The Power of AI for the Legal Industry – Techstrong AI Podcast EP39

Hello and welcome to the Techstrong AI podcast. I'm Amanda Razani. I'm excited to be here today with Scott Stevenson.

He is the CEO and Co-founder of Spell book. How are you Dude, Allison. Uh, thanks for having me, Amanda.

Well, that name is so interesting. Can you share a little bit about spell book and what services do you provide? Sure.

Um, so spell book, uh, was the first sort of generative AI copilot for lawyers. Um, we launched back in 2022, and the name came from this idea of giving lawyers like a, a magical spell book that they could cast to help them do document and, uh, drudgery. Um, we mainly help with contract drafting and reviews.

So a lot of what our product does is helping lawyers find issues and contracts like say a hundred page, uh, legal agreements, um, and help and helps fix them or helps tell them when something is kind of off, uh, from normal. Um, we help them draft and, um, you know, do a whole bunch of other things. Um, we have over 2,500 law firms and in-house teams using us today.

Um, and, uh, yeah, customers se seem to love the product so far. Wonderful. That's a lot.

Well, I don't think that we've ever talked about how AI is impacting lawyers and the legal industry, I don't believe on this show. So can you share from your experience, how are legal teams using ai? Yeah.

Um, so we primarily service transactional lawyers, so that's lawyers who are working on, uh, commercial transactions like sales transactions, setting up companies, um, employment agreements, uh, real estate, um, all of these sorts of things. So the, and the other side of law li litigation, so we don't deal in litigation. Um, yeah, I I would say one thing I one stat I will mention at the high level is, uh, AI is being adopted five times faster by law firms than the cloud was.

So it is being adopted incredibly, incredibly quickly across all practice areas of law and faster than almost any technology has been adopted since maybe like email or, or word, word processors. So that's how big it's, and um, at a high level, I think, uh, there was a lot of sort of pent up potential energy. Lawyers wanted to adopt technology for a long time and nothing really helped them that much because what lawyers deal with all day is unstructured text and most software out there like say CRM, the software, um, uh, I don't know o other sorts of like DD everything out there that has come out over the past like decade has really been a lot of database software, which is good for like categorizing structured information and law is the opposite of that, is it's big documents of unstructured text.

And, um, when we started our company we're actually originally called Rally. Um, we were another company trying to shove legal text into a structured format. We had a templating engine where a lawyer trying to make an employment agreement could create a template and then, you know, push out, uh, say employment agreements really quickly, uh, using this templating engine.

And it just never worked very well because, um, lawyers would always say, well, the deal I'm working on is too bespoke, you know, it's, it, it, my work, my work is really bespoke for my clients and any kind of template or automation is just not doing it for me. And this has resulted in legal fees staying like really surprisingly high for a long time, um, because no technology really helped lawyers that much. Uh, and then, uh, generative AI came along and totally changed the game overnight, um, because all of a sudden we could help lawyers with unstructured text, we could help them find issues in the unstructured text, we could help them draft unstructured text, um, we could help them understand and ask questions about un unstructured texts, and that's what they're doing all day.

And, um, yeah, it's, it's all, all of those things I just talked about, um, are how, uh, AI is being adopted today. And I think there was so much drudge work built up in this industry. I think it's, a lot of our investors talked about it as like one of the last unturned rocks.

It's like software never figured out how to help lawyers until very, very recently with ai. Um, so, um, yeah, it's, it's used to really almost in some cases 10 x the speed and efficiency and accuracy of things like contract reviews. Um, that's where we see it being used most.

So, um, say you're doing a big m and a transaction, um, there could be millions of dollars at stake. Um, you might have 20 core agreements that you're working on, like a share purchase agreement and that, uh, a voting agreement and investor rights agreement. Um, and then you might have a data room of the company that's being acquired in say, an m and a, um, that has thousands of legal documents in it.

And if one document in there has something that's wrong, it could blow up the whole deal or could, it could be too big a risks to take on. So AI helps in this type of transaction tremendously. One in those thousands of documents.

It can help highlight, you know, a, a game changing or kind of critical issue, um, that's gonna block the deal. Um, or, uh, it's, and it's gonna help in the drafting of those like 20 really complex sensitive agreements in, in, um, making sure all the right data gets in there. And also it's gonna help in the negotiation process so that both sides know they're getting like a fair deal or a decent legal document that's normal and not off in some, um, uh, important way.

So, uh, and I, I really think it's, it's a true like 10 x to 100 x gain in, in value for, uh, the clients of law firms, um, as well. So yeah. Yes.

I can't even imagine having to read through pages and pages and pages of text and or write them and how much more efficient it would be using generative ai. Yeah, it's, um, way more way, way more efficient. Yeah.

Like the scariest thing that I always think about, like the drafting text is difficult. Um, but the scariest thing is when your lawyer and your client hands you a hundred page document and they say, Hey, can you review this and make sure nothing's wrong? It's like, where do you start on that problem as a lawyer?

Are you gonna boil a pot of coffee and then read through a hundred pages and make sure that there's nothing, um, you know, dangerous, um, in that agreement? Uh, I mean, that's literally what you used to have to do. Um, and now with ai, you have this extra layer of checking that can go up happen where within say, 10 seconds, you can find out that on page 80, you know, this contract says you're giving up rights to your intellectual property that you didn't know you were giving up.

Um, and yeah, it's, it's totally game changing. So is it right now just an extra layer, an extra tool? Is it still necessary to have a human also still read through all of that?

How accurate would it be to just let generative AI handle all of it? Um, yeah, so one thing I mentioned is we do not let generative AI handle all of it. We don't recommend that our customers do either.

Our UI is built to be, we say it's like an electric bicycle for lawyers. So, um, when a lawyer's using our product, we are giving them suggestions that they are reviewing and they have to go through and review every suggestion and say, yes, I agree with it, or no, I don't. It's, there's never a point where our tool is just kind of doing everything and says, here you go, here's the result.

Um, and that's a really important part of our product design. Um, and this is one of the issues with, with chat GBT, like chat, GBT is great, um, for certain things, but it kind of just spits out answers to you and you know, you can take it or leave it. Um, whereas in our products, um, you know, we take each suggestion and we say, Hey, it's a suggestion.

Um, and you have, but you ultimately need to decide whether it's a good one or not. Um, uh, I think that's really important and I think we'll see more products designed that way, um, in the kind of months and, and years ahead. So it is improving the outcomes, but is it still also, uh, speeding up the outcome and, and reducing the amount of work time?

Yeah, I mean, it, it depends on the type of transaction and whether it's work time sensitive, but I would say yes, in many cases it can cause transactions to be completed much, much faster. Um, and it's also improving the work product at the same time because yeah, like I was talking about with sale 100 page agreement, I mean, a human just is not very good at finding a needle in a haystack of a hundred pages. Um, and it's just not something we're very good at.

So I think you, you'll see both better work product, um, lower time spent on drudgery and, and reading and copying and pasting, um, and more time for lawyers to work on the more strategic work, which began as more value to the client. If the lawyer's able to spend more time on say, the human negotiation element, um, that can provide a lot more value to the client than the lawyer, you know, spending their time copying and pasting Microsoft Word. So do you foresee in the future that as this tool advances and it does speed up all the work time, do you think that would make the legal pricing more affordable in certain areas to clients?

Um, yes, a hundred percent. And that's like the mission of our company, uh, when we started, was to make legal services accessible that is like our actual top level mission. Um, and we're doing it by reducing all the drudgery, uh, for lawyers because lawyers don't feel good about sometimes what they're billing either.

Um, and my co-founder Dan, he's the lawyer on the team, and he went through law school and then was billing clients, you know, hundreds of dollars per hour to copy and paste Microsoft working. He, he didn't feel good about that. Uh, for me, I had a small business before I started SPEL work and one day I got a legal bill that took half, half the cash out of my bank account and for to do some like pretty rudimentary things.

Um, and that was a really surprising moment. So yes, I think the price of legal services will generally come down a lot where it's price sensitive now in big law and really large transactions. And if you're doing say a $15 million transaction, you know, people are less worried about how fast does it take and is, is it cheap enough?

Um, and people are more worried about quality. So it depends on the type of transaction. If you're a small coffee shop and you're just figuring out how to get incorporated and how to get your business set up, um, yeah, you're gonna get lower prices, um, people are gonna be able to do those basic things more efficiently.

If you're like Coca-Cola and you're trying to do a big, uh, m and a transaction or something like that and it's worth a hundred million dollars, you probably don't care that much about the legal fees. Um, what you do care about is that everything is done really, really well and that nothing is missed. And I think in both those circumstances, uh, AI is gonna be, uh, really helpful.

From your experience, I know that we're still trying to get some good regulations and, um, laws around use of ai. So, uh, from your experience, what should lawyers be looking at and listening to? I know we've seen some recent bills come through in regard to ai.

Um, how does that weigh in? Um, yeah, I think we should be looking at regulation. Um, I think it is still very early to understand how these technologies are gonna impact the world.

And I'm ner I'm nervous a little bit that we're gonna kind of react to early. Um, and I think you've seen some early kicks of the can, uh, for regulation that just didn't end up making a lot of sense based on how the industry has already evolved. Um, in terms of our customers and our company.

Um, we don't see any, um, uh, threats or issues on the horizon. Like for instance, we don't use any customer data for training our models. I think that's probably the biggest concern in law is like, is my data being used to train a model and um, is my client, or actually more important is my client data being used to train a model.

Um, and I think that's an important, uh, area where regulation there should be some kind of regulation and oversight, um, because people don't want their information baked into these models and spit up, spit back out, um, you know, in, in a way that they didn't know was going to happen. Um, so I think that's really important. Um, we already as a company have all sorts of safeguards in the place to make sure that doesn't happen and we just don't use customer data for training at all.

Um, except we do allow firms to have sort of their own personalized model. So we, I think you're, um, because of new advancements in ai, like in in context learning, um, we are now at a point where say each of our customers can have their own personalized system, um, that has learned based on their own data, but just for them, um, not shared with anyone else in the early stages of LM and generative ai, you would have, um, you know, all of the data being funneled into one model and then everyone uses that one model. Um, but now because, um, I know that this is something you've talked about on the show, but now context windows are very, very big, which means when we go to use an AI model, we can shove a ton of extra information into there and that allows us to do something called in context learning, which means, you know, when law firm a goes to use spell book, um, we can actually stuff in all their preferences into their use of, uh, spell book and they can get hyper personalized results out.

Um, while we're not actually doing any say training per se and we're not sharing any of that data with any other customer and that can, that can be done in a really tight and controlled way. So, um, all that's to say, uh, I think privacy is probably like the most important aspect of regulation right now. I think we're very prepared as a company to provide very like privacy friendly solutions to our customers so we're not too nervous about, um, the regulation.

Awesome. Well if there's one key takeaway you could leave with our audience today, what would that be? Um, it's a great question.

Uh, I think AI agents are the next few thing that, so we didn't get to talk about too much. Um, and I'm sure you've talked about agent song here, but like we are another company that really, really deeply believes that AI agents are kind of the next big thing. I think it's gonna be bigger than chat was, um, because we're gonna go from having these chat assistant that help us here and there to now having sort of these AI colleagues that can take out whole projects.

Um, and we, we actually launched the first full-fledged, uh, agent for legal work as well. It's called Spell Book Associate. Um, we're seeing early reactions from lawyers and now these agents can go and do these complex multi document drafting, um, workflows.

And it's like the reactions we are getting from our customers are just holy, their minds are blown as much as when they first saw chat GBT. So I think that's the biggest thing that we're seeing now is like agents are really real. Um, they're definitely coming.

It feels so much like chat GBT and Ls in the early days. There's the skepticism in the air, is it really gonna work? Is it really reliable?

These are all the things you heard when G GBT three came out and no one was sure if it was gonna work and then it did. Uh, and then GBT four came out and so on. I think you're seeing the same thing with agents now, and now the oh one model just came out, which is really good for powering um, these agentic workflows.

Um, so yeah, I think in 2025 we're just gonna see a ton of adoption of agent based tech and um, yeah, it's gonna be like we have ai, AI colleagues. Awesome. It will be interesting to see how that technology unfolds.

Alright, well thank you so much for coming on our show and sharing your insights with us today. Awesome. Thanks for having me at Amanda.

Alright. And thank you to our audience. Stay tuned.

There's more.