00:00What's so interesting is about the human in the loop, the oversight, the control. How do they
00:04maintain that when you're leaving AI agents to go and do work for multi-step processes?
00:08Yeah, thank you so much for having me. I'm excited to be here also for Legal Week, which has basically
00:13all of the legal tech industry in New York for one gathering. The way that we think about it is
00:19in law, it's incredibly complex and it's multiplayer. And so if you're building this
00:23agentic system that is kind of the infrastructure for completing legal work, you need it to be able
00:28to actually interact with all the different specialized agents to do that task, the multiple
00:33law firms working on a project and the Fortune 500 that is thinking about buying another company or
00:38involved in a litigation or something like that. And so it's a really complex process of integrating
00:43with all the correct data, having the humans in the loop when they need to be and handing off between
00:48the different experts that are specialists in that particular domain. Let's talk about your
00:52specialization and the fact that you have all these integrations and have such a wide scope of legal
00:57companies already working with, because the age-old criticism is you're just a rapper of GPT or
01:04whether it's anthropic, you're agnostic to the models underlying. How are you making yourself so
01:08necessary for the legal profession? Yeah, the best way to think about this is in these incredibly
01:13highly regulated industries, the work that they have to do is very complex and all of the data is in
01:19all these different silos, right? And again, it's multi-party and collaborative. And so what you need to solve is
01:24which model is the best for this particular part of a task? How do you route that to the human?
01:29And then
01:29what's going to become really important is how do you actually review all those outputs and evaluate
01:34them for accuracy? If you think about what's going to happen in the next 10 years, think about how
01:39complex an international merger is going to be, right? Now that we're using AI to produce more products,
01:44to produce more MSAs, more contracts, et cetera, reviewing those, honestly, in 10 years, I think you can't do it.
01:51A human cannot do this without AI systems that can integrate with everything else.
01:55Winston, I want to hold you to this to be succinct. This is a tool that makes a practicing lawyer
02:01more productive
02:02and better at their job. Or it is a tool that in the future displaces the role or function of
02:08certain team members
02:09within a legal practice. Yeah, the best way to think about this is this is task automation. And I think
02:16it will always be task
02:17automation, right? So if you think about like a very complex fund formation that involves 50 LPs or something
02:23like that, there are certain tasks like doing comment memos or something like that, that are going to get
02:29automated away. But a lot of the high value is that strategic advice on top of those tasks, right?
02:35And so as these tasks get automated, the job of a lawyer is just going to level up over time.
02:42I ask you that because one of the show's biggest fans, my father, is almost certainly watching,
02:48who has more than three decades of practice in English and European law. And I remember as a kid,
02:54the whole billable hours thing, right? You know, lawyers across the world used to scan barcodes for 15
03:00minutes of work. Really interested, therefore, for how you price this. You know, I think you guys run
03:06a subscription-based model, right? Which can also be priced depending on the volume of use of any
03:11customer. But that's the debate. Where's the value? Yeah, exactly. So right now, that is how we do
03:17all of our pricing. Over time, what we're starting to do a lot more of is how do you do
03:22custom build for
03:23a lot of different customers? And the reality is doing these custom builds, these forward deployed
03:28solutions, is with how good the coding models are getting, a lot, lot stronger gross margins, right?
03:34And so a lot of what we're starting to do is we'll work with, like, a large bank or private
03:38equity or
03:38telco and a law firm that's partnered with them and will actually redesign or transform how they deliver those
03:44services entirely, right? And that we are going to price much differently and probably closer to how do the
03:51outcomes work and how does consumption on top of that work?
03:54You've got some heavyweight investors, Sequoia, KP, A16Z, GV, the list goes on. They want to hear
04:00about how you're able to monetize this. When someone is sat within the innovation part of the
04:06legal business and they're like, do I do a Claude plug-in? Do I do Harvey? Do I keep with
04:11LexisNexis?
04:12What's your answer? Yeah, the best way to think about this is everything else out there is a point
04:16solution. And what we're building is a platform, right? And with that platform, you have all of the
04:21integrations, you have security, you respect ethical walls, permissions. We grab the correct
04:27data so we ground it in truthful data instead of bad data that is going to give you a bad
04:31answer.
04:32And as this scales over time, you really just have to outrun these model providers that are trying to
04:37build the general system, right? And the bet that I have, and I'm very confident in my team's ability
04:43to deliver on this, is we can race faster in the legal vertical than they can generally.
Comments