AI & PI Part 3 with Adam Greene
- 22 November, 2023
- 45.9 MB
In this insightful episode of the Justice Team Podcast, Bob Simon sits down with Adam Greene, an engineer specializing in AI products at Filevine. Their discussion revolves around the innovative applications of AI in legal practice, highlighting both statistical and generative AI’s potential. Adam introduces several of Filevine’s AI-driven tools designed for the legal domain, such as Lead Docket, which predicts fees and summarizes leads. Adam also underscores the critical need for legal professionals to embrace AI tools to enhance operational efficiency and stay competitive in the rapidly changing legal industry.
Adam Greene, Filevine
Transcript
Bob Simon (00:08):
Welcome to this episode of the Justice Team Podcast. You can find us at the Justice Team Network on the Apple Channel or on YouTube or just a website. And today we’re going to continue our segments in AI and PI. And we are so blessed to actually have an engineer here who actually builds the AI who’s passionate about the product, who is currently at Filevine. So why don’t you introduce yourself and then we’ll get into the nitty-gritty?
Adam Greene (00:30):
Hey, my name’s Adam Greene. I’m a developer over at Filevine and I work on the product and also the new AI products that have been coming out over the past year. And really just passionate about technology in general. I’ve been building software in different industries for 20 years or more. And now the coming trend of AI, it’s really exciting to be able to take all these new use cases and really bring this technology to bear on new problems.
Bob Simon (00:59):
So, if you could just summarize for our listeners, what is AI and how is it applicable to legal practice?
Adam Greene (01:05):
So AI, actually, I split it up into two segments right now. There’s the traditional, more statistical AI, where you’re looking at big sets of data and statistically trying to figure out how to predict things, cluster things or do things in that manner, trying to predict the outcome of something. Lately, we all have seen LLMs have come to the forefront, and that’s a little bit different. It’s all AI but basically-
Bob Simon (01:31):
What’s an LLM?
Adam Greene (01:33):
LLM is a large language model. So something like ChatGPT, that’s one type of LLM. Google has an LLM. Facebook has launched open source LLMs and all of that, and that’s called generative AI, meaning it generates content or it can look at content and extract from that content. And so, both of those areas are ripe for use in many fields, but especially in legal, for a lot of reasons that we can talk about.
Bob Simon (02:03):
And I see you’re also a girl dad. I can see we have matching bracelets that our daughters make for us.
Adam Greene (02:07):
That’s true. This is for my daughter. My wife also has breast cancer. We’re getting through that right now, so she has a little campaign going on where she’s building these pink bracelets and getting them out to the community.
Bob Simon (02:19):
We’ll be sending good vibes to you and your family.
Adam Greene (02:20):
Thank you.
Bob Simon (02:21):
And it’s something crazy, where it’s like one out of eight women are diagnosed with breast cancer.
Adam Greene (02:25):
One out of eight, yeah.
Bob Simon (02:25):
It’s crazy.
Adam Greene (02:27):
And younger. It’s starting to be younger and younger.
Bob Simon (02:30):
And they’re also … well, we won’t divert too much, but now they’re using AI in medicine to help predict these things faster. So, it could be used for good, it could be used for bad, it could be used poorly. So let’s talk about specific use case, because a lot of lawyers listen to this show, and I know I’ve been able to vet a few of these products. I’ve used the AI fields and also the AI on the deposition, a contradiction model on our platform that we have. So, let’s just talk about different … where it’s going, because I think what the mistake a lot of lawyers make is they think they could just go to the language model, the ChatGPT4 and go in and just ask questions and think it’s just going to be verified. But that pulling from Reddit on the internet. It could be anywhere.
Adam Greene (03:12):
Yeah. Wikipedia, basically just huge corpuses of information that’s out there.
Bob Simon (03:18):
All right. But the products that you’re building are specific for specific things and legal?
Adam Greene (03:23):
Tailored to the legal space, the legal information domain, and then the legal workflow. If we take a step back, the way I look at it is there’s a workflow of a case that goes through a law firm. It comes in, it’s a lead. Things happen on the lead front, then it goes in, it becomes a case. If the case is signed, things happen on that front and then it goes through a new bill and there’s all the invoicing on the backend. So, what we’re doing is taking that workflow and we’re finding the places there’s so much to do right now. We can bring AI to bear on so many different places. We’re trying to prioritize and figure out what are the highest value areas that we can address right away? So, on the lead side, we have Lead Docket and we’re building a model now to help predict fees when they come in the door.
Bob Simon (04:14):
It’s crazy. And I was talking to Josh who was … Did he found that product? Lead Docket?
Adam Greene (04:21):
Did he … sorry?
Bob Simon (04:22):
Was he the founder of Lead Docket.
Adam Greene (04:23):
No, Eric Coffman was the founder. Josh was a client who was working-
Bob Simon (04:26):
Wow.
Adam Greene (04:26):
… at a law firm and then came on as our product owner because the fit was just so great, he understands it so well.
Bob Simon (04:31):
He markets, so so well. Yeah, and correct me if I’m wrong, but where it will start to go is that as the leads come in, you can start to predict the value and if it’s a yay or nay case potentially for you.
Adam Greene (04:42):
Right, so we look at millions of data points that have come in into Lead Docket, and we can use that and build … this is the statistical model. This is not using an LLM or generative AI. Build a statistical model of what that looks like. And then when a new lead comes in, it can match up the data and say, “We’ve seen this before, and based on what we’ve seen before, we predict that this is likely, highly likely, to turn into a lead, or unlikely, highly unlikely,” or sometimes unknown, just don’t have enough information. That’s where we’re starting in the future. And we’re starting with motor vehicle accidents, that’s the most [inaudible 00:05:17]-
Bob Simon (05:17):
Yeah, there’s probably 10 to 15 data points you could do quickly that’ll be able to qualify, right?
Adam Greene (05:23):
Yeah.
Bob Simon (05:23):
Okay. So if you go through that, and that’s actually … I mean, it’s a huge lift. It’s harder to predict the value of higher … catastrophic injury cases with larger policies. That’s obviously a case that they’re going to take and explore more. But this is a good product. I would assume for listeners, a quick yay or nay, especially the nays take so long to sort through. It could really help you be more efficient.
Adam Greene (05:46):
That’s where we are now. There’s interest in figuring out if we can predict value buckets to say a lead coming in is going to be from zero to 5,000, 5,000, 10,000 or whatever. So that we can start to use that to allow firms to forecast and just know, “Oh, this is going to be around this size. Is it something I want to address or not?” Because there are different business models that different firms take. They take different levels of cases. And so, knowing that information would be great if it can be done. I’ve heard word out in the market that some of our competitors are coming out with things like that. So I have a feeling that’s going to be something [inaudible 00:06:24].
Bob Simon (06:23):
But the struggle with those is always you need so much data on what the case is resolved for and all these other data points.
Adam Greene (06:30):
Jurisdiction.
Bob Simon (06:31):
Yeah, it’s all different, but a lot of-
Adam Greene (06:33):
That’s definitely much more difficult than just predicting fee or not.
Bob Simon (06:36):
And for this, the intake AI what’s the intake AI call that you’re coming out with
Adam Greene (06:41):
Lead Docket?
Bob Simon (06:42):
Well, Lead Docket is-
Adam Greene (06:43):
Oh, it’s called LeadsAI.
Bob Simon (06:45):
LeadsAI. And how far away do you think we are from being able to access that?
Adam Greene (06:48):
That is going into beta as we speak, and it’s two parts. So it’s the prediction, the fee prediction, yes, no. And then it’s also summarization of leads. So, a whole lead comes in and right away within a few seconds, you get a summarization right there. So, you can take a quick look at what this lead is and help you look at … So it’s a package of those two pieces of functionality.
Bob Simon (07:10):
And that’s a very good product, especially for people that are getting a lot of different calls, a lot of different leads but how are they … So this is, for those of you don’t know, there’s a product called Lead Docket where you can get leads directly either from a digital marketing campaign ad that auto-populates, or if somebody takes the call and manually enters the information. Is there any other way that it goes into Lead Docket?
Adam Greene (07:33):
It can come in off the web, it can come in from phone. There’s an automatic phone call hookup-
Bob Simon (07:37):
Oh, cool.
Adam Greene (07:37):
… that it can come in through. There are other ways to calling the office, typing it in by hand, but again, it’s hooked up to all the digital channels that are out there, that leads could come through.
Bob Simon (07:50):
And then in the PI space, which a lot of our listeners are in, it’s like if you find out what the policy limit is, sometimes that can be the value of the case pretty quickly, right?
Adam Greene (08:01):
Yeah.
Bob Simon (08:02):
So that’s an exciting product, especially for firms like ours, which we’re getting brought in a lot of cases. If we can see how leads are scored when they’re given to us, it makes it a lot easier. Especially we’re getting a lot of cases call in and we’re like a 90% rejection firm, but we have to find potentially another firm to partner with. It’s a good tool.
Adam Greene (08:19):
And the model learns that from the data because we have data on all different kinds of firms. So, it knows if you’re a high volume firm, but maybe you have a low conversion rate, or if you’re a low volume firm with a high conversion rate.
Bob Simon (08:31):
Wow.
Adam Greene (08:31):
It takes all of that into account as it’s figuring out statistically, should this be a lead or not? So it’s pretty advanced and pretty cool technology.
Bob Simon (08:41):
And I know we’re working on our integration with the Attorney Share marketplace where it can go directly through Lead Docket. And if you don’t take that case, you can push it to somebody that specialized. And that’s where it can get exciting. I mean, imagine having your AI predict who’s the best specific lawyer in the nation to monetize the case.
Adam Greene (08:59):
Right. Yeah, I mean, we haven’t even thought about that. That’s the kind of stuff that comes to us every day, all these things.
Bob Simon (09:04):
Yeah, we have workflows [inaudible 00:09:07].
Adam Greene (09:07):
Automations, you don’t even have to touch the lead, right?
Bob Simon (09:09):
Nope.
Adam Greene (09:10):
If you get comfortable with the prediction. And there’s going to be some time where clients, I think have to get comfortable that the predictions that it’s making are the right predictions, which they will over time, but we understand that. But once you get comfortable, why not hook up an automation that says, if this is a highly likely lead, sign it, send out a Vinesign contract, have them sign it, come in and push them right into Filevine, and you’ve got a case
Bob Simon (09:31):
And if it’s not, if it’s outside your wheelhouse or too low for what you take, it goes somewhere else and you can split the referral fees. That’s what we’ve trying to create. So, that’s to walk us through AI fields. I know this is a product that we’ve test driven and do like to use. So, walk us through how that works.
Adam Greene (09:46):
So, from the workflow perspective, you’ve got those two products on the lead side, the fee prediction, the summarization, then you do all your stuff in Lead Docket, comes into Filevine, and then you start aggregating documents, uploading documents into the platform around the case. AI Fields allows you to build in automated intelligence around extracting data from these documents. So if you-
Bob Simon (10:08):
It’s extracting documents from your own case management software that already currently exist there on the cloud?
Adam Greene (10:12):
Right, and it’s a workflow piece. So, you say whenever a medical form comes in or whenever a police report comes in, on the medical form, get me the CPT codes and get me the treatment information. Pull that out, stick it in the field, and then it’s right there in front of you. Police report, similar stuff. And then it can also look for discrepancies. “We had in our summary of the case that this happened, but the police report said this, there’s a difference here. Focus in on that.” And it can bring that to people’s attention much more quickly than a human reading through all this stuff and trying to figure out what’s-
Bob Simon (10:45):
And that’s a product that’s available now for its users?
Adam Greene (10:48):
Available now, yes.
Bob Simon (10:50):
So how far are you away? Are you going to do a ask me Anything within … have a bot that asks you anything within Filevine [inaudible 00:10:59]?
Adam Greene (10:59):
So we have something called Sidebar AI, which is that. So I think that piece, that chatbot type thing, is going to be one of the biggest pieces of AI that comes into legal tech. Having this kind of copilot that sits next to you and you can work with, we’re building that, and right now it’s helping with support and helping do some basic things around the case. It takes time to figure out how these things work, and that’s the generative part, using the LLMs.
Bob Simon (11:26):
But that one seems like it’s such a … for the firms that have so much data in there and so many cases, that seems like a very hard lift for you to build. Again, I’m not an engineer, but that seems really difficult.
Adam Greene (11:38):
The challenge is around making sure that the LLM, the AI, is looking at the right data. Because we’ve all heard of these things called hallucinations where … maybe not?
Bob Simon (11:51):
No, explain it for … yeah.
Adam Greene (11:52):
Hallucinations are basically when you ask it a question and it comes back with something that makes no sense and is just completely out of the blue. And that happens because when it doesn’t know the answer, it’s trying to answer you with something. So, it finds the closest thing it can find and spits it back.
Bob Simon (12:08):
Interesting.
Adam Greene (12:08):
So, we’re doing the work to figure out how do we handle those type of things in the legal universe? And that’s part of the hard work that goes into building one of those applications.
Bob Simon (12:20):
I assume if there’s a hallucination event, you just ask another question and get back on track?
Adam Greene (12:24):
Sure, sure.
Bob Simon (12:25):
Because that’s an exciting product, I think for a lot of Filevine users, how far away do you think we are?
Adam Greene (12:30):
It’s close. I don’t know exactly when launch is for that, and then it’ll be built on over time.
Bob Simon (12:35):
Because I always wanted to have something. Imagine you’re sitting there in your case management platform, and I get this all the time, I’m trying to find a deposition of somebody. I remember the name of the person, but I don’t remember the case or anything else, and I need to get somebody a link to that because they need it. Just ask this, your assistant, to get you there.
Adam Greene (12:54):
Yeah. I mean, Microsoft is coming out with … they’ve got a whole copilot strategy, but supposedly they’re integrating a copilot into Windows. So, you’ve got this thing that just sits next to you and you can interact with and ask questions, and it remembers things about you
Bob Simon (13:07):
That’d be sick. Right now, I just use that search bar in my Windows to find stuff, and then I don’t remember … I still don’t remember what I named it, and then it’s a problem. So, that’s an exciting product. What else do we have on the horizon? We have AI blocks, what’s that one?
Adam Greene (13:20):
AIBlocks is for when you’re building contracts and putting documents together, it allows you to use one of these LLMs to … Let’s say you want an indemnity clause with a certain twist. You say, “Oh, give me this.” And it shoots back and it gives you some content, and then you have to review it, obviously, and you have to make sure it’s right.
Bob Simon (13:37):
And for all of our listeners don’t fall victim to just boilerplate, signing off on the product. I give it the 80/20 rule, gets you 80% of the way there. You still got to finish it off.
Adam Greene (13:46):
We’ve all heard of that case wherever it was, where the lawyer just took it to the judge and reads.
Bob Simon (13:51):
Idiot and then doubled down on it too. What an idiot.
Adam Greene (13:55):
So, that lets you build contracts using a LLM to back you up and help you gather the information, come up with new copy.
Bob Simon (14:06):
Where is that pulling from? Your own data set that’s within your case management or somewhere else?
Adam Greene (14:11):
Both. So yeah, you can build up content in … So we have a product called Outlaw, which is our document creation.
Bob Simon (14:18):
Oh yeah, so you guys acquired them last year?
Adam Greene (14:19):
We acquired them. And inside there where you’re building out your contracts, there are different pieces that you can use. You can build up a library and you can build up a set of prompts and basically use it to help you generate contracts in a much more efficient way.
Bob Simon (14:34):
Wow, I forgot about that Outlaw acquisition. And then there’s the DemandsAI and ImmigrationAI, right?
Adam Greene (14:40):
DemandsAI. Before we get there, there’s one called AI Document Review, which is similar to AI Fields, but AI Fields is something you build into your workflow. Whenever a new case comes in and it has these types of documents, run the AI on it. A document review is a personalized chat with your document. Maybe chat is going too far at this point, but you can ask questions of your document and you can say, “Oh, I’ve got this specific document, I want to know this.” So, I send a prompt to it and it comes back and it gives me the information. And you can drill into that thing and then have it there right inside your case and-
Bob Simon (15:14):
Okay, so I find the biggest struggle is having lawyers adopt use of this. It’s like literally pulling teeth sometimes. How do you get them to adopt and use this stuff?
Adam Greene (15:29):
Well, I think that’s the benefit, and that’s one of the key things that Filevine brings to the table is, it’s all in one place.
Bob Simon (15:36):
It’s already there.
Adam Greene (15:37):
And it’s part of your workflow. And we have a user experience team, which focuses on making it easy for the user. So, I think in the past, the hardest thing was just getting law firms to digitize and to use systems to help run. If they can get over that, take that leap, which I think in general, the market has taken at this point in the last five years, now-
Bob Simon (15:58):
You would hope so.
Adam Greene (15:59):
… you would hope so, right? Now you have a platform where we’re doing the hard work of bringing those tools to you in a way that you can use them easily, and hopefully that will get you over that bridge, across that bridge.
Bob Simon (16:10):
And then how does it work with the price point of using these AI tools? I don’t know if that’s probably not your department though.
Adam Greene (16:17):
Yeah, I don’t know how pricing works.
Bob Simon (16:20):
Yeah, I’m just wondering if it’s a per use type thing, because for me, the concierge bot would be life-changing, to have something like that.
Adam Greene (16:28):
There are different pricing models, something, and this isn’t anything from Filevine vision or anything, but just as a technologist, I feel like these LLMs, the generative AI is going to get, like every technology, is going to get less expensive rather quickly.
Bob Simon (16:45):
Yeah because there’s more use, more data, easier.
Adam Greene (16:46):
Well, it’s kind of like the technology has been built and that first version, it’s not … This has been being built over years, but that first version of these LLMs came out and it’s great and it’s a paradigm shift and it’s a breakthrough, but people haven’t done the work to optimize those things yet. And now, you see all these things coming out every day where they’re optimizing.
Bob Simon (17:07):
All this works. I mean, there’s been so many companies with … AI companies that started in the past six months.
Adam Greene (17:11):
Oh, every day something new comes out, new models, new companies, everything.
Bob Simon (17:16):
So give us, Adam, run us down some of your background just to understand how you get professionally where you are.
Adam Greene (17:22):
So, I’m actually a software developer by trade. I’ve been writing code and I’ve been consulting for years, going out to companies, working with them, learning new industries and building software to solve business problems. That’s what I’ve always done. I have a computer science degree, so I learned … I know programming and all of that. And then just in the last few years, I’ve become interested in AI, so I’ve started learning that.
(17:49):
And then honestly, these LLMs and this generative stuff just came out of … It’s been building an industry for quite a while, but in the marketplace it came out of left field and all of a sudden there are these new use cases that can be built with much … Number one, those use cases weren’t even available, but now you can build them and the effort to build them is way lower than in the past, you’d have to build your own model, which would take forever. Now you use someone else, large company has built it, and you just use it.
Bob Simon (18:19):
Wow. Then how long have you … When did you come to Filevine? What’d you do before that?
Adam Greene (18:25):
Been with Filevine for two years. Before that, I was doing a mix of consulting and software development.
Bob Simon (18:29):
Cool, man. Yeah, I was wondering because it’s like … So when you meet people to build product and stuff, but just I’m always interested of how long you’ve been doing this stuff, because to me, it just seems so technical and overbearing for me is it that-
Adam Greene (18:45):
For me, the law is the same way, I-
Bob Simon (18:47):
Imagine marrying those two things together. So yeah, the other products DemandsAI and I assume if you’ve got Fields AI that you can just take what it can draw directly from Fields AI and then automate your demand letter?
Adam Greene (18:59):
Yeah, yeah. It basically makes creating demand letters much more efficient using mix of AI and services and all of that. But yeah, you just send your demand letter out and you get it back much more quickly than you would-
Bob Simon (19:11):
Wow.
Adam Greene (19:11):
… if you had it in-house or anything like that. And then there’s Immigration AI, which is using AI to help bring the lengthy, document heavy, data heavy immigration process, and just take the pain of gathering and entering, and entering, and entering the information into the system using machine learning to do that.
Bob Simon (19:34):
Wow. I mean, those automations can help a lot of … because immigration lawyers who I think are doing God’s work, it doesn’t pay as well as it should. So they have to be able to be super, super efficient on the backend to be able to help. Any other products that you want to go over while we sit here? I think we covered a lot of stuff that’s coming out.
Adam Greene (19:54):
Yeah, I mean specific products that we’re building now, I think we’ve covered, there are other things on the cooker, but it’s too early to talk about it. It really is a latent space right now. I think everything is up for grabs in terms of where you can apply AI.
Bob Simon (20:09):
Wow.
Adam Greene (20:10):
So, I think what we’re going to see over time is these point solutions are being built to solve specific problems. And at Filevine, we’re trying to build it into a coherent workflow also. But over time, it’s going to start to consolidate-
Bob Simon (20:24):
Consolidate others.
Adam Greene (20:25):
… either inside a company or M and A.
Bob Simon (20:27):
Well, yeah, I mean, it should be to the point where having all these different AI tools built into one ask that it can go pull from, right?
Adam Greene (20:37):
Right, yeah.
Bob Simon (20:37):
That would just be … I feel like we’re only a year away from that, would you say?
Adam Greene (20:41):
Yeah. I mean, I call short term as we’re building these solutions for specific use cases. Medium term, whatever that is, one to three years, we’re integrating it in. And my personal opinion, not speaking for Filevine here, is I actually think the interface for software is going to change over the long term. I think that instead of going in and clicking and using a mouse and typing, you’re going to start … and I don’t know what it’s going to look like. Whether you’re talking, like Alexa, you’re talking to your application, or you’re maybe typing, but you’re asking it questions and it’s giving you answers.
(21:13):
So, for reports like today, you have some canned reports and you have to go in and filter and set the fields and blah, blah, blah, run it. Wait, I want to say, “Tell me about my top 10 leads in the last three months at these companies,” and boom, it just comes back to you [inaudible 00:21:29].
Bob Simon (21:28):
Or, “Hey, draft me,” and I email to these three people and send them my schedule for the next two weeks.
Adam Greene (21:33):
Yeah. I think long term that that’s where software, not just legal, software in general is going.
Bob Simon (21:40):
And if you had one piece of advice for the lawyers or law students that are listening to the show or watching, what would you tell them as it pertains to this industry?
Adam Greene (21:48):
Well, it’s funny, and I’m paraphrasing here, but our CEO Ryan Anderson always says that AI will replace lawyers, and then he waits and he says, “Who failed to adapt.”
Bob Simon (21:59):
Yeah, I saw that keynote that he gave in Utah, yeah.
Adam Greene (22:01):
And I think it’s true. I think lawyers have the opportunity now to become much more effective because they can focus on higher value add tasks, but the ones who do it the old way are going to lose out to the ones who are using new tools to make them more efficient just on an economic level, right?
Bob Simon (22:18):
Yeah.
Adam Greene (22:18):
You’re cutting your costs of doing any task, your margins are going to be higher, and if you’re doing it the old way, you’re not going to be able to compete with those other types of firms.
Bob Simon (22:27):
That’s true. So start adopting out, and I think easy for our listeners, if you start to get on any case management platform, please, there’s one thing you do tomorrow, make sure you’re on it because these workflows are coming. It’ll make your life a lot simpler.
Adam Greene (22:41):
But always Filevine first.
Bob Simon (22:43):
So, that’s where you can pull right now the AI Fields, and I’m excited for the LeadsAI myself, because that’s where it gets interesting for me.
Adam Greene (22:52):
Super upfront, knowing, being able to [inaudible 00:22:56]-
Bob Simon (22:55):
Because for me as a trial lawyer, I’m seeing either the very end, or the very end of the case, the very beginning of the case, not so much the middle. So, those ones were very attractive to me. They’re pretty good model at the end. And then what it could be at the beginning is very good.
Adam Greene (23:06):
Great.
Bob Simon (23:08):
Well, Adam Greene, thank you for coming on. We appreciate it. How can people reach out to you if they have any other questions?
Adam Greene (23:14):
At Filevine, my email, [email protected]. That’s probably the best way to reach me.
Bob Simon (23:20):
Green E at the end, or no E?
Adam Greene (23:22):
E at the end, yes.
Bob Simon (23:22):
Adam Greene with e at the end filevine.com. So, any questions, go to justiceteamnetwork.com. If you have any case questions on case evaluation or seeing the other shows, or go to the Apple Podcast channel and look forward Justice Team Network. Thank you, Adam Greene from Filevine coming on.
Adam Greene (23:36):
Thanks for having me, Robert.
Bob Simon (23:38):
Easy peasy.
Adam Greene (23:39):
All right.
