This episode of The Geek in Review podcast provides an in-depth look at how the AI assistant Paxton, created by Tanguy Chau and Mike Ulin, is transforming legal work. The hosts speak with the founders of Paxton to explore the pain points their technology aims to solve and how generative AI can enhance lawyers’ capabilities.
Tanguy and Mike discuss their backgrounds in AI, regulatory compliance, venture capital, and management consulting. This diverse experience informed their vision for Paxton as an AI assistant specifically built for legal and compliance professionals. They explain that Paxton is trained on millions of legal documents and regulations, allowing it to search this vast knowledge and retrieve highly relevant information rapidly. A key feature they highlight is Paxton’s accuracy in citing sources, with every sentence linked back to the original text.
One of the key features of Paxton is that it can automate repetitive, low-value legal work to make lawyers more efficient. Tanguy notes that tasks like reviewing thousands of sales contracts clause-by-clause or compiling 50-state surveys that once took weeks can now be done by Paxton in minutes. Mike discusses Paxton’s advanced document comparison capabilities that go beyond keyword matching to understand meaning and intent. This allows quick, substantive analysis of contracts, marketing materials, and more.
Exploring the future, Mike predicts that like software developers, lawyers who embrace AI will become much more productive. But higher-level strategic thinking will remain uniquely human. Tanguy shares an analogy of a human on a bicycle outpacing a condor, the most efficient animal. He believes combining human creativity with AI tools like Paxton will enable radically new levels of efficiency and capability.
Paxton.AI’s Tanguy and Mike make a compelling case that AI-powered tools such as Paxton will fundamentally transform legal work. By automating repetitive tasks, AI will free lawyers to focus on high-value, client-facing work. Overall, this episode provides great insights into how generative AI may soon become indispensable for legal professionals seeking to improve their productivity and capabilities.
As a special treat, we wrap up the interview with a demonstration of Paxton.AI’s capabilities. (YouTube only)
Greg Lambert 0:02
Welcome to The Geek in Review the podcast focused on innovative and creative ideas in the legal profession. I’m Greg Lambert and I am flying solo for a little bit. I’m hoping Marlene can jump on while we’re doing the interview, if not, everyone is stuck with me today. So I want to start this off by warning the audience that we’re probably going to get pretty geeky on this week’s show, as we dive into some examples of what generative AI can do with specific legal task with the guests that we have this week. So we’re really really excited to have the cofounders of Paxton.ai Tanguy Chau, CEO and CTO Michael Ulin. Glad you guys can both join us. And hopefully we can all geek out here. So Tanguy and Mike, welcome to The Geek in Review.
Mike Ulin 0:55
Thanks for having us, Greg. Yeah, excited to be here and a big fan of the show.
Greg Lambert 1:00
Thanks. Thanks. Thanks,
Tanguy Chau 1:01
Greg. It’s such an honor. Thank you.
Greg Lambert 1:04
So it’s great. Great having you here. So Tanguy, let’s start this off. Because not everyone may have heard of Paxton. Yet. Do you mind giving the audience a little bit of your background? And the history of Paxton and why you and Mike decided to launch this new venture?
Tanguy Chau 1:23
Yeah, absolutely. Well, Greg, like every good company, the story of Paxton starts with the relationship between Mike and I as friends and as founders. Mike and I are AI engineers. We are enterprise experts. I did my master my PhD, my MBA from MIT. Mike and I met a decade ago, we work together while at McKinsey and Company. This is the largest management consulting firm. And we were strategic advisors to a lot of fortune 500 companies. But really what sets us apart as legal tech founders is that both Mike and I have a deep understanding of Legal and Regulatory Affairs. Now on my end, I spent almost a decade as an investment advisor in the venture capital industry, eventually serving as the Chief Compliance Officer for an SEC and FINRA regulated, firm. So it’s really through this personal experience and shared desire to help regulatory and legal professionals make sense of millions of laws, rules and regulations that Paxton was born.
Greg Lambert 2:35
And Mike a little bit about you.
Mike Ulin 2:38
Yeah. Thanks for having us on. As Tanguy alluded to, we got a, we met and became friends while doing a stint at at McKinsey and Company. And then after that, I made my way into the world of legal tech, working in data science for a company called RPI x, which was involved in serving the General Counsel’s Office of large tech firms on intellectual property matters. So work they’re developing natural language processing algorithms, working with with patent data. And then I was part of the founding team, along with a few other ex McKinsey colleagues of a company called zesty AI, where I was the founding AI engineer and, and head of AI for XSD, where we focused on building enterprise AI models for the US property insurance industry. So we were actually the first AI model approved by the California Department of Insurance for use in underwriting and pricing of property insurance in that state. So worked very closely with regulators there, and then worked closely with regulators in 10 other states to get our AI models have proved. So bring, bring a deep experience in AI to the space and we’re really excited about the possibilities of applying generative AI in the legal context.
Greg Lambert 4:09
Let’s say we’re riding up the elevator together. So I’ll let one of you pick who wants to give the speech here. But if you were going to introduce me to Paxton and what it does, what what’s the what’s the elevator speech that you give?
Tanguy Chau 4:26
Yeah, I’ll volunteer for that. You know, Paxton is really a generative AI assistant that is built for legal and regulatory compliance professionals. That’s really what it does. And unlike other assistant it is really trained and build from the ground up. We have painfully collected millions and millions of regulation rules. Statue use and use this material to train a very accurate legal AI assistant.
Greg Lambert 5:08
And Mike, I want to cover some. Because usually, when people create something like this, there’s some problem that you’re set out to solve. And so can you. And I know that a lot of what you’re what you’re doing and you’ve elaborated on in the media is about the biggest pain points and inefficiencies that you see in the legal sector. And so can you elaborate on on some of those pain points that you’re trying to solve? They’re Tanguy.
Mike Ulin 5:41
Absolutely great. So a lot of this is kind of born out of our experience in the legal and regulatory space. So, Paxton is really designed to take a lot of the pain points away with regards to legal and regulatory research. So having done a lot of regulatory work in the insurance sector, you know, insurances is one of those areas that is not regulated at the federal level, it’s actually regulated at each of the individual 50 state levels. So if you’re a a national player, like some of the companies I worked with, at my prior company, we we worked with companies like Berkshire Hathaway, farmers, AAA you have to figure out the different regulatory regimes of each one of those those 50 states, and then talking to legal professionals, this can be a pretty daunting, tedious and, and very labor intensive task. So you know, what we’ve done with with Paxton is, we have all of the laws, rules and regulations from not only the federal level, but also each of the 50 states. So one, one particular use case that’s been gaining a lot of traction with our clients in, in financial services, or the insurance industry is conducting 50 State Surveys in a matter of minutes. So you can simply ask Paxton for what the regulatory regime is on a particular topic in each of the states and get a response back that you can then use as a as a starting point for your research, rather than, you know, spending a lot of money doing this, or, you know, we’ve heard that these things can take months to complete,
Greg Lambert 7:39
I can verify that it takes a long, long time to pull those together. And, and so now I think a lot of partners are going to have to figure out new projects to give summer associates because that’s a big task that a lot of summer associates get is take this issue and give me a 50 state survey. Now a smart associate would then do it in the 10 minutes, that it would take and then wait around for a couple of days and then hand it in to them. So it didn’t seem like it only took 10 minutes.
Mike Ulin 8:18
You know, and we want to we want to empower smart associates here at Paxton. Here we go.
Greg Lambert 8:24
So, Tanguy, you recently wrote an article that was in Forbes, and it was called unlocking the 10x lawyer how generative AI can transform the legal landscape. And in the article, you discuss how generative AI can help lawyers become more efficient by automating low value tasks. So I think most of us probably have some examples. But I want to I want to see if you can provide, you know, some examples of low value legal work that you specifically think can be automated with with AI. And, you know, how much time do you estimate that this could actually save a lawyer say, even on a weekly basis?
Tanguy Chau 9:11
Yeah, absolutely, Greg. So let me start by defining what we mean by the 10x lawyers. And what we mean here is that with the proper tool, with the proper support, a lawyer can be 10 times as efficient as a lawyer that doesn’t have access to tools like Paxton. And so, you know, let me let me share in this piece, right, we analyze a survey that Wilson Sonsini, one of the largest law firm in the world did, and that survey explains that 67% of in House lawyers feel buried in low value, highly repetitive work. And let’s give some samples here, right? You look at an m&a process, and a lawyer might, a law firm might need to look at the sales contracts of 1000s and 1000s of vendors, and go through the through the language of all of these sales contracts, and it’s very repetitive. They have to look clause by clause over all of these documents, and, and identify variations in the language as it’s being used. It takes hundreds of hours to review each of these contracts. This is a perfect use case, for an AI agent such as Paxton. A second example, for example, would be, you know, when when we talked about 50 State Surveys and, and looking and reading vast knowledge databases or fast amount of information, it requires reading through 1000s of pages, just to identify and find critically important information that is relevant to the to the case, Paxton, for example, is able to read through 30 million pages of legislature has perfect memory, perfect recollection and accurate retrieval of the most mundane, but critically important information. And so what we think is, you know, this is this is what Tanguy is outstanding for handling very mundane, very boring, repetitive work, and to free attorneys to focus on the high value work, the highly engaging the client facing work that all attorneys and all lawyers love to do.
Greg Lambert 11:47
So, I want to ask you, because we had a guest on a couple of weeks ago in Marlene has joined us as well. Hi, Marlene.
Tanguy Chau 11:57
Hi, Marlene, Marlene.
Greg Lambert 12:01
Tanguy, I did want to follow up on that we had a guest on believers last week that talked about how you can structure the generative AI, even even when it’s trained on, you know, legal specific things that because of the because of its ability to be creative, you know, it could take it could take liberties with the with quotes on it, change things around. And so he talked about some ways that they did the multi token, I forget how multi token authorization or multi token something but basically where material that is like quoted stays the same so that you don’t get the any type of loosen hallucination. What are you guys said
Marlene Gebauer 13:00
they’d like they’d like Link words. So it wouldn’t be just like separate words. They’d link it so that things wouldn’t get separated? And it was there less of a chance for hallucination
Greg Lambert 13:11
and how it worked. Yeah, and especially in legal where you want to be very specific and precise. When you’re quoting, legal materials. So how are you looking at? If I’m going to cite something, that I am assured that the answer that I’m given matches what’s in the documents? Is that do you link to the documents? Or do you have something else going on in the back end to kind of make sure that you’re, you’re getting very good answers?
Tanguy Chau 13:49
Yeah, absolutely. I think this is a this is a very important problem that we spend a lot of time thinking about how do you actually do this accurately? There are many different ways that one can approach this problem, both from a technical approach different techniques, and machine learning techniques that one can can use. What we have focused on and what has resonated with our users is very accurate sourcing. So for any, you know, the use of AI, really, it’s really important that the user trusts the answer. And there’s a saying that says trust but verify. And this is what, Paxton is really good at every sentence, every word, every paragraph is has a footnote that you can verify the where the information is coming from directly to source text to source regulation and it takes you to the source of the information. So this is or approach or approach is you The the ability to very quickly verify that the answer is given. Because eventually, you know, it is still on the user’s responsibility to check and make sure that the information that the tool provides, is, is accurate when they when they file a motion, for example.
Marlene Gebauer 15:23
And is it a footnote? Or is it like a reference in in the text?
Mike Ulin 15:27
It’s a, it’s an actual citation in the text. And, you know, folks are free to, you know, check it out for themselves at Paxton.ai. Ai. You know, we’re offering free free beta access right now. But you can actually, when you get a response, every response comes with a footnote, as Tony was saying, and that is a link direct back to the primary source material. So you can link to the regulation or the statute or the rule, or, or materials that you upload yourself to Paxton. So you can verify what what the system is saying is accurate? Yeah.
Marlene Gebauer 16:07
And it’ll go back to that specific area. For reference, yeah,
Mike Ulin 16:11
it won’t. It won’t just, you know, link you to the entire
Marlene Gebauer 16:15
California section of the tax code.
Mike Ulin 16:19
Just an entire chapter. Yep, there you go.
Marlene Gebauer 16:25
There’s my weekend.
Greg Lambert 16:26
Yeah, I’ve actually played around with it. It’s, it’s really cool. And thank you guys for letting it during, during this phase of you guys launching, letting everyone take a look at it for free. because not everybody’s doing that.
Mike Ulin 16:45
Ya know, we’re very feedback driven culture here at Paxton. And, you know, we’d love for and thank you for, for testing out the the tool, we’d love for your audience to take a look and let us know what you think we’re always on the hunt for ways to improve it.
Marlene Gebauer 17:03
How specifically, are you designing Paxton.ai use of artificial intelligence to augment the capabilities of lawyers? What is unique in the way that you’re leveraging AI now that you may not have that may not have been possible, just a year or two ago?
Mike Ulin 17:20
Sure. Happy to have to take this one. So I’ve been in the field of AI for for, you know, my whole my, almost my whole career over over a decade now. And we really designed Paxton with with the needs and workflows of an attorney in mind. And really taking advantage of the new generative AI, the new capabilities that the generative AI ai offers, offers today. And, you know, if folks want to get a sense of, of how that’s how that’s different. You know, you can really see it for yourself, you know, just just take a look at what’s available, you know, when you do autocomplete on, on your phone, or autocomplete on, on Gmail, you know, you would just get a phrase or a couple words, as a suggested next thing to say, and that was kind of the state of the art for a while. But but, you know, today, and you can experience this with yourself for yourself on on, on Paxton, we can now generate, you know, entire documents, based on on really accurate source materials. You know, we have models at Paxton that can take into the, you know, can take into account like the context of about a book worth of material at a time, versus, you know, what we were what we’re dealing with just a few years ago, where it was like, what’s, what’s the next word that I want to take someone? So I think that that is kind of the the change in how AI has, you know, what AI systems are available today that’s really enabled these types of use cases.
Greg Lambert 19:18
So, you guys have a really good presence on medium. And I think you’re doing a really good job of kind of reaching out beyond the, you know, the Twitter, LinkedIn, although you guys do a great job on on those as well. But, you know, on medium, you write some, some very specific and interesting ways about how to use Paxton AI. And so we want to go over just a few examples of some of those articles that that we’ve read through and Tanguy I’m going to start with you. There was an article on Paxton’s drafting capabilities. And in the article, it mentioned generating documents from templates. So can you expand more on how lawyers can customize those templates? And do they have some flexibility to develop templates attune to their own specific needs?
Tanguy Chau 20:24
Yes, absolutely, I think, let’s remember, right. Paxton really is an intelligence layer that sits atop different knowledge databases. And so you can really ask Paxton, to perform different tasks to adjust and to adapt, you can use the curated database that we collect, and we maintain, and we monitor, such as, you know, the, the federal and state statutes, rules, regulations, and more to come, you know, obviously, but you can also ask Paxton.ai intelligence, and apply it to your own knowledge databases. And you do that by simply connecting it either to your document management system, or uploading your own set of documents or cases, or your own templates and the language that you want Paxton to use. And so you can then ask Paxton to read, and Paxson will adopt your template, and respond in the style and interest in the style and then draft following, you know, draft motion drafts legal memos, following your custom firm templates. And so this is really how we’ve made Paxton easy to, to find for our users to fine tune to their specific knowledge, as well as their specific templates.
Greg Lambert 21:59
And, and just a follow up on that one. Because I can hear my, my IT security guy screaming right now. So how do you work with, with with firms, with lawyers, with your users on making sure that that information is secure, and that someone else won’t be able to get access to that?
Mike Ulin 22:26
Yeah, so we have a very deep background and serving large enterprises. So so both Tanya and I served very security, conscious, fortune 500 companies in our, in our prior life at McKinsey, and, you know, when we were developing AI models for the insurance industry, you can imagine those, those companies also had very strict security requirements as well. So we really took that that experience and you know, that knowledge of what, what organizations care about, and that’s, that’s built into the design of Paxton. So we meet all the requirements for for sock two. And, you know, we work with security teams at law firms all the time to understand their individual requirements, and really tailor Paxton.ai security settings and data retention policies to meet the needs of the firms that that we’re working with.
Marlene Gebauer 23:31
So as law librarians, the article on Paxton Boolean search composer caught our eye and, you know, I think there there are still librarians out there that say, you know, boolean is the best is the best way to search and that, like natural language is kind of the, the death of the ability to, to get really down and granular in your search. And while while, you know, experts, searchers are very good at this sort of thing, you know, sometimes more casual searchers are not. So, you know, what are some of the examples of complex legal research questions that the search composer can simplify that, you know, lawyers and maybe more casual users struggle with? You know, how does the explanation help lawyers learn in the process?
Tanguy Chau 24:23
Yeah. Yes, this is, this one is a very, very popular feature within the Tanguy platform. And it’s a feature that was requested by our community. I think, I think it speaks to like the how Mike and I are thinking about being very building a very user and consumer centric for a company first so. So for the audience. We built a, a Boolean composer and Uh, what we heard is a lot of our users didn’t enjoy or didn’t like writing Boolean queries. You know, they’re complicated. They’re unintuitive. Shocking, shocking, right? Yeah.
Mike Ulin 25:17
This net natural language thing is just a fad. Yeah. Right.
Tanguy Chau 25:22
But Marlene, I mean, it’s so true, right? We were forcing users to adopt the language that machines use, right? It’s not like you and I, like speak using Boolean, right to read speak using natural language. And so instead, we thought that instead of asking users to adopt the language of machines, we reverse this process. And we, you know, we had Paxton learn the language of users instead. And so, you know, instead of coming up with complex Boolean expressions, or users have the ability to very simply explain in plain English, what do they want from Paxton, and it might be retrieving all of the cases for over the last 10 years for a particular issue. And, you know, and if and when needed, Paxton can transform these requests into clear Boolean expressions that the users can then use in tools like LexisNexis, and Westlaw. And then Paxton will also explain how this expression was built stage by stage what each of the Boolean queries is used for. And so makes it very easy for a user to draft a Boolean query, but also adjust it adapted and, and takes a lot of the mystery out of Boolean searches.
Greg Lambert 27:00
Well, that’s interesting, because I feel like definitely welcome. Yeah. Well, and one of the things that and I know we do it here at my firm, and I’m sure other firms do the exact same thing is, like, like, if somebody is searching on Westlaw, and they’re having problems, we suggest that they call one of the research attorneys at at Westlaw, that helps them basically create a search structure for that. So you know, I don’t I don’t want to run the people out of out of a job up in Eagan, Minnesota. But, you know, sometimes you don’t necessarily want to talk to somebody, you want to get something that you just want to figure it out on your own. And I like the fact that you’re not just giving giving an answer, but you’re actually explaining how you’re getting to that result, which maybe that will teach some of the attorneys how Boolean works. So
Mike Ulin 28:01
absolutely. Yeah, you know, we’re just a very, I mean, this this, as Don, you mentioned, this was just a suggestion from from one of our users that we then took and adapted to the platform, just just to make the lives of attorneys just a little bit easier. So we’re glad folks that our folks are getting use out of it.
Marlene Gebauer 28:21
And the transparency builds trust, so yep, that’s good.
Greg Lambert 28:25
Absolutely. So So Mike, the there was another piece that featured document compare. It talks about what what’s the logic behind the the document difference? In highlighting? So does it just look at the wording of the document itself? Or is there some kind of more semantic analysis going on? It’s identifying the, you know, the substantive differences in the documents when you’re comparing them?
Mike Ulin 28:59
Yeah, great, great, great question. And, you know, this this feature of Paxton is really meant to go beyond just like the the redlining or track changes that that folks might be used to, it’s not limited to just keywords or, you know, sequences of words in the document, it actually looks at the meaning and intent of the text in the document. So you can use it to say, compare it to NDAs. You know, maybe you have a template NDA that’s been approved and one from a vendor that you want to understand you know, what their their similarities and differences are. And Paxton can really quickly in just a matter of seconds, highlight what what the substantive differences between those two documents are. And it doesn’t matter if those claws Is are in a different order. Or if you use a slightly different language to describe something, Paxton is able to just tell you, you know, this particular clause covering this particular issue is present in one document, but not in another. And that really can save folks a lot of time in what used to be a pretty, pretty tedious exercise. You know, other other clients are using it for, say, marketing collateral review. So if you have a set of requirements or, or a regulation that that covers your marketing material, you can just plug that into the document compare tool and see if the document you’re evaluating like a webpage or a piece of print collateral complies with the requirements that you want to evaluate. And it just saves a lot of time versus versus what folks were used to doing before.
Marlene Gebauer 31:02
So you seem to have a little fun with the topics you use for your articles by looking at real world examples of how Paxton can be used on things like the congressional UFO hearings, or the ability to quickly go through most of the recent indictment of former President Trump. How was that marketing working? I mean, are you getting a lot of response from that?
Mike Ulin 31:26
Yeah, we, you know, we like to have a bit of fun, and, you know, I think demonstrating a lot of these, these capabilities on on topics that are all topical, and you know, in the public consciousness helps really bring home the value, you know, even you know, even though folks will probably not use this to analyze a UFO hearing, they’ll probably use it.
Marlene Gebauer 31:54
Mike Ulin 31:58
But a lot of UFO attorneys in your audience, Greg and Merlin, yeah,
Marlene Gebauer 32:03
we’re gonna find out.
Greg Lambert 32:04
It’s definitely a growth industry. So we’ll see how that will speak. Speaking of the congressional UFO hearings, in your article on that, you the analysis mentions being able to instantly process a lengthy transcript. So what’s the key to Pakistan’s natural language processing that enables such a quick turnaround time on that? And and how do you train the AI model to handle that large amount of text?
Mike Ulin 32:43
Yeah, and, you know, we wanted to pick this more fun example. But you know, what, what we’ve heard from from users is, you know, there’s a real pain point about working with video content, whether it’s an agency hearing a congressional hearing at the federal or state level, or, you know, depositions or other types of long form content that, you know, you may be pressed for time, and you have to really get the key takeaways from it rather quickly. And what, what really enables this is Paxton.ai ability, like, like we discussed earlier, Paxson is ability to understand a large amount of information at one time, you know, in the generative AI space, this is what’s known as the context window. So the the amount of information that Paxton is going to consider before providing you an answer. And, you know, as opposed to some of our competitors, Paxton is able to take in a lot more information at once. And, and we really see the returns on that, in terms of the quality of our answers. And in terms of the quality of the responses. Paxton generates, so Paxton.ai Paxton.ai trained on on legal data, and, you know, it really becomes a very flexible tool. And so it doesn’t matter if we’re applying it to UFO hearings or something a bit more serious. We can we can use this tool in a wide variety of areas.
Marlene Gebauer 34:30
When Paxton does analysis on these large documents or transcripts, what specifically does the summarization capacity focus on extracting? I mean, are you using you know, semantic are you using Are you it’s the amount of information submitted generally is the amount of references of a particular thing. So, you know, does it identify or does it specifically, you know, identify people charges, key evidence quotes. How does it work?
Tanguy Chau 34:59
Yeah, I mean I think this is this is what’s really special about about this generation of of AI tools, right? The flexibility that it has. So for example, if talking about the documents of the Trump indictment, you can use Paxton to summarize indictment. Generally, you can ask Paxton to identify which people were charged, what are what were they charged with? What are the key evidence? What are the specific quotes. And I think this is something that wouldn’t have been possible a few years ago was prior generations of AI systems. But really generative AI has an understanding of the intent behind the queries and an intent of the source documents. And and that’s what makes this so powerful, especially in the legal field is that it’s no longer searching for words, it’s no longer searching for phone numbers or pattern it really is searching for what is the intent of the user. And this is, this is something that, you know, users games have tremendous flexibility and how it’s being used.
Greg Lambert 36:27
So we kind of have danced around this a little bit. But they’re kind of a common phrase, it’s saying that it’s being said, and that is that generative AI tools like this aren’t going to replace lawyers specifically, but it’s going to give a significant advantage to lawyers who leverage the AI tools against those who don’t. And so as we, as we probably see this happening more and more as as AI TAS take on more of some of the base work that that lawyers do, how do you how do you see the day to day work of a lawyer changing? And do you have any suggestions or thoughts on some of the skill sets that the AI enhanced the lawyer is going going to need as far as a skill set, or even a specialization?
Mike Ulin 37:31
Sure, you know, and I think this is, this is just another another tool. And I think that’s what, what really makes us makes us human is the ability to leverage and use tools that can actually make us more efficient and enhance our work. You know, just think about the switch from looking up materials and in a book or a library versus now you have Westlaw and LexisNexis, you can think about this as kind of a new iteration or wave of that. So I think really partnering with these, these tools, and learning how to utilize them will be will be critically important for attorneys in the future. But you know, those things that are really, you know, kind of uniquely human, higher level strategic thinking, thinking about how different pieces of of a problem fit together, and really, interacting and guiding clients. Will will be really, I mean, they’re important skills today. And I think those are going to be even more important in the future. But hopefully, you don’t have to, you know, banging your head against the wall reading, reading a lot. A lot of Boolean queries going forward.
Marlene Gebauer 39:03
Damn. Well, speaking of attorneys in the future, yeah, Mike, and Tanguay, we ask all of our guests are crystal ball question. So please take a peek in the future for us and let us know what you see as a change or challenge in the legal industry over the next two to five years.
Mike Ulin 39:26
Yeah, happy to go first. You know, I think you know, it’s notoriously hard to predict the future but I’ll give it a shot. You know, one of the professions that has really readily adopted the use of of generative AI I think a lot more quickly than a lot of the others has been software engineering or software development. And I think if you want to understand a bit of what the future is going to look like we can we can see how this person Session has adapted to generative AI. And, you know, I think there was a study out of Microsoft around, you know, software engineer, software developers that utilize generative AI tools are up to 60% more productive than their peers that do not. And, you know, just thinking about the way I code when I, when I still have time to code. And, you know, our team at Paxton, when they’re coding, utilizing these tools, it’s really dramatically changed their workflow from from a year ago. And they’re a lot more efficient and a lot more productive. And I think we’ll, we’ll see a lot of this in the legal profession, you’ll, you’ll have an assistant like Paxton, that that will make you more effective, more productive, and allow you to get back to, you know, give you more time in your day back and allow you to focus that, on higher value, add sorts of things. And I think you’ll see these, these AI assistants proliferate, not just in legal but but in other professions, like medical as well. Tanguy,
Tanguy Chau 41:16
and yeah, you know, I think, to look 1020 years into the future I, I start by drawing inspiration and looking into what happened in the past. And so I have a, there’s a, there’s a fun analogy, that I think we can learn a lot from. And so in the 70s, I don’t know if you’ve seen that study, but Scientific American did did a study on locomotion across species. And so they calculated the amount of energy that was required to travel a given distance. And so, you know, the study concluded that the Condor was the most efficient animal, you know, an expense, the least amount of energy to go from point A to point B. And, you know, on the other hand, humans spend a lot of energy to perform the same task. And we’re kind of in the bottom half of, of efficiency. But what was really interesting about this study is that it also measured the efficiency of a human on a bicycle. And surprisingly, when, when a human is aided by a machine, they expended the least amount of energy to perform the same task and was even more efficient than than the condors. And so I think this is a really good analogy of what’s going to happen in the next 1020 years. There are a lot of tasks that we perform today, where we’re really not very efficient. And AI assistants such as Paxton.ai, I see them as the bicycles of tomorrow, you know, you combine a human with a machine, and we invented the most efficient way to perform a task, while minimizing the energy spent. And that’s the true power of innovation. And whoever adopts AI tools, like Paxton first, will do the same work faster, better and more efficiently.
Greg Lambert 43:19
Mike, I want to make a quick comment on on your analysis with the software engineering field in law. And this is just a personal anecdote. And that is, so I’ve been using the generative AI tools out there to help me kind of relearn some programming. And, and so it’s been, you know, I would say, measure it in that it makes me much more efficient, because it can give me a great place to start, say, I’m doing a Python script. So it can give me some something to start with. And, you know, and I get to a good result much faster. But I will also say that on the flip side is the net motivates me to now take on additional projects that I would have never have done before. So you know, I just wonder, do you see that in? In the future for lawyers is that yeah, it may may only take me now 10 minutes to get to, you know, something that usually took me two hours to do before. In Do you see the same analogy with say, Well, now I can take on, you know, these other things that I didn’t, you know, wouldn’t have taken on before, as well. Is that how you see it?
Mike Ulin 44:46
Yeah, no, I think that’s a really great example. And there’s there’s a lot of great stories of folks like yourself, being enabled by this, this new set of tools And, you know, you can, you can look at, you know, what happened with with the spreadsheet a few decades ago, you know, in the time before Microsoft Excel, you know, the bookkeeping profession used to be a lot more prevalent. But after, after we made the transition from from physical spreadsheets to software based spreadsheets, we saw a dramatic increase in the number of accountants, then you can think of accounting as a, you know, it’s a more higher level form of that group of work than bookkeeping. And I think I think something similar will happen. With the legal profession with with other professions, when it comes to generative AI, you’re, you’re able to dedicate a lot more time, whether it’s personal time to activities you enjoy, or activities that are more higher value, add more strategic in nature. And, you know, hopefully, that that that’s inspiring for folks, and you know, you, you get to spend a lot more time on things that you actually like doing versus versus just trying to make do with with inadequate tools.
Greg Lambert 46:23
So Well, Mike and Tanguy. From Paxton, Ken, thank you very much for taking the time to talk with us. And tell us more about your work there at Paxton AI. Thank you.
Mike Ulin 46:36
Yeah, thanks for having us.
Tanguy Chau 46:38
Thank you, Greg. Thank you, Marlene.
Marlene Gebauer 46:41
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 social media, I can be found on LinkedIn, and I gave our M on x. And also look for The Geek in Review on threads.
Greg Lambert 46:59
So and I can be reached at LinkedIn or glambertpod on X or glambertpod on threads or everyplace else. Plus, we are launching our Geek in Review channel on YouTube. So that one as well. So Tanguy and Mike, if someone wants to learn more about Paxton, or reach out to you, where can you guys be found online?
Mike Ulin 47:27
Yeah, if folks want to learn more, just check out Paxton.ai. And right now, as we mentioned before, we have free access to Paxton so people can really experience it for themselves. Just sign up with an email at Paxton AI. And if you want to reach Tony and I you can just send us an email at Hello@Paxton.ai AI. And as we said, we’re really hungry for feedback on on Paxton or if you guys want to chat and chat about AI or anything else we’re always happy to.
Marlene Gebauer 48:04
And listeners, you can also leave us a voicemail on our Deacon review Hotline at 713-487-7821. And as always, the music you hear is from Jerry David DeCicca. Thank you, Jerry. Thanks, Jerry.
Greg Lambert 48:15
All right. Marlene, I will talk to you later.
Marlene Gebauer 48:19
Okay, so yeah,
Greg Lambert 48:21
for those of you watching on YouTube, we get a little extra for you that we’re not going to have on the audio podcast. And so I asked the guys coming out of left field here to give us an example of something that they think is cool. So I think Tanya, you’re going to share your screen. Yeah.
Mike Ulin 48:48
Greg’s way of making sure the software works.
Tanguy Chau 48:53
Quality control. So yeah. So there goes so. So this is the plaxen platform. And so you can see you can query curated databases here laws, rules and regulations. You can add videos or audio or deposition transcripts or connected to your databases, you can use a document comparison features. This is the the famous Boolean composer that that we talked about. And so you see we connect to these sources, starting with the administrative code of Alabama, all the way to this Wyoming State statutes. And so if for example, I am a I’m an investment manager and I want to understand what are the rules to sell under Rule 144 This would be my eye select my source. Here, I’m going to query the Code of Federal Regulations, we’re working on an updates where Tanguy searches through the entire body of knowledge. And then I simply hit submit. And what happens when we click Submit is passing reads very quickly and very rapidly, but very accurately through the 30 million documents and growing that are in the database. So you can, it does a semantic match around the language that is here. So the rules to sell under 144, and then retrieve the relevant information. And so you can see Paxton writing, and I’m going to increase my screen here, you can see the response that passage gives, and how every sentence that we have is properly footnoted, when you click on the reference material, you see where it pulled the information from. And so here, we’re connected directly to the Securities and Exchange Commission. We pulled this information from title 17. And as you click, you see the actual language of that Tanguy used to provide the answer. So this is an example of what we do for body of, of knowledge that we curate. In this version, you can, you can create a new folder, a new matter, and then drag and drop 1000s of documents here, and then adapt and, and adapt the knowledge base that Paxson has to your needs. You can also upload videos. And see for example, this is a video that that of the joint hearing about President Biden’s border policies in Arizona and you can query those videos, we have like a quick summary that we provide. And then you have a document comparison, and a Boolean search tool. Such as, for example, if I wanted to look at know, what are all of the rulings, please find me all of the court rulings needed in the last 10 years for someone that illegally operated a motorboat in the state of Hawaii for for example. And I wanted in, in Boolean query to use in in a platform like like Westlaw. This is the search query that we will provide. And this is how the search query is, is built. And so the user can simply just copy the search string and then use them on their platforms. So that’s that’s a quick overview of what Paxton the plaque detection platform.
Greg Lambert 53:01
Awesome. Wow. Very cool. Yeah. enjoy that very much. Thank thank you for sharing for sharing that with us.
Tanguy Chau 53:11
Thank you for giving me the opportunity. All right.
Transcribed by https://otter.ai