Welcome to the Mind Over Law Podcast, where we break the traditional rules of practicing law.
Our focus is helping you first to become a better, happier person, which in turn will make you a better, happier lawyer both in and out of the courtroom. We will combine mindset and energy practices grounded in ancient wisdom along with cutting edge neuroscience give you those skills. Plus, I'll have deep conversations with some of the most thoughtful leaders that will share their life stories, their leadership journeys and their legal practice wisdom.
I'm Lexlee Overton and my promise is that each episode will offer practical insights and strategies to empower your law practice, your leadership skills and most of all, your personal well-being. Join me and I promise you'll become a better you lawyer and leader. Welcome back to the Mind Over Law podcast. I'm your host, Lexlee Overton, and today we're diving deep into a subject that is transforming the legal landscape faster than ever before, artificial intelligence.
And my guest today is someone who is leading this transformation in the legal field. Jay Madheswaran is the CEO and the founder of E, an AI powered platform built specifically for plaintiff law firms. With a background as an engineer at Facebook and a leader at Rubric, Jay has spent his career at the cutting edge of technology, and now he's bringing that expertise to the legal profession. Jay isn't just here to talk about what AI can do, he's here to explore what it should do.
And his mission with EVE is not just to automate legal task, but to fundamentally reimagine how law firms operate, which helps lawyers to do more of what they're best at without burning out. O excited for this conversation? Let's dive in. Jay, I'm so excited to have you here today.
Likewise, I'm so excited to be here, Lexlee, thank you for having me on.
Yes, this is going to be a fun conversation, so I'm just going to dive right in. I have not been practicing law for the last 10 years, but I do know that in the last 10 years when AI was introduced in some ways and then like even in the last three years, three to five years, that the first thing that I heard from people were that you shouldn't trust it. There must have been some hurdles for you coming into this space with lawyers and law firms. What we're some of the things that you saw and how did you overcome those challenges?
But that is such a good question, right? Law is all about getting the facts right and being as accurate as possible and as smart as possible with how you leverage those facts. And obviously one of the biggest challenges with AI when we started the company from actually even before LLMS, right? So we had something called Transformers, which is like the precursor to large language models and ChatGPT.
This was always an issue is the way the technology relies on is it tries to guess the next word. And this presents a problem for us in the legal industry because it has a chance of hallucinating, which is a term I think everyone is sort of. But the most obvious ones being making up case law, but less obvious being making up the client's name, making up the address, making up basically be overly enthusiastic about answering your question to the point of making things up. And as you can imagine, this was the biggest challenge we had early on in being able to take this use case and actually serve as the really broad variety of tasks that happens in litigation and plaintiff litigation general, where you have these life cycle of very different questions, right?
If you think about it during intake, they're typically trying to understand the basics, right? And if you're in Pi, if they get in an accident, did they have some sort of injury? Did they have insurance? So on and so forth.
Relatively straightforward questions and you're just trying to make sure you're catching that accurately and you have that information in before you can do further analysis with it. In L&E, you might need to go deeper into some questions. If it's in a workplace violation, OK, then ask these follow up questions, etcetera. And after that point, that's when some of the creativity starts, right?
So after that, you may want to get a bit creative and depend later in terms of persuasive argument styles or imply certain damages are being done. That would basically bolster your claim a bit more. And there you actually want creativity, not just pure facts. And as a result, the hardest thing for us to solve was how did we actually make Eve cure it across the entire case life cycle with all these different questions where sometimes you want exploratory answers like discovery request responses.
Let's say you wanted to account for what might the opposing counsel ask or what is their strategy like and do a bit of that game planning beforehand. Those questions are very different than the questions you do for a factual analysis. So what we did to account for this problem is a lot of machine learning work. So you have to be able to generate.
We have a creative process of creating a data set that we can rely on these questions and answers and we're doing some clever things behind the scenes given we don't use customers data to train our central large language models. We have created a system where we have a very good understanding of what are the types of questions people ask through the case life cycle. And we've optimized accuracy across the stack to make that happen along with very basic legal specific optimizations like it tells you when it sites case law and it verifies it and if it and it tells you when it can't and makes it very obvious. And when you pull together quotes, which is often times something you'll be citing in legal documents, it'll actually go and check using non language models.
So using traditional technology think control, it lets you go do that for you and verify the data is correct. So those are the ways in which you've built trust into the system. And trust is like a holistic problem for us, going from machine learning training to how we actually analyze how our users use the product and head off any major, oh, like, I mean, citing case law that doesn't exist. It's a huge hole.
And we have multiple steps in the product to prevent you from doing it.
OK. So I have some questions. I want to know if I was a, let's say, a plaintiff's lawyer that's coming to you. I have a personal injury firm, which is what I had for a long time.
Generally, I know how I can use AI to find some information, things like that. I also know that there's the whole thing about using it for legal research, and that also has to be still verified and things like that. But if I'm coming to you and I'm a potential client, I think a lot of people probably don't know all of the things that you're doing with your company. Give us an idea of like how you're using II to help law firms.
What specific things can it be used for? So you've mentioned something about intake, you've mentioned a little bit about discovery. Can you give us more in depth on that of what is some of the services that are the ways that it can help us day-to-day in a law practice?
That's a really good question. And actually this is one of the biggest value I think we've learned that we provide is we've invested a lot behind our client success team that partner with law firms individually and get them on the path of becoming AI native law firms. And the hardest part is actually to know where to start and how to be thoughtful with where you use it for. Some practical advice I can give is if you're in personal injury especially, a lot of your work tends to be pre lit and think think gathering documents, medical records, bills, etcetera and then putting together some sort of demand letter or a medical chronology that might be the precursor.
Demand letter is a lot of what you're doing in prelit, at least the time consuming pieces.
And is that something you know, that's very time consuming, right? I used to have employees that they did nothing but gather medical records, read those records, create chronologies by exactly looking is summarizing every visit for what was relevant and also not relevant because you never know what they're going to bring up and to say something about your client. The when you look at the trust factor on that, it's a little different because it's analyzing documents you're giving it. So talk about that process and what it does.
Yeah, absolutely. So as you can imagine, EVE actually does help with medical technologies in demand letters and that's a pretty popular use case in personal injury for us. And the hard part there is exactly what you mentioned. How do you be cured and how do you be fast?
That also matters, especially for the larger firms. And then how do you actually use that information for everything downstream? Because once you have that chronology, often times you have to update it when you get new bills and records over time and you might have to redo a demand letter or redo some work downstream as a result of changing information. So what we've done is we've actually consolidated all of that into the EV platform.
And how EV works is for every single page that you upload onto EVE, we have invested in significant amount of machine learning to make sure we're able to extract the important information accurately. And we also give you sources of where we get it from. Every time we give you any sort of information at all, they actually tell you exactly where we get it from. And we also verify that we got it properly by using some tricks, right?
Where we don't actually use AI, we can do other things to know that is accurate. And that's actually part of how Eve is able to accurately and quickly give you actual real answers. The second piece is iteration, right? So once you have that, you often times might want to massage it into your own framework.
For example, Eve is already capable of pulling out things like bad facts you find in medical records and just the data set in general. People, our customers also have their own unique ways of digging deeper in that information. And Eve actually comes with capabilities where you can teach it how it is that your firm analysis and looks at that type of information. And you can encode this into blueprints is what we call it.
Then with the click of a button, Eve was actually able to go deeper into that data and do that analysis, which could either be double checking work on Eve or deeper analysis in terms of key pieces of insights you want to pull out that other firms normally don't pull out.
Is there you have a an example of that on top of mind?
I think some of the most common ones is actually happens a bit before medical chronologies where it's ACE evaluation. In case evaluation, we've seen number of firms actually make pretty intelligent small tweaks in terms of how they decide which case to take on. And the most common ones to some firms try to optimize for kind of higher value cases and they have a lot of signals they come up with for what might be a higher value case. And then based on how that case settles, maybe there's one of the really interesting things I think which will happen with AI and what Eve is already doing is if we think about it, typically traditional practices and plaintiffs go from like intake some sort of pre lit discovery or demand and then lit trial.
That's a funnel, right? That's how most people view it. What happens with AI and what happens with Eve is that's going to turn into a loop where now every time you actually finish a case and litigate a case and settle a case, you can actually take that information, gather insights intelligently and feed it back into how you take on even new cases to begin with. And this is a type of danger I think most law firms have tried to do intuitively, but they haven't been quite as systematic in doing that.
If you think about why is it that tech companies are doing so well in the stock market and growing, this is something they're all doing across the board is they've all found a way to take their data and make their products and goods and services significantly better year over year versus like in the plaintiff law world. That's be very hard for us to do right. We're chasing new clients all the time. We are just getting the job done and adapting to typically new laws and regulations and other peaks.
But very rarely, only when you really get to a certain scale do you start analysing that data and scale and make really intelligent day-to-day changes in your entire process end to end. But now with AI and like how you operates and actually seeing through the entire case, life cycle actually becomes doable. And I think it'll change how law firms compete with each other and how people think about running a law firms to begin with. Wow.
OK, so case evaluation, how are you using it for intake?
So that's actually a really cool thing. So the new feature we actually announced on June 9th is around intake, right? So historically, the way to think about AI is increasingly AI is going to be able to do more and more. So today it primarily helps with understanding text, but what's coming soon is its ability to actually understand speech and actually be able to talk as well.
So soon, you know, what I expect a lot of law firms will be doing is actually allow AI. And I think call of recording is going to become a larger and increasing piece of how you practice law where you actually want to take that data that you're having during communicating to clients or taking new data in and new information about a case into your corpus of knowledge. Eve is actually going to do that for you. So when you have a call with a client and it'll basically work with all the different call recording softwares out there and your local computer in being able to take that information, upload a ride to Eve, and then immediately start delivering insights from that UPS information how it pertains to your case in addition.
So that's how it helps with intake. Of course, addition to that, the other really cool part Eve is given it can now talk, it actually will be able to handle overflow calls, right? So if you have overflow or after hours services you, you pay for or manage in some way or if maybe you don't do it all today, given the cost, you will actually be able to feel it for you. And it sounds shockingly like a real person.
It's probably still won't be as good as a real person, but given how AI goes it likely will improve like substantially over time.
So that's what is one of the things that's coming and the people talk about it. They is going to replace people. And So what we're literally saying is now it will be able to, might not be perfect right now, but it's growing, right? It learns and we're learning as we go and building it that it would be able to take and take calls, get the all of the information, etcetera.
What would you say to someone who would say there's a whole nuance about connecting with a client in a way that they would be that you sell them into hiring this firm and that happens. We've been an intake.
It's actually really fascinating, right? Which is why we have two components of this whole intake product, one which helps the existing staff and one which handles overflow and intake. In the one where it helps the existing staff, it lets you turn that into a science. Often times when you're training you staff, you're teaching them some principle you believe in, and so maybe you have an empathetic style of actively listening and signing your staff in terms of following a script that looks like that others tend to follow.
An alternative talk track is one that's very scientific, right? So you would literally keep iterating on, you ask this question, followed by this question, followed by this question, and you track conversion rates and funnels. Some firms operate that way. And I would say there's no necessarily right answer, but there is a certain salesmanship involved in terms of doing both understandlexling what the case is about and representing yourself well, and also not establishing, find an attorney a relationship right off the bat.
So there's like a whole set to it, right? What you've let you do is turn that into a science, even the empathy part. So what's really interesting is maybe this is a little risque to say, but AII think is capable of being more empathetic than humans over time. And basically I think what's going to happen is you can bring it to follow a particular type of style of empathy that's specific to you and your firm and how you sell.
And I think that's what's going to happen. What is interesting though, at least what I've noticed in using and also using other AI tools, is it actually helps you weirdly in connecting more with the people and worrying more about the people side of the problem, because less of your brain is actually working on the kind of drudgery of everything, the basics.
Of the record, everything so you're able to focus on. Yeah. And it's one of the things I see a lot when I'm helping to train someone in cells, especially with lawyers, right? So caught in their lawyer brains of trying to take down all the facts that they're missing the opportunity to connect on a deeper level with the client, which is really what's going to make them trust the lawyers.
I could totally understand and see, that's pretty fascinating. Yeah, that's amazing. But.
I think that's actually the really interesting part, right Lexlee? Like this is what I mean by you can tweak it per firm is some people naturally are more empathetic when you hire an employee to do intake. Some people connect really well, but they might forget to ask some pretty important piece of information about their connecting or maybe their time management isn't as good or what what have you number of different problems. This is where you can actually step in as you're talking to the client and because she is with you during that call, she can actually coach you live and give you suggestions on what to do, which is pretty cool.
Today, the process actually, some of our customers are already using it like that before we launched the voice and intake product, but some of our customers are already literally putting customers on hold and then asking Eve, hey, what should I do in this case? And then exactly start going back. So that at least is going to get a lot better. And some others might be a lot more scientific, right?
They might just be like, oh, let me just get this and this and not realize the other person is suffering and telling you something very personal and emotionally you're just moving on to the next thing. And just some other people need guidance there. So that's where like training and coaching the staff and being with them is going to become an increasing part of our life, I think not just for intake specialist, but let's see, I think everyone.
So the new part that you just announced is the voice feature?
Yes, the voice and intake. So holistically, Eve is going to help you with taking data in that is a voice in nature speech and often times it does have an intake and then second Eve is now able to field intake calls for you if you wanted to.
Wow. OK, so tell me what else is what are the ways that you their primary ways that you see that we're using AI so we know and that you've just given me open my eyes to something with intake. I know we've talked about the pre let part. Tell me about how it helps in litigation and prepping and all of that.
Absolutely. In litigation in general, like one of the most common places that's both time consuming, but one where Eve can help a lot is discovery. We're dealing with discovery responses. Sometimes you just don't fully control the timeline on these things, but then you have timelines imposed on you.
In these type of situations, Eve is a lifesaver. Many for customers literally are. They get discarded proponent on them and they have to respond and it's 40 hours of work. Eve can take that down to less than 5 or more.
I've never experienced it, I've never even seen it demo to me, and there are probably a lot of people listening who have not either. So how does it work? So I get a set of discovery and what are the practicalities of that? How does it actually work?
How do you do it?
This is actually the cool part of Eve is because you're working with it through the case life cycle. He actually remembers and understands what the case is about, what the files in the case are about. Especially if you have an integration set up with your CRM or your DMS of choice so there's no nothing really needed. She already knows everything about that case, alongside potentially strategies you're pursuing.
But you get the strategies that you're pursuing and what you're looking for to support each of those strategies.
That's right. That's right. And on top of that, like the, when you do have to respond to discovery, oftentimes it's if it's a form interpretory or if it's a, you have like a list of questions you have to go through and find supporting piece of evidence for and understand whether you want to answer it how you want to answer it. And this is something Eve is able to do because she has such a good understanding of the case, right?
She already knows for each question, do I have materials to support this question? If so, she will respond for you in your same style. You can also train it to respond like you would by giving some of the past discovery response you've done. And it actually just goes and does it 1 by 1.
And it goes through the materials that you have it'll and the responses that it gives you, it's going to tell you where it's pulling those from.
That's exactly right. That's an important part of how he works. So every time he pulls together the information, she will cite where she got it from and it will let you easily double check it. This is one of the biggest problems with AI in general, at least early stage of AI when things were so working but not fully working, it was very difficult to go figure out how do I trust it?
And sometimes you might spend more time checking the answers than like actually just doing it yourself to begin with. So one of the things that we made super, super easy and evil is citations and knowing exactly where she gets her information from. And what's cool is she does it even for things where it's not a direct citation, right? So let's say you're summarizing a large lot of documents and just getting one piece of insight out of it.
She's still capable of telling you exactly where she got that summary from. Potentially hundreds or thousands of documents you have.
I had a conversation with a lawyer about this a couple months ago and it's still is. I'm wondering about this. Even though it can do all that and it can tell me where it pulled it from, there's a part that do we still have an ethical obligation to go back and look to see if it actually is right?
In my opinion is that if you're the attorney producing any type of work, you should do the work required to get confidence in the quality of output. That's my generic high level answer. Now, what does that mean? Yeah, if the technology is there where let's say it's been inaccurate or if it's been not working well, you should learn the tool enough to know how do you go and check the sources and accuracy of those information.
And I think that's going to keep evolving, right? Where we were in AI two years ago or a year and a half ago even, because you probably had to go and check every piece of information it put out, which basically for most part negated using AI in the 1st place. If you want to have high degrees of confidence, Where we are now is hallucination rates are way down, right? So as a result, depending on what you want to do, you could start setting up instead like a blacklist, right?
These are the type of information that is really critical that I check. Could be a saw citation, could be, I don't know, names, addresses, could be other piece of important information that as a process you want to make sure you're training your attorneys to double check that information. And we've actually worked with our partners, a lot of our early customers and our existing customers in actually creating an AI guidelines. When you adopt Eve, we will give you, if you want how you should be thinking about rolling out AI and training your staff on what to double check what not to what the process should be to double check which is still safe and reliable.
And we actually give you that whole guideline. But I would say for most people today, if you're using a platform like Eve, you can probably get away just creating your own mini process for double checking the important pieces because Eve is pretty good and accurate enough at the others. I think it's probably at human levels of accuracy already. Like the mistakes it'll make are probably in the same level of realm of what humans will make, but given it is AI, it probably is worth your time to be able to check the things that are really mission critical for you.
He does. It makes it easier to double check.
Yeah. It's interesting that you say that it's like to the level of what human mistakes that they would make, because I was thinking as you were saying that if I'm a managing attorney, I still have to check my people. I can't just assume that everybody's going to get it right. And there are certain things that I still need to look, especially if, like, I'm training lawyers, certain things I still need to look that they got that they may not know.
Even if we do that anyway, it's not adding something. It's just learning to do it with Eve or doing it with the AI system is basically what you're saying, yeah.
That's exactly right. Eventually you're getting a feel for the system and then you develop your own understanding of where you can trust it, where you can ask honestly, how you on board people too, right. When you have a new employee to the firm you're trying to see, OK, first I'm going to check everything. And then after you, you have some level of confidence, you're understanding, OK, these are the places where they seem pretty solid.
I'm going to trust them. These are the places where they've made mistakes that are still have trouble and you start forming that mental model yourself and people are actually farther than technology, right? Because technology is going to make the same type of mistakes again. Again, people are a bit more complex with how they make mistakes.
Sometimes they regress in places they were good before and those are you have different problems with AI. So I think learning that initially is a little challenging for a lot of people, which is why we have this whole guidelines, right. So instead what's easier to teach is it's easier for people to learn processes in general, right? Let's say a simple process is just double check case law.
So it's easier for people to just learn that and you as a firm owner can have some level of assurances that from a worst case scenario, at least those boxes are checked and maybe it makes other mistakes. Those mistakes are within the realm of acceptableness. So I think that's like the easiest place to go right now. I think there is a more advanced one where some people do give out such a good understanding of AI that there I would say another level of competence with how fast they go.
And I think it's another part that's fairly important that you mentioned is that Eve is making it easy to check, say a little bit more about that.
Yeah. So like I mentioned, double checking is an important workflow because similar to how you're building trust with a person, right, That's how we view it too. We want you to be able to trust Eve and you're building features and improvements across the board to help you across to better and better, but also obviously do a better job over time as well. Eve itself doing a better job over time.
So those are the two metrics we push on continuously every single day. And what we actually do to check is we ask ourselves like, hey, let's say I want to check every single piece of answer that Eve spit out in this instant. How do I do that in the least amount of time possible? And that's what we rate on.
So for every single piece of information that Eve spits out, you can click a little button that comes out with the answer and they'll tell you exactly where it pulled it out, and they'll make the process for you of verifying and double checking that work really easy.
Yeah, it's just fascinating. Can you give us a specific example or a case study of how Eve helped affirm increase in efficiency and or revenue hopefully revenue too as a result of?
That absolutely. So what's really fascinating on average across all of our firms that are not does, which is now in the hundreds, people are taking on 40% more cases the next year than they did before with the same.
Number of team.
Same number of teams and sometimes less, right? Because you do churn employees over the year and that's what we're seeing across the board. I think within each case, within each law firm, there are some really significant outliers and some firms, it's actually really interesting how smaller firms react versus larger firms. The benefits smaller firms have is they have they can implement AI significantly more quickly once they get bought in.
They are able to train their staff much more quickly than a larger firm and make some bold decisions here on they see early data signs that they can handle more cases per attorney. They will start spending more the next month, right, which historically it's maybe a little bit risky to do. But with Eve they can dial up and dial down the knobs as much because our pricing model kind of scales with their case volume, not like a set amount for forever like a like it's not a salary. So that's one thing that the small offers screwing on.
I think one of them Eve Frontier Law have been he's an L&D. We've been working for a long time with him, but he's I think almost tripled his case volume and the time he's been with us. And then even larger firms there is more complex. So someone like Mike Morris, he has the largest firm in Michigan and there we are.
He's using us to drastically decrease the time it takes to get demands out and pre mitigation work done and they're adopting and mitigation and that is going to because it's a larger firm and he has understood attorneys. It takes longer to instill everywhere, but the Prelit team has been able to take advantage of that instilling. It goes team by team there and benefits are quite immense in terms of time to value. The other hidden benefit is because you're doing lots of your work on drudgery, you're actually able to spend more of your work on things that excites you in some way or things you're good at.
And what's subtle, But we haven't found a good way to keep track of this. So I don't have good numbers apart from what I hear from customer calls, but settlement rates are going up across the board because you're able to do more of what you want to do. And surprisingly, because if we're costly improving how we deliver our various different answers, it's going to make you at a certain baseline. So if you're below that baseline right now, it's going to least at minimum make you meet it.
And then on top of that, you can then use the stuff you're naturally good at to push the envelope even more. And that's something we're seeing in Frosted Board as well, which is really fascinating.
It's really fascinating. This is amazing. Thank you. This has been such a great conversation.
It could just go on and on, but let's wrap up. We always ask three questions of our guests at the end. And I think you might have already answered this, just like with your recent announcement, but what's one thing that you're excited about right now that you're creating?
Yeah, I think I did answer this. I'm very excited about being able to understand voice and to be able to also speak. That's the feature we've been working on for a while and people have been asking for a long time, and we have a very high bar for Jersey and making sure it actually works. So we've been iterating on it for a while.
I expect that we're now on the path to where God's getting better and very quickly that's similar to how large language models were. So I'm really excited about it and I'm looking at other modalities as well.
Wow, that's amazing. OK, second, what's one rule you would tell lawyers to break?
That's a really good question. I think ultimately a lot of lawyers still view their I'm as very precious, like how much can I do in an hour and how much value can I get in the hour? I think one of the things that seems obvious with AI is it's going to be a lot more about impact, right? So what's the maximum way I can actually improve the impact over a period of time that I'm delivering to either my clients or to the firm or what have you really across the board is another way of generalizing it.
So I think impact optimization is going to matter more than time optimization. They might be the one of the same depending on how you break it down, but I think that mentality shift is useful to get into.
So I was just thinking, that's a mindset shift.
Exactly.
That's huge. And finally, you're an incredible leader and really leading in your field. So we always want to ask people, what's something that you do is some kind of practice or habit that helps you to be a healthier, happier person, that helps you be healthier and happier in every role that you take on.
That's a really good question. I guess mine is a little more personal life based.
Like I haven't.
I have a kid and have a pretty challenging job where I'm working pretty late all the time. I think what I've found really valuable is like setting up routines with my family and child around connecting, but also getting time for myself that wake up pretty early at 5:00 and then actually work out every single morning with my wife. And it's been a way to connect with my wife. We enjoy more time together and we get healthier together as well.
And we have a little bit of time before before the day starts, which I found to make a pretty big impact on my mental health at least.
Yeah, I love that. I think morning routines are so important. But really, what you're saying is intentional time for connection with others and intentional time for connection with yourself.
That's exactly right. One of the geeks about trying to live longer. So I do a lot of like research and the random Googling around those concepts. And a lot of the people that live longer are typically they have a full emotional circle where they actually connect and they have friends and they have typically good diets and health practices.
The basis is probably most people are aware of. We just have to find a way to do it. That's always the hard part is doing it.
Yes, and those factors are for living longer, but they're also for living happier.
That's true, actually, very true.
This has been so wonderful. Thank you so much for your time and all of your wisdom, and we would definitely have to have you back as things continue to change and grow.
It's been so fun, Lexie. Thank you for the really thoughtful questions. I'm happy to come back anytime. Thank you so much for inviting me.
You're welcome. Thanks for listening to today's episode of Mind Over Law. We hope that you're walking away inspired and ready to embrace your life and law practice in a more holistic, healthier, happier way. Don't forget to hit subscribe and let us know what you think.
Leaving a review helps us to spread the word to others. Be sure to follow us on socials for all the behind the scenes fun and sneak peeks of upcoming episodes. Until next time, remember with intention anything is possible.
