessly hidebound.