This week, we welcome back Serena Wellen, Vice President of Product Management at LexisNexis Legal and Professional, to discuss the newly launched Lexis Protégé tool. This advanced AI assistant is designed to enhance legal professionals’ workflows by personalizing responses based on individual user profiles, including practice areas, jurisdictions, and document management systems. Wellen emphasizes that the future of AI lies in personalization, enabling Protégé to understand user habits and preferences, thereby improving its utility in legal settings.
Serena elaborates on how Protégé collects user data and integrates with existing document management systems (DMS). The setup process includes heavy lifting from LexisNexis, ensuring that the AI understands user roles and permissions, and can access relevant documents seamlessly. She stresses the importance of transparency and user control in the personalization process, allowing users to opt in or out of certain features as they see fit. This approach ensures that the use of Protégé aligns with the unique needs of individual law firms and practitioners.
Integration with Microsoft 365 applications like Word and Outlook is another significant feature of Protégé. Wellen explains that legal professionals spend a considerable amount of their time in Microsoft tools, and the AI’s ability to enhance productivity within these familiar environments is crucial. With features such as smart drafting tools, automated legal requests, and contextual awareness of user actions, Protégé aims to streamline workflows and reduce the time spent on repetitive tasks. The seamless connection between Protégé and Microsoft applications represents a shift toward more efficient legal research and document drafting processes.
Serena also addresses the pressing concerns of security, privacy, and data protection in AI applications. LexisNexis prioritizes the security of customer data by utilizing advanced encryption and private cloud infrastructures. Importantly, she clarifies that customer data is not used to train their models, maintaining confidentiality and trust. By ensuring that users have control over their data and how it is used, LexisNexis aims to alleviate fears surrounding the adoption of AI technology in the legal sector.
Finally, she shares insights on future developments for Protégé, including the incorporation of voice commands and horizon scanning features, which will further enhance legal research capabilities. As AI continues to evolve, Wellen envisions a future where legal professionals can engage with technology in more intuitive ways, allowing them to focus on the complexities of their work without being bogged down by manual processes. Overall, the conversation highlights the transformative potential of AI in the legal industry, underscoring LexisNexis’s commitment to supporting legal professionals with innovative, personalized tools.
Listen on mobile platforms: Apple Podcasts | Spotify | YouTube
Twitter: @gebauerm, or @glambert
Email: geekinreviewpodcast@gmail.com
Music: Jerry David DeCicca
Transcript
Marlene (00:01)
Welcome to The Geek in Review, the podcast focused on innovative and creative ideas in the legal industry. Live from New York, it’s Marlene Gehbauer.
Greg Lambert (00:09)
And live from Houston, it’s I’m Greg Lambert and today we have with us returning actually is Serena Wellen who’s vice president and product management Lexis Nexis legal and professional and She’s returning to the show to talk to us about the newly launched Lexis Protégé tool. So Serena, welcome back to the geek and review
Serena Wellen (00:34)
screen to be back.
Marlene (00:37)
Yeah, so Lexis Protégé is described as a highly personalized AI assistant that is tailored to each legal professional. So can you elaborate on how Protégé customizes its responses and workflow suggestions based on individual lawyers or firms habits and preferences?
Serena Wellen (00:57)
Sure. So our vision for AI at Lexis is that it will become increasingly more useful as it becomes more personalized. So as we like to say, the future of AI is personal. And Protégé uses basically three methods for personalizing its responses. The first is a user profile that contains multiple attributes about a user, starting with things like a practice area.
jurisdictions, subscribed products, formatting rules, even preferred writing style. And then second, the user’s context. What are they working on? What’s the conversation that they’re having in the product? And then the third thing is the user’s own documents. So either through a connection with their document management system, their DMS, or via document upload. Eventually,
The user’s profile is going to include the user’s past behavior in Protégé, such as past documents viewed. And when Protégé appears within our other product offerings, such as our Microsoft 365 add -ins and apps like Lexis Create Plus and Lexis Connect Protégé will know the context of what the user is doing in that other product and respond accordingly.
Greg Lambert (02:23)
So Serena, that, especially like the user profile, is there a lot of upfront work that is done during the initial phases of bringing in Protégé into law firms so that it kind of, you know who the user base is and you, I’m assuming like through the DMS and other ways, you kind of build that profile, is that how it works?
Serena Wellen (02:54)
Generally, I would say yes, we come in knowing certain things about our subscribed users based on the setup that happens for a new account that’s subscribing. And especially if you’re connecting in your DMS, there’s work that we do to ensure that we have the right connection to the right documents and all of the roles and permissions and.
all of the other governance is set up in the document management system is transferred so that we know that, appropriately. So we, just want to say here that we do a lot of the heavy lifting when it comes to, setting up these profiles and then it’s, it’s up to the customer and the users to determine a, do they want to opt into the system? Do they want to use personalization? Do they want to use their proposed profile or not? And then, you know, what, what.
types of things do they want included because it’s not an all or nothing proposition. It’s really important when we talk about personalization to understand that this needs to be transparent to our users and they need to control it. Otherwise, it doesn’t work.
Marlene (04:05)
are some of the personalization options? Can you kind of give us some flavor on that?
Serena Wellen (04:10)
Right. So one example I’ll give is, is drafted my style. So, I was, I was an attorney for almost a decade, I think maybe more, before I came to Lexis. The main thing that I did was draft documents. I was a litigator. I also spent time on the other side doing transactional work, but either way, most of my time was spent drafting documents or preparing to draft documents, by, you know,
finding starting points, looking for clauses, et cetera. And the thing that I felt was special about my work and that I spent the most time on was my own drafting style, how I wrote, my voice. I put a lot of time into drafting and redrafting so that the arguments were as persuasive as they could be. They were clear and that they represented me as an attorney.
And so this is something we’ve heard from customers and we want our, our profile in Protégé to be able to use samples of the, the users writing style and apply those to the documents that we’re generating on their behalf. So that’s one example of what we’re going to do with personalization.
Marlene (05:35)
And when you bring the, you had mentioned that I guess in the next iteration you’re going to be able to sort look at past documents. So I take it that the Protégé is sort of learning from that in terms of what was done in the past.
Serena Wellen (05:52)
So yes, the idea of being able to look at a user’s behavior, past behavior, past work, and leverage that and base the response upon that, I think makes a lot of sense intuitively. And so I think people will understand and actually expect, as we’ve talked to users over time, that we will know them. We will know their work. We will know what they need. And they don’t have to tell us the same thing every time.
they log on to the system, which is really frustrating.
Greg Lambert (06:25)
Well, I’m going to dig a little deeper into the DMS here in a minute, but I do want to kind of talk about the integration with Microsoft 365. And so from my understanding is, know, Protege integrates with, you know, Microsoft Word, Outlook, Teams, Co -Pilot, with the idea of enhancing the workflow for the legal professionals.
Do you mind sharing just a couple of use cases where, you know, attorneys would use this integration with Office 365 and how you envision that impacts their productivity?
Serena Wellen (07:10)
So I’ll start by saying Protégé meets our customers where they want to work and for. Yeah. Well, people really want to work in outlook. Yeah. that’s a sort of a, the, more of a metaphysical question, but, but, we do know that our customers spend, you know, upwards of 80 % of their time in Microsoft, in some cases more.
Greg Lambert (07:16)
Which is an outlook. Word.
Marlene (07:20)
now.
Ha ha ha ha ha ha ha ha ha ha
Serena Wellen (07:40)
So, and I think there’s a reason for that. I’m a big fan of Microsoft. We worked very closely with them in our partnership and I think they’re doing some amazing things right now that is making what they refer to as the modern office or modern work really, really possible. And I would refer to it as enlightened compared to how things used to be maybe as recently as five years ago.
So what we’re doing because of that is embedding Protégé within our add -ins and apps that, that customize Microsoft applications like word for legal professionals. So what this does is it gives you powerful drafting tools with functionality like smart redlining or clause comparison, proofreading and
Our comp in our comprehensive drafting add-in which, which is called Lexis create plus, and you’ll have that alongside all of the generative AI capabilities of Protégé. So it gives you tools also for automated legal request intake matter management and collaboration in Microsoft teams and outlook with our MS teams app for corporate legal, which is called Lexis connect.
Also alongside all of the generative AI capabilities of Protégé. So you can kind of see what I’m, what I’m sketching for you is that in our, customized for legal Microsoft based tools were, which are already very powerful and workflow oriented. We’re bringing in this context aware now generative assistant that can, that can and does.
know your data and know Lexis data. So another point I think I’d like to make here though, is that Protégé again is contextual. So if you’re in Lexis Create Plus, it’ll interact with the canvas in Word. If you’re in Lexis Connect, which focuses on matter management, it will interact with your matter data within the application. And I think what’s
very interesting to me in this space is that there’s other Microsoft data available to Protégé. So if you think of Protégé as contextual in the Microsoft space, it can look at and explore data in the Microsoft graph, which is so rich. There’s so much there about each of our users and Protégé will be able to leverage that.
Greg Lambert (10:31)
Just to call back on something that you had mentioned before where you want to kind of write in your style. When you are in the Microsoft products, because I can tell you one of the first things that one of, actually more than one of my attorneys said when AI came out was, this would be great. It can learn me and it can write in my voice. so when I go to…
respond to an email, it can basically have an outline or a draft in my voice that I could quickly edit. Is that something that you think this combination with Copilot and Protégé that I could tell my attorneys that this is now possible?
Serena Wellen (11:24)
Yeah, I think that this space where we are building a custom engine copilot for Protégé. So we’re bringing Protégé into copilot. And this means that you can leverage the best of both worlds. So you’ll be able to use copilot and all of the knowledge about say drafting your emails that copilot knows.
You’ll also be able to leverage what Protégé knows and all about your legal data and about Lexis data and your profile that you use with Protégé, which includes examples of your legal writing style. So maybe your email writing style is, let’s say it’s professional, but it’s not the same kind of voice and the same kind of cadences and language that you would use in an appellate brief, for instance.
or especially something like a contract or a clause, very different types of styles. We all have so many, right?
Greg Lambert (12:30)
Do a little code switching on the type of writing that we’re doing.
Serena Wellen (12:32)
Exactly.
Marlene (12:35)
I want to keep this theme of integration going. So the integration of Prodigy with the firm’s DMS, that obviously is like a game changer. And it seems that the acquisition of Henchman earlier this year seems to springboard. So how does Prodigy use the firm’s DMS? How does it access? How does it use the firm’s DMS to provide insights?
and what specific tasks or challenges does this capacity solve for legal professionals? And will it help me make PowerPoint slides?
Greg Lambert (13:13)
Answer the last question first.
Serena Wellen (13:15)
yeah. Yeah. Yeah. well out of the gate, right? not right. Yeah. I hear you. I make a lot of PowerPoint slides myself out of the gate. Not yet is what I would say on PowerPoint. that’s something that, we’re to have to work on a PowerPoint and then I think we’ll, we’ll test the, the enthusiasm for Excel as well. That’s another tool that, that I use a lot.
Marlene (13:20)
That’s all I care about right now.
there’s a lot of enthusiasm for Excel. You can just put, you write that down right now.
Greg Lambert (13:42)
You
Serena Wellen (13:45)
Okay, I will. It’s going to go on my list. I think Game Changer Marlene is really a perfect way to describe how we see this. So since we started working on Lexis plus AI, and even before we launched it, customers were telling us that they wanted to be able to use their own data alongside Lexis data with Lexis plus AI. And since we launched last year, it’s really been their number one request.
And I think, you you guys completely understand why, but the power of being able to combine, authoritative Lexis content and all of our know -how and things like Shepard’s with the customer’s own data, and unlock what’s trapped, within document management systems.
or the goodness of what’s trapped within document management systems. I think we know that there’s a lot that maybe needs to stay in there. But there’s maybe, maybe you shouldn’t be in there. Yeah. We all have closets. It’s the same kind of thing, you know, closets and drawers. I think there’s tremendous potential here across so many use cases, but I’m just going to give you a couple of examples.
Marlene (14:48)
Maybe shouldn’t even be in there.
Greg Lambert (14:50)
you
Serena Wellen (15:09)
And really focus on the drafting space. Cause it’s one that’s been really top of mind for me as I, as I do this work with henchmen and with organizations, document management systems. So a key problem in drafting any agreement is introducing risk with untested language. So with Protégé, you can check your language, your clause language against
the common language in your DMS, as well as, or at the same time as checking against what’s market from Lexis. And that right there, even though I think it seems fairly discreet, that’s a huge amount of time. And I think worry because, you know, drafting agreement is really all about avoiding the risk that in the future, something is going to fall apart.
bite you, result in a lawsuit, et cetera. And so there’s so much effort that goes upfront into trying to figure out how do I come up with the best language and make sure it’s not an outlier, it’s not outside our negotiating positions, and it is or conforms to what’s in market for that language. So that’s one.
Another problem is misinterpretation or failure to adopt the best negotiation position when you’re working through an agreement. so understanding the evolution of a negotiation in a very granular way by being able to look at the history of what’s changed. It’s almost like, you know, the DNA of an agreement.
and even a clause, seeing how a provision has evolved over time is super helpful in avoiding this departure from negotiating positions or taking positions that are going to be contrary, or even understanding how we’ve arrived at the position or the particular language that we’re at now. And then,
Marlene (17:26)
And is this for just sort of one document or set of documents, or can you sort of look across different types of negotiation documents and sort of see if there’s any sort of parallels or patterns?
Serena Wellen (17:43)
That’s a really interesting idea that you just mentioned, but it’s really right now intended to focus on one set of documents, versions of the same. exactly. And then another example that I wanted to mention is unlocking the expertise in the organization. So particularly in large organizations, big teams, big law firms.
Marlene (17:54)
Yeah, as an attorney is working through it.
Serena Wellen (18:11)
I want to be able to see who has most frequently drafted an agreement type, even a clause type for a client so I can leverage that expertise. Maybe I know something about that person or they’re the key partner on an account, but I also want to see who’s done the most work in an area and be able to tap that person.
or maybe look for some of their other work in an area. So this whole kind of like knowledge management zone for law firms in particular, I think it’s a very rich one and one that they’re very concerned about because their people are their greatest asset.
Greg Lambert (18:55)
Well, we couldn’t talk AI this year without talking about safety, security, and privacy. The three things that probably are the biggest barriers to bringing in new AI technology. So, given all of this concern around things like data privacy and legal work, how does Lexis Protégé ensure
Marlene (19:06)
SSP.
Greg Lambert (19:22)
like the internal work product, the sensitive documents and user data remains secure and still take advantage of all the AI -driven insights as well.
Serena Wellen (19:38)
I will say the security and privacy of our customer data is our number one concern. We are having a lot of conversations with our customers around Lexis plus AI and their data. And to be perfectly honest, I’m surprised that they hadn’t happened sooner. I think these conversations are really good and valuable and important. The last thing that we want to do.
is expose any risk to our customers. So this is really top of mind every day. And we devote a lot of resources and a lot of focus to it. So a few things that we do, we employ state of the art encryption. We use commercial cloud infrastructure, private models deployed to our own private cloud with every user session completely walled off. So that’s just the starting point.
I would say personalized AI really further raises that bar for obvious reasons.
Greg Lambert (20:45)
Yeah, you’re predicting my next question.
Serena Wellen (20:48)
Yeah, yeah, yeah, I am. I’m ready. Yeah. So, so this is why the personalization features in Protégé are completely transparent to the user and the customer and controlled by the user and the organization. So whoever uses these features has to opt in. They have, and they have to know what they’re opting into.
Marlene (20:50)
She’s ready for you. She’s ready for you. She’s a pro.
Serena Wellen (21:17)
So there’s a lot of transparency and information sharing that needs an education that we are doing to ensure that all of that happens. And once they do opt in, they need to be able to turn it off or on and have the data purged whenever they want. So this idea of control is central. I just wanted to point to a couple of other things.
customer documents in the DMS. So that’s my dog’s really cute bark.
Marlene (21:51)
It’s pod dog. Pod dog two. Pod dog two.
Greg Lambert (21:52)
get. Introduce the dog. What’s the dog’s name?
Marlene (21:57)
Yeah, please, please introduce.
Serena Wellen (21:58)
Yes, so that’s Lotus. She’s a wire haired fox carrier.
Marlene (22:04)
very nice.
Greg Lambert (22:06)
and a handful I can tell already.
Serena Wellen (22:08)
Yeah, she’s looking at me. She’s looking right at me. Barking saying, something. Speak. So, customer documents in the DMS. So Protégé uses all of the same controls and permissions that are contained within the DMS, which makes them, you know, the kind of fortresses that they are.
Marlene (22:09)
definitely, that’s good.
Whatcha doin’ mom?
Serena Wellen (22:35)
Protégé uses all of that so that the customers have the same governance of their DMS data, whether it’s connected into Lexis plus AI and Protégé, or it’s just sitting within their DMS. And then I would be remiss in talking about security and privacy if I did not say this next thing, which is super important. We do not use customer data to train our models.
Greg Lambert (23:02)
So let me ask you a technical question then. The personalization of it, is that, because I think I could see where people might think, well, if I’m training it to know me, is that training your large language model or is it kind of a separate way that it understands my writing style and my previous work?
that leverages it. what do you tell people to kind of set their mind at ease on the personalization?
Serena Wellen (23:39)
Right, we’re not training our models with our customers’ data. So when we think about training, there’s kind of two things. Large language models are pre -trained, meaning they consume a lot of data, and that goes into the parameters and weights. Training is often used to refer to what we call fine -tuning.
And fine tuning happens after the model is already trained and you can basically instruct the model to do certain things in a certain way, use certain information. And in that way, when you would, for the most part, when you do fine tuning, are not training, using any customer data to train the model. So the model is not actually learning.
in a permanent sense from the customer’s data. In the fine tuning sense, a lot of prompt engineering and instructions to the model are referred to as fine tuning. And this is a lot of what we do. It’s not all of what we do, but it’s a lot of what we do in being able to use customers past behavior and their data.
with the model to personalize a response. So if you can imagine, we can take, let’s say, a chaff history, and we can attach that to that user’s profile. That profile will stay private and secure. And then when the model needs to know that information, we give it temporary access to that information for the purposes of providing a better, personalized response.
Marlene (25:30)
So, Protégé is set to introduce features like voice commands and multimedia processing and horizon scanning. I need to know what that means in 2025. So, how do you see these features transforming the way legal research and document review work?
Greg Lambert (25:40)
Yeah
Serena Wellen (25:50)
Yeah, so horizon scanning is, again, I think about it as managing risk. And so how do you prepare in the present for the future? If you can scan the horizon and sort of see those blips and be able to make predictions, that will allow you to prepare in the present for the future. And when we refer to horizon scanning, that’s what we mean. It’s really about being able to use this
as much as we know about the past and the present to predict the future. So if you think about our data and all of the data that we have in news, in legislation and in law, being able to connect all of that together and use it to forecast, that’s what we hope to do with Horizon Scanning.
Marlene (26:42)
She used the word forecast and predicted. So, I guess, how does that work? Is it recommendations or is it based on this? This is where we think this is going.
Serena Wellen (27:02)
Right. I think it’s going to take a lot of different forms depending on the particular use case where the forecasting is needed. So it could take the form of recommendations, could take the form of analytics or a visualization that represents data and allows the user to explore different options, how things might turn out based on different factors.
So we’re looking at a bunch of different ways to work with both our data and potentially the user’s data where that might be appropriate. I do want to talk a little bit about voice because this has been kind of a passion project of mine for a few years now. If you don’t mind.
Marlene (27:56)
go ahead. I asked the question and you, what do we say? We say that you answer the question. Do you want to have it answered? So you go ahead.
Greg Lambert (27:56)
No.
Hahaha
Serena Wellen (28:02)
All right. This isn’t a deposition. I can explore. So, so briefly, I began exploring the application of voice technology to lawyers workflows several years ago, around the time that Alexa and other voice based assistance became widely available. just, became very curious about this and how it might be used.
to solve some of the problems that our customers were having. And we built several prototypes and tested one of them with law students and several attorneys. And the problem that we began targeting based on our research is one that I’ll call the interruption problem. So every time you stop reading or writing something to look up a term that’s unfamiliar, find an example, consult a colleague.
There’s a cognitive tax. You have to reorient yourself and find your train of thought once again. Of course, this has made worse in, you know, the distraction era of so many things coming at us on our screens. So everyone is familiar, but there is a, there’s a mental exhaustion factor here. There’s also a time loss because you would be surprised how much time it actually takes to reorient yourself to the work and pick up your train of thought.
tons of time is lost here. And because of the way they work, this plagues legal professionals, really. Not only are they constantly being interrupted by emails and other requests, by the nature of their work, they can’t know everything that they possibly need to know while they’re pursuing it, while they’re drafting a document or even talking to a client on the phone or something like that. They need to be able to stop and consult.
Marlene (29:36)
you
Serena Wellen (30:00)
other things, other documents, follow up on workflow. And the testers in this project loved the idea of having a voice activated service that could answer questions, find information and guide them while they stayed in their, the flow of their original task, more or less. The problem was the technology of our prototypes was just not good enough for lawyers.
It really needed to be precise, accurate, needed to be correct. And the voice technology language understanding just wasn’t there at that time, but it is now, which is so exciting. And not only is it there now, the translation, so being able to move from voice to text, text to voice, into other languages is all there. So I’m really excited to solve that interruption problem.
with a voice enabled Protégé. I’m looking forward to that. And I think, you know, that’s just one example of what voice technologies can do. I think you guys can think of a lot of other use cases where this would be super useful.
Marlene (31:11)
We’re pretty soon we’re going to hear like, hey, Protégé.
Serena Wellen (31:14)
I love it.
Greg Lambert (31:14)
Hahaha
So let me pull on that thread just a little bit. Do you think that the voice command is going to be very, that the younger attorneys are going to be the users of that? And then also, do you see them using it, voice command, at their desk? Or do you see them using it on?
like on the phone, when they may be in transit or just, I don’t see a lot, I actually see more older attorneys using voice at their desk than typing, but how do you see it as far as the new generation coming in?
Marlene (31:50)
On the move.
Serena Wellen (32:02)
Mm
Yeah, I think that they, the new generation is very oriented towards, towards voice and video. They use it in their, their daily lives all the time. I, when we were doing the research, they, they told us as much. said, we, don’t like to type. I don’t like to text. I want to just be able to say what I need to say and move on. think one of the constraints for law students and young attorneys,
has been that talking out loud in office environments and in places like the library and the classroom is not that acceptable. But now that we’re in these hybrid environments where we’re working in our homes a lot and we have privacy, can start talking and listening without disturbing other people or having them listen in on our business too much.
So I think they’ll really like it. And as we know, older attorneys have long used voice and dictation as a way to accelerate their work and spend less time doing things like summarizing depositions. But of course, now you don’t need to do any of that. Large language models can summarize depositions for you and create timelines and all that sort of wonderfulness.
Greg Lambert (33:36)
So one of the things that I think Lexis got a lot of credit for this when you were doing the commercial preview and testing of Lexis plus AI was that you got a lot of client feedback. I remember there were portions where you actually paused anyone from using the product while you actually
looked at the client feedback and then created kind of the next version to test out. So, you know, what are some of the, I think, the key improvements or features that you’ve been able to incorporate based on the feedback that you’ve been getting on Prodigy and what’s kind of your methodology going forward for making sure that you continue to get
that type of feedback and then build that into the product itself as needed.
Serena Wellen (34:43)
So this involvement of our customers in our product development, I think has been so exceptional. And the willingness of our customers, the enthusiasm that they have for engaging in this process has been one of the best experiences of my product management career. Product management, as you know, is really all about focusing on customer problems and trying to understand what they need.
But I think in particular, doing this customer preview program with Lexis plus AI, where basically the customers are learning alongside us and giving us really deep feedback about their experience. And then we go ahead and quickly turn that around and it comes back into the product and they can see that it’s really built this two -way relationship
of trust and a kind of a deeper communication that’s really enriched what we’re able to do for our customers. And I think it’s, it’s enriched their understanding of the system that we’re building for them. So I can’t say enough positive things about these preview programs. We invest a tremendous amount of time and effort.
in them because it’s so worthwhile and I know our customers do as well. What I’ve been hearing in particular around Protégé, a lot of feedback on drafting. Again, this isn’t the email you’re hearing from me during this whole podcast, but because it’s so important to the daily lives of attorneys and they’re really focused on it. What I’m hearing is they want to be able to draft
many different types of documents. They want to draft longer and more complex documents across transactional litigation. And of course they want to use their own documents as grounding for these generative documents. so drafting complex documents is a really hard thing, whether you’re a human or a model. And so we’re very focused on
iterating in these areas and really refining our approach so that we can provide a good high quality result for Protégé users. So this is one area where we’ve been really focused and the customer feedback has been super, super helpful in helping us refine our approach.
Marlene (37:29)
So, Serena, based on what you’ve told us about the integration, the customized profiles, the ability to learn from past work, know, Protégé really seems like poised to redefine how legal research is conducted. So what would you say the specific advantages are to lawyers and other legal professionals, you know, in terms of speed, in terms of accuracy?
and even depth of insight when conducting legal research.
Serena Wellen (38:05)
Legal research has basically been about finding the right information when you don’t necessarily know either what you’re looking for or exactly how to find it. Certainly this was my experience when I was a lawyer and I will tell you as someone who has worked specifically on search that my Boolean searching skills are not awesome and never have been.
Greg Lambert (38:31)
Yeah.
Serena Wellen (38:33)
full admission there.
Marlene (38:35)
And she’s so comfortable telling that to two librarians.
Greg Lambert (38:38)
Yes.
Serena Wellen (38:38)
I just, know, it’s, like some people are good at math. I’m just not good at, at bullying and searching, but so, you know, if you’re like me or a lot of legal researchers doing those things involves a lot of searching, reviewing, then searching and then reviewing and doing it again and again and again. And, know, you guys know this very, very well.
And you have to do that to make sure there isn’t anything you’ve missed. And it’s a ton of work and no one loves it. Well, there may be a few people out there who love it and bless them. but most people really don’t like it. And a lot of people hate it. And that’s just in preparation for the actual task itself, which is, you know, the drafting of the motion or the analyzing of the contract.
Greg Lambert (39:21)
Ha
Serena Wellen (39:37)
that required the research in the first place. And so if you think about this as the fundamental problem, there’s a few ways that Proge is transforming that process I just described. So it’s fast. It does the searching, the reading, and then could take that next step in your workflow to use that information to draft the motion or analyze the contract, for example. And
We, know, you can also think about fast in another way around performance. so Protégé is the fastest generative AI platform for legal research on the market. And, know, if you’re doing something and it takes, you know, three, three minutes to load on the screen, as opposed to, know, I can start reading it, in, know, after a few seconds. Over and over and over again, that makes a huge difference.
It’s also true, Protégé is accurate. So Protégé has all of our Alexis know -how built into it. So every response is going to be informed by authoritative content, relative relevant metadata, shepherd’s data, our taxonomies, et cetera. And we’re refining our system constantly. So accuracy is super, super important. It cannot be accurate enough, right?
It’s conversational, meaning as you’re progressing your work through dialogue with the system, you’re, you’re engaging in a more efficient, in a more natural way. And you’re going to be much more successful than searching and getting lost on your own. Searching is a transactional thing. It’s like search review, and then that it’s over. And I start all over again in a conversational system, the system.
has holds the thread of the conversation, just like you do. And it can quickly learn through the course of a short conversation, much more about your intent and what you need. And then it’s personalized as we’ve been talking about. It’s Protégé knows you. So it’s going to use your past work to inform its responses. And you aren’t starting from zero with every query like you were before.
Marlene (41:55)
So you had mentioned it breaks it down into like little steps. I mean how would you describe that from a technical standpoint? What is that?
Serena Wellen (42:10)
Working with these models is so different from the way that we’ve worked with AI in the past. Yeah. So, so it’s a form of fine tuning where we instruct the model. We can instruct the model to tell itself how to do things, or we ask another model to tell a second model how to do things.
Marlene (42:16)
Is it a big model? Is it mini models? Is it fine tuning? What exactly is it?
Greg Lambert (42:19)
Hehehehehe
Serena Wellen (42:39)
And then the second model informs the first model about, you know, okay, here’s what I found. Now what do you know, you do something with it. So, and the way in which we compose these complex systems is in a lot of ways, at least from my experience, quite different from how we operated before. Because you are.
you’re creating a set of instructions that is often in natural language.
rather than having to be translated into code and various other languages in code. Because the models are fluent in the same language that we’re fluent in, which I find fascinating. And it also has kind of transformed the way that we work with this technology.
Marlene (43:37)
So would that, Greg even knows what I’m going to ask. He’s smiling. So would you consider that like agents or is it something different?
Greg Lambert (43:41)
Ha ha.
Serena Wellen (43:47)
Yeah, so, agentic models is definitely an area that we are working with experimenting with. And I think that what we just talked about starts to get into this area. So, agentic models are models that have a much higher ability to plan and reason. And so, they can
They can do things like if you say, I want to do a certain task, make a plan for me.
and then, I’ll review that plan and then execute the plan and then go back and look at your resulting output and check it against the plan to make sure you haven’t missed anything. So that, that’s, and that’s a very simple flow, but that gives you an idea of what an agentic, model can do, which is.
I think much more successfully elaborate than the earlier generations of models that we were dealing with even as recently as last year.
Marlene (45:06)
In this case, you might do the research to draft the document, draft the document, and then the agent would check and make sure that the document was reflective of the research that you did.
Serena Wellen (45:19)
Yes, it’s also possible that the model would do all of that itself with your human
Marlene (45:31)
draft it and then just check and be like, okay, you’re good to go.
Serena Wellen (45:34)
Well, or I think a researcher, a lawyer would be more comfortable with being the human in the loop so that at every step the model could ask, do you want to review what might work so far?
Greg Lambert (45:51)
All right, before we get to the crystal ball question, this whole conversation and this may be a crazy thought that’s just coming from me, so take it for that. But you’ve got this personalization, this individual personalization of how Protégé in Lexis works with the individual attorneys. You’ve got this conversational
system that you’re setting up. You’re doing voice and multimedia over time on this. So I could see attorneys almost having a relationship kind of with their legal research tool.
Marlene (46:43)
No, it’s her. Stop! No! Don’t go there.
Greg Lambert (46:46)
So, but I mean, this is a question that we’ve been having with the chat GPTs and whatnot that, especially like with the voice interactions that there’s this kind of super personalization of it where they’re almost thinking like it’s, know, they’re anima -phorth, eh.
They’re turning it into a person, since I can’t say the other word. Is this something that you guys have thought about as far as, you know, it’s such a personalized experience? Has there been any conversation about either enhancing that or maybe pulling that back a little bit? don’t know what kind of conversations you’re having.
Marlene (47:14)
You know me. You know me.
Greg Lambert (47:42)
on that front if you’re having those at all. And like I said, this may be a crazy idea just coming from me from this conversation.
Serena Wellen (47:49)
I think it’s an interesting idea. I’m certainly aware of in the general purpose model world where you have companies like character .ai that are creating very realistic sounding and seeming representations of people to act as companions and guides and helpers and more.
we’ve been very focused on making very powerful, useful tools for attorneys where they stay in control and they are the human in the loop.
So what I do know is when we talked to legal professionals about voice, whether it was law students or lawyers, the thing that they were most interested in was not the personality of the voice, but that it sound professional and authoritative, but not
overly informal or friendly. I found this very interesting when we did this research. Now this was some years ago and again, voice was relatively new. But I think if you asked customers and law students the same question today, you might get very much the same answer. They’re not looking for a friend or a companion. They’re looking for an assistant for help, knowledgeable help, professional help.
And I think that, you know, this is an area that is really ripe for exploration. How best to do that? How best to create an assistant that is going to be the most helpful in a personalized way? So then what that means, what the kind of voice or the intonation that will vary from user to user. So I think you’re right.
in what you say that people will want to have different manifestations depending on who they are and what they’re doing. But as far as taking it up into something that is a virtual legal assistant or something like that, that’s not something that we’re focused on at all. And it’s not
Greg Lambert (50:22)
Mm -hmm.
Serena Wellen (50:29)
something that we’ve heard from our customers that they want. I think it’s an interesting idea, but not something that we’re focused on right now. We’re focused on being helpful and useful in the way that our customers want.
Marlene (50:30)
you
you
Greg Lambert (50:44)
Well, so no Scarlett Johansson voice or Matthew McConaughey voice, although that would be kind of cool.
Serena Wellen (50:54)
It would be cool.
Marlene (50:55)
All right, all right, all right.
Serena Wellen (50:57)
No.
Greg Lambert (51:00)
All right, well now it’s time for the crystal ball question. So Serena, what do you see as the challenges or the changes that over the next couple of years that you think we’re gonna face with this type of technology?
Serena Wellen (51:19)
So two to five years seems like a really long time. So we’re all agreed on that caveat. In terms of market, I think we’re just still at the beginning of a sea change. I think we all really know that. That this is, know, the wave is just starting to form and we all don’t really know where it’s going to go. But I think…
Greg Lambert (51:21)
It’s so long, so long.
Marlene (51:24)
How about six months, right?
Serena Wellen (51:49)
Legal organizations who are prepared for their AI moment, if you will, because they’ve retained and organized and prepared their data are going to pull ahead, way ahead of others who haven’t, who are still struggling in this space and continue to struggle. the organizations that have done that, that’s a big investment. That’s hard. And so.
You know, if you haven’t started, this is the time to start making those plans and really learning about what you need to do in order to catch this wave. The other thing I would say is the challenge for legal information providers like Lexis is to support that transformation. We need to help.
our customers unlock the deep value in their data while we continue to evolve alongside this rapid technological change. And I think we’ve been able to make some really good progress in both of those areas. I’d also say, when I think about this, our advantage at Lexis is our customer focus and understanding the authoritative legal data that we have, the know -how.
that we have in this space and our ability to use the latest emerging technologies to give our customers what they want and what they need.
Marlene (53:31)
Well, thank you, Serena, for that. Serena Wellen, Vice President, Product Management, Lexis, Nexis, Legal and Professional. Thank you so much for coming on and talking with us today.
Greg Lambert (53:41)
Yeah, thanks.
Serena Wellen (53:43)
My pleasure.
Marlene (53:45)
And of course, thanks to all of you, our listeners, for taking the time to listen to the Geek in Review podcast. If you enjoy the show, share it with a colleague. We’d love to hear from you, so reach out to us on LinkedIn.
Greg Lambert (53:56)
Serena, we’ll put some links in the show notes, but what’s the best way for people to find out more about Prodigy or to reach out if they have more questions?
Serena Wellen (54:09)
For more information on Prodigy, can go to LexisNexis .com and navigate to our Lexis plus AI product page. Lots of information there. And if you have questions for me or would just like to connect, please reach out. I’m on LinkedIn. So you could just search for Serena Wellen.
Marlene (54:27)
Thank you. And as always, the music you hear is from Jerry David DeCicca. Thank you very much, Jerry.
Greg Lambert (54:34)
Thanks, Jerry. All right, Marlene, I’ll talk to you later.
Marlene (54:36)
All right, see ya.