I thought the post was good overall. However, saying 90% error is the same as 10% is an exaggeration. Let’s grant that 90% error is entirely worthless. Now imagine you’re under a tight deadline, and you only have time to investigate one thing. A 10% error service could help you choose how to use your time best.
Are the secondary sources 100% correct? Yes it helps that “another lawyer read everything for you.” But is the chance of them having overlooked or misinterpreted something 0%? It must be the case that they are not entirely perfect, what is, and yet not worthless. Similarly, is it helpful to have a junior associate do a first pass of something? Aren’t they likely to make mistakes?
As you anticipate, the AI will get better, and I expect it to outperform average lawyers in our lifetimes in the next 30 years. For an even longer time, the most skilled analysis will come from lawyers who leverage AI to their advantage.
Stepping back, everything is on a reliability spectrum. Just because something isn’t perfect doesn’t mean it’s worthless.