This week, we welcome back Kara Peterson and Rich DiBona of Descrybe to talk about the company’s rapid growth and its expanding role in legal research. Since their last appearance, Descrybe has not only built out new tools but also entered academia by joining the curriculum of more than 350 universities around the world. Kara reflects on her earlier career in legal education and how this new partnership feels like coming full circle. Together, she and Rich share how Descrybe is positioning itself to fill the gap left by other providers while keeping affordability and accessibility at the core of their mission.
A major highlight of the discussion is Descrybe’s unique approach to legal citators. Unlike traditional tools that often provide a blunt “treatment” of a case, Descrybe’s citator allows issue-level analysis and even introduces a “backwards citator.” This means researchers can see not only how later courts interpreted a case but also how the judges who wrote the opinion cited and treated earlier authorities. Rich explains the technical challenges involved in training their system on 30 million citations, while Kara describes how these innovations give researchers new storytelling and analytical power when building arguments.
The conversation also dives into the Legal Research Toolkit, Descrybe’s paid tier that offers a collection of tools designed for professionals who need more advanced case law analysis. While the company continues to provide free access to its core research platform, the toolkit adds features such as issue explorers and advanced citator functions. Kara emphasizes the company’s deliberately simple pricing model, which prioritizes trust and accessibility. At just $10 a month for non-commercial use and $20 for commercial users, the service is priced more like everyday software than the traditional high-cost legal research platforms.
The discussion moves into broader industry trends, including the wave of acquisitions by major players like Thomson Reuters and Clio. Kara and Rich note that while consolidation is reshaping the market, it also leaves space for new entrants to innovate. With data becoming the most valuable commodity in legal tech, Descrybe is building curated and clean datasets across statutes, regulations, state constitutions, and even attorney general opinions. Both guests highlight the importance of accuracy, data hygiene, and minimizing hallucinations, explaining how their closed-system approach helps ensure that results remain grounded in actual legal documents rather than speculative AI outputs.
Finally, the episode touches on ethics, recognition, and the future. Descrybe recently won the Anthem Award for Ethical AI, a nod to its safeguards against hallucinations and commitment to transparent data practices. At ILTACon, the team found themselves impressing not only potential clients but also leaders from larger companies who were curious about how such a lean startup was able to achieve so much. Looking ahead, Kara predicts the pace of change in legal technology will only accelerate, challenging law firms to keep up, while Rich warns of the commoditization of AI capabilities and stresses the importance of staying ahead of the curve. Together, they bring both humor and insight, reminding listeners that the legal research market is shifting quickly and that affordability, accuracy, and ethics will shape its next chapter.
Listen on mobile platforms: Apple Podcasts | Spotify | YouTube
[Special Thanks to Legal Technology Hub for their sponsoring this episode.]
Blue Sky: @geeklawblog.com @marlgeb
Email: geekinreviewpodcast@gmail.com
Music: Jerry David DeCicca
Transcript:
Nikki Shaver (00:00)
Hi Greg and Marlene, it’s Nikki coming to you live today from Rome, Italy. Just dropping in to let your audience know that Legal Tech Hub is holding one of its mid-law events on September 3rd in Washington DC. Now for those who don’t know, our mid-law events are pitched at law firms from 20 to 250 lawyers, and they have wonderful content for the entire day for managing partners, office managers, lawyers, about how generative AI can be used to grow your business, how to deploy it, what kinds of solutions are available in the market. So for those interested, check out our events on LegalTechHub, LegalTechnologyHub.com is the URL. Use the little drop down for LTH events and you’ll find the Midlaw event in Washington DC and you’ll be able to register. Hope to see many of you there, bye.
Marlene Gebauer (01:02)
Welcome to the Geek in Review, the podcast focused on innovative and creative ideas in the legal industry. I’m Marlene Gabauer. So this week we welcome back Kara Peterson and Rich D, also known as Rich BiBona from Descrybe Kara and Rich, it’s so great to have you back on the Geek in Review.
Greg Lambert (01:09)
and I’m Greg Lambert.
Kara Peterson (01:19)
Thank you. We’re so happy to be back.
Rich DiBona (01:22)
Super excited. ⁓
Greg Lambert (01:23)
We were calculating and it’s been what, nine or 10 months since you were on last, but you have been really busy since we talked.
Marlene Gebauer (01:33)
Gangbusters
right?
Kara Peterson (01:34)
Yeah.
Greg Lambert (01:35)
Descrybe has grown quite a bit since then, including into academia. So let’s just kind of start off, tell us what you’ve been up to. And I’d like to learn a little bit more about this push into law school since we last talked.
Kara Peterson (01:50)
Yeah, that’s great. And I don’t remember if I said this the last time I was on, but I actually worked in a law school once. I was the marketing director. I did. And I was the marketing director at Suffolk Law School here in Boston, which I’m sure a lot of people know they have some awesome technology, like innovative thinkers there and they have for a long time. So I feel like I got the bug when I was there. So I feel like this is kind of full circle. So, and then I worked in higher education for a long time.
But anyway, so this was at a tech show, right, Rich? We were at tech show when this happened, right? think you don’t know what I’m talking about. You don’t know what I’m talking about. Yeah, it’s going to be a surprise. we’re a tech show and which was really a great, a great place for us to be like, cause it was like the right market and all this stuff. And we met a guy named Doug Lusk who runs, he’s the CEO of the national society for legal technology. And he was scoping out.
and talking to different legal research companies, legal research providers there because he had a space to fill in his curriculum that goes out to about 350 universities. He does law schools and paralegal programs across the world. Actually, it’s like, I think, 11 countries. But anyway, so we started talking with him and we were first on the list for who would fill the shoes of, da-dum-dum, CaseText Because CaseText…
⁓ is no longer, as folks know, probably listening to this offering, free search. So yeah, so we are officially rolling out this fall into a legal curriculum that’s again, in 350 schools. And we have some pretty good company in that program and happy to tell you more if you’re interested.
Marlene Gebauer (03:27)
Yeah, I think we’ll, we’ll touch on that a little bit, a little bit later. ⁓ but I want to talk about the citator. So that offers issue level citation analysis. So, I mean, that’s really quite different than, than traditional citators. So, but, but, you know, how do you ensure accuracy, ⁓ in the ambiguous legal context? like, you know, mixed treatment cases, you know, are human validators part of that process?
Rich DiBona (03:56)
Yeah, I mean. No, no. Yes, Rich D that’s my I have my DJ name on there. We actually have advisors who are legal librarians and people at university and most of them are academics. So we’ve done a lot of testing with them, but it’s a lot of trial.
Marlene Gebauer (03:58)
Is it you?
Kara Peterson (04:00)
Rich D. the human.
Rich DiBona (04:20)
an error, like before we did anything, it took like two months just to get things right because what we did was just analyze, I think it’s close to 30 million citations to see what the treatment was. And then you find ones where it’s like, it cites a case, but it’s really like a parenthetical for another case. And so it kind of confuses the AI, like this thing was overruled and so it thinks it was overruled, but no, it’s the other case.
that overruled it. So you kind of have to just work through a lot of these edge cases. So it’s just a long and arduous thing. One good thing is we’re like a tier five open AI provider from all the money we’ve given them.
Kara Peterson (05:05)
Sam calls us once a week to say thank you for for how the money we’re spending
Rich DiBona (05:09)
Yeah, I But we have access to some of those advanced models that are super expensive to run. And so you could actually use, like you paste in things that seem to be ambiguous, like, and ask it, like, what would you say the treatment is? And I would do this against like several different models and say, like, what do think the treatment is? And then then you kind of just work on things. It’s just so
gradual, it’s not just, I don’t even know how to explain it. It’s just two months, three months, six months, whatever, that I just kind of became a big blur.
Greg Lambert (05:46)
I remember those days.
Marlene Gebauer (05:49)
Yeah.
Greg Lambert (05:49)
Let me follow that up a little bit. if I’m a researcher, a lawyer, a librarian and I’m on Descrybe and I’m using citator, what’s going to be the big difference that I’m going to see as I’m using it compared to other tools I may be using? What’s going to stand out to me?
Rich DiBona (06:08)
I think the main thing that’ll stand out is that you see the citator treatments in every search result. So you don’t have to click a button that says treat or, you know, to quote the trademark name, Shepardize anything, you could, you see that the treatments right in line with all of our search results. So it kind of tells you right at a glance what’s happening. So.
That’s because we pre did everything we’re not doing, calling LLMs when people use our platform. And I think that’s a big difference for all of our platform is, is when you make a request to Descrybe, we’re not calling an LLM and then giving you the results back from the LLM. We’ve, we did all that work upfront. So, that makes sense. Go ahead, Kara.
Kara Peterson (06:53)
Yeah,
so from the use, Rich obviously is the technical side, so he answers these things in sort of like the technical perspective. We’re also just back from ⁓ a big conference. But anyway, so for the user’s perspective, it has a couple of new features too that don’t, I mean, we’re not going to say they don’t exist elsewhere because that’s.
always not smart to say because we don’t really know, but there’s a couple different ways that our citation sort of process works, which includes the issue level citation and the backwards citator. So in some ways you have different ways to look at the same question. And of course, where the real differentiator comes in is when you’re paying for it, when you’re paying for your monthly costs. But Rich, maybe tell them a little bit about how the issue level and the backwards work, because that is pretty unique.
Rich DiBona (07:42)
Yeah, yeah, yeah. So, since we were going in and analyzing all these treatments across the 30 million things, we decided instead of just saying how the whole case gets treated by this citing case, so when one case cites the other, like how did it treat the whole case, we said what issues did it treat on and how did it treat those issues. So we found a lot of examples where it will follow
on some issues and then overrule just on one particular issue. it’s like it likes that case for some things, but not for this one thing. But in a lot of cases, the whole case gets overruled in the indicators. there may still be issues you can use from that case, but you know, you just it gives you more specificity in your research. Does that make sense?
Greg Lambert (08:36)
Yes, now talk about the backwards citator.
Rich DiBona (08:38)
Okay,
so the backward citator. typically the way a citator works is it’s how cases that cited the case going forward treated the case. But what we did because we had this data is when the judges wrote the case, what cases did they cite and how did they treat them? So you see what they were thinking in
in history. like with Plessy versus Ferguson, for example, if you look at the backward citator, it’ll have a whole bunch of cited cases about like segregation on railroad cars and separate but equal. And it follows all those things because that was the way things were done. So the backward citator is all these followed cases. But then,
going forward, the forward citator on that is obviously all overruled with Brown versus education So you kind of see what happened with the judges thinking at the time. It might be more for researchers, but it was something that we thought was really cool. So we just decided to give it to everybody.
Kara Peterson (09:46)
Yeah, because once you have that data set, like, and that’s what we’ve been playing with and a lot of the other things, like some of the legal issue analysis and all this stuff that’s in the legal research toolkit, like almost like endless thing you can do. And of course that’s easy for me to say, cause I’m not the one building them, but you know, there’s the kind of endless ways to parse and look at that data and, and, and understand it and find like deep legal research rabbit holes to fall down. And it’s, it’s pretty cool. And I have to say like, that’s what people seem to get pretty excited when they see how that works. Cause
in the past that was humans that had to do that kind of work and it was to do that kind of analysis and it was cross prohibitive so nobody did it.
Marlene Gebauer (10:23)
Right.
Rich DiBona (10:24)
And to the earlier point about not calling the LLM, like you could go down these rabbit holes in like one second. you just click around and see everything like as fast as you’re using Google or something, as opposed to click, wait 40 seconds. ⁓ click again, wait 40 seconds, all that. So.
Marlene Gebauer (10:44)
I’m just thinking about this backwards site citator. It’s like, it’s kind of like a storytelling mechanism where you can kind of see where breaks happened and then be able to explain, you know, in your argument why those, those breaks happened, or you can kind of look over time at a number of, similar decisions and sort of see, the paths that they’ve taken. I mean, that’s kind of my analytic hat coming off.
Rich DiBona (11:10)
Have you ever had anything like that before?
Marlene Gebauer (11:12)
Have you,
Greg Lambert (11:13)
doesn’t sound doesn’t sound familiar. We kind of did some some things like this when I was at the Oklahoma Supreme Court years ago, not this this complex.
Marlene Gebauer (11:23)
So Greg beat
you to it.
Kara Peterson (11:25)
Yeah, good job, Greg.
Greg Lambert (11:27)
We call ours the citationizer.
Kara Peterson (11:29)
There you go, yeah,
don’t sue us for copyright infringement, The citationizer.
Greg Lambert (11:35)
The running joke was ⁓ at the time there was a company called Lois Law and right after we came out with Citationizer, they basically did the same thing and called it the Loisizer.
Kara Peterson (11:46)
my gosh. I thought you were kidding.
Rich DiBona (11:48)
Because
our thing is called, our brief checker is called the citationator. So I thought you were doing like a riff on that. That’s so funny.
Kara Peterson (11:57)
We definitely all are like comic book nerds in legal tech. I’m sorry.
Greg Lambert (12:01)
Yeah, well, I had somebody who’s throwing cool ideas my way to and saying saying make this happen. So rich, I feel your pain.
Kara Peterson (12:08)
Yeah, yep.
Greg Lambert (12:10)
I want to talk about the toolkit, but before I jump in, Kara, you mind just kind of giving us a broad overview of the toolkit and what it offers?
Kara Peterson (12:21)
Sure. folks that know us will know that we have a free tool and we started out free. So for like the first two years, we launched in summer of 2023, which is like, you know, the Stone Age of AI at this point, right? And so we had a free search and we still do, but it became clear over time that, you know, professionals need obviously different level of tools. And the first thing that was always sort of pointed out in a helpful way, but also realizing if we wanted to have a
know, serious tool we’d have to deal with this was the citator itself, right? So, or something that could do the citation like that. it pretty quickly became clear that we needed to create a professional level tier. And so that’s exactly what we did. We created what we call the legal research toolkit. And it’s a collection of maybe 18 or so different tools that are meant for people who need more advanced case law.
know, research and analysis. And so that is what it’s about. we, with our sort of hope to be increasing access to the law, we put it out at a price point that we thought was very, very reasonable for what you’re getting ⁓ without compromising on the quality of the product itself. So kind of a win-win situation was the
Greg Lambert (13:31)
And I was reading in leading up to this that you’ve had a good number of record setting, guess, number of paid signups since June. And so what are the toolkit features that are driving the conversation and how are you balancing the monetization with the mission that you have about democratizing legal research?
Kara Peterson (13:55)
There’s some ritual, why don’t you tell about the tools? Cause those are always fun. And then I’ll talk about the other stuff, which is fun too.
Rich DiBona (14:01)
So I’m actually not sure what is driving people, what tools specifically are driving people to sign up. We know that people like we have an issue, legal issue explorer where you could just start typing in legal issues like maybe like retaliation, document retention or something and it’ll show you all the cases that have that as an issue or
you know, anything you could type in and it’ll bring you up cases. have the citator. think it’s hard to find a citator at this price point. So that might be a thing too. What we’ve heard anecdotally is that there’s a bunch of lawyers who they have a need for legal research, but they can’t justify paying for the $500 a month thing if they only need to do a
a couple times a month or something. they, you know, they use ours and then, you know, yeah, what do you have to add to that, Kara?
Kara Peterson (15:03)
Well, okay. This is, yeah. So this is again, why the engineers are usually like not allowed out of their rooms. Because when you say, when someone says, are people adopting your product? And you say, we don’t know. Like that’s probably not the best answer. So, love you and all that. Yeah. Love you and all that honey. But yeah. Okay. So what we do know. So one of the reasons we
Rich DiBona (15:05)
Kara always bails me out.
Kara Peterson (15:26)
don’t know at that granular level is because we also don’t track our users in the way like a lot of tools do. So we really kind of want to give people the freedom and the space to like use our tools without feeling like they’re going to be getting like, you know, monitored every second and you know, we’re tracking them with cookies. So that’s part of our value proposition is we’re very much like, Hey, do what you want to do. Even our pricing, I probably shouldn’t admit this.
Our pricing has two tiers and if you’re using it for non-commercial use, it’s 10 bucks a month. If you’re using it for commercials, 20 bucks a month and you know how we check, we don’t. We’re just like, whatever, you know, we trust you. It’s like, you’re gonna game the system and you know, you lose $10, like we’ll survive. So what we do know is that people are converting like at a scale that’s like increasing really rapidly because people do find our case, our case summaries through Google, which Rich can explain.
But what we hear over and over and over isn’t necessarily which particular tool it is. It’s just how easy and intuitive and simple they are to use. Like it doesn’t mean maybe what the information that they get back is simple. Lord knows that like legal research is not simple, but it’s so intuitive and so quick and fast and easy to use that I think that is the part. So it’s a combination of between the price point, the ease of use.
And then the quality of the output that I think makes the magic sort of happen. And that’s the marketer’s answer.
Marlene Gebauer (16:52)
And it’s a good answer. Okay. So, you here’s another opportunity to, to market. so our, our friend Bob Ambrogi had some very nice things to say about Descrybe lately. So I was hoping you could tell us a little bit more about that.
Kara Peterson (16:54)
Thank you.
Yeah, so Bob’s awesome. Like he’s, he’s another Boston guy. So we’re, up in Boston.
Marlene Gebauer (17:16)
But you know we’re a little nicer. ⁓
Kara Peterson (17:19)
You’re
totally nice. mean, I’ve never been on his podcast. Let me put it that way. So you already beat him. Bob, you’re out there listening. I know. I had better. But no, so he’s, he’s, he’s great. And of course, like, ⁓ you know, Boston thinks it’s this big city. Let me tell you, it is not everybody is like a small place. We all know each other. So anyway, so Bob’s been great and he’s been following our journey from the beginning. And you know, we were in a startup alley and all that, like at the tech show, we had a very funny.
Marlene Gebauer (17:25)
You’re gonna get an invite like that. Invite as soon as this comes out.
Kara Peterson (17:47)
sort of valley experience, which is a whole separate story. But ⁓ so yeah, so you know, but he’s also a journalist, right? And you know, he’s a very serious journalist. And a lot of times in industries like this, or any really industry, you know, there’s people who write about stuff. And that you’re not sure what you and you know, because you do a show, like sometimes you can take it on as face value. Sometimes you’re not sure because it’s like as a person, you know,
whatever more in the marketing space than the journalism space, but like Bob’s a real deal. I don’t think he says anything he doesn’t think. when he gave us, when he covered the release of the toolkit, he gave us a really nice sort of compliment in that, ⁓ you know, we were poised to fill the space that CaseText is leaving in the market in that, you know, there isn’t affordable.
another option really for affordable case law research like that. So that was really nice to read because I who doesn’t want to be compared to a company that got sold for $650 million? But even more importantly than that is that, you know, it’s about the need and the need is not gone. The need is still there. And so we’re really happy to be able to even do a small part in filling that need.
Marlene Gebauer (18:59)
So I want to follow up on that then. Are you exploring other areas to fill that gap? beyond case law, like statutes and regulations, other things?
Kara Peterson (19:13)
Yeah, I laughed just because poor Rich. Yeah, Rich, why do you… This is probably why he’s like…
Rich DiBona (19:18)
I know
Marlene Gebauer (19:19)
talk to
Greg Lambert (19:19)
has grand plans.
Kara Peterson (19:21)
Poor
guy.
Rich DiBona (19:22)
I talked about like the six month dark place I went into for the site data stuff. So we decided that what we need is beyond just caselaw We’re actually pulling in where we’re along in the process. We’ve done like the 10 biggest States and but we get statutes, regulations.
Marlene Gebauer (19:28)
You’re back in again.
Rich DiBona (19:49)
state constitutions, court rules, attorney general opinions, and there’s one more. But session laws. Yeah, so we’re pulling all those and making them all searchable as well so that you could kind of search across jurisdictions or types and just find things like if you search for toxicology rules or something, it’ll show you all the rules about.
Kara Peterson (19:57)
Trial, was it a trial? Session laws.
Rich DiBona (20:16)
how to deal with toxicology or nursing homes or amusement park rides or however you would normally search for that. So it’s just a way of getting. I know, just, that is, toxicology and carnival rides probably goes.
Kara Peterson (20:24)
That’s a strange combination.
in a nursing home. Yeah, I don’t know.
There’s probably a case, if there’s a case that exists about that, you can find it and Descrybe.
Rich DiBona (20:40)
Mad Libs. Yeah, so we’re working on getting all of that in because that’s really hard to find too. And I’m not sure how many other platforms offer that that aren’t, you know, a lot of money. So, yeah. So we’re working on that and we’re doing some cool things to help make that really user friendly and easy to use. And then the key is
You know, we don’t plan on charging any more for it. We’ll just include it. What we realized that we were at ILTA is I was thinking about our pricing, like the $20 a month. It’s like the cost of a Microsoft Office license or subscription. So, you know, 240 year or whatever, which is a normal price outside of legal tech. Like 20 bucks a month is like a normal price for things. So we’re like charging.
based on cost rather than based on how much we can charge.
Kara Peterson (21:38)
Yeah. And just to add to that, like, I do think because it’s been, I mean, AI is going to change everything. We know this, right? That’s obvious, but the cost per user and all these things is so artificially high in a lot of ways and in the legal tech space. And I think it’s almost like that boiled frog scenario where people just don’t recognize it. Maybe don’t even think about it. Cause when, when we’re saying that to people like, yeah, you could get Microsoft Word or you could get like Canva or whatever. It’s like, yeah, that’s a normal price anywhere else.
And I’m not saying, know, Canva and like some very, you know, serious like LexisNexis or Westlaw, you know, thing are comparable, but it’s just, you know, there’s just a very, very wide gap there.
Greg Lambert (22:17)
Yeah.
You talked earlier about CaseText and getting bought up by TR for $650 million. We also just had ⁓ vLex that had acquired Fastcase and now vLex was acquired by Clio for a nice cool billion dollars.
Rich DiBona (22:38)
You gotta do that with the Dr. Evil thing.
Greg Lambert (22:41)
Yeah,
she’s gonna do the Austin Powers finger. how does this, I think, maybe open up an area where there may be an area to fill the innovation gap for affordable legal research out there? how are you looking at this? Do you see that as a gap that needs to be filled?
Kara Peterson (23:01)
Absolutely. it’s interesting because you kind of can do the Fastcase CaseText Descrybe thread, right? Like, and whether or not we end up, you know, anywhere near that scale, you know, TBD, but it’s kind of a neat narrative. And so there’s always somebody coming in to fill that space, right? And it probably will continue no matter what happens. But I think the lesson it gives me, and of course, I care a lot about the access to justice, but
The other lesson that I’ve been hearing from the market from this is just how important good data is right now. this legal our timing is almost perfect where it’s like this case law research, this kind of clean data sets of really good curated case law and other things that Rich was mentioning before are really a hot commodity right now.
And that’s pretty exciting, to be honest. So yes, it’s needed. But it’s not only needed for the access to justice or sort of like open access to information, but I think the big players, they’re also all realizing it’s quite critical to their, to their, I don’t know, to their business as well. I don’t know what you guys think about that. You’re you you deal with this all day long. Does that strike a tone?
Greg Lambert (24:19)
Yeah, I mean the data we’ve we’ve had multiple guests on in a row that you know talked about You know data is the new oil data is the new electricity data is the new whatever And I think that’s one one of the things in fact the the question that we’ve been asking them over and over again is yeah, you know
Yes, the vLex’s the CaseText have been bought by the TRs and the Clios of the world. But since data is going to be so important, does that open up the TRs and the Clios of the world to now be bought up by a Microsoft or an OpenAI? That so, you know, it’s definitely something I think you’re seeing the processing power.
Kara Peterson (24:55)
⁓ interesting.
Greg Lambert (25:07)
is improving, but the thing that has been kind of capped is the data. And so that’s, I think it’s something that there’s definitely a need for the data to be out there. So, Marlene, any?
Marlene Gebauer (25:25)
Yeah,
I mean, I think that the challenges organizations face with data hygiene that they’ve faced for years, I think it’s catching up because, we want, they want to use, some of these new tools on the data, but, if they don’t have it structured and clean, then they, these wonderful, high tech tools are not as effective.
Kara Peterson (25:50)
In fact, they cause harm. They can make things worse.
Marlene Gebauer (25:55)
And
so I agree with you 100 % that data is the focus.
Rich DiBona (26:02)
There’s also this user expectation, I think, like when you go into these tools because of how fast, like there’s so many amazing tools out there. was at ILTACon I was watching demos of some of the tools and you could, as a software engineer, could tell how much work went into these things, but there’s an expectation from users that you could just go to one place and get everything now. And so if you’re missing that one part of it, it’s like,
The companies probably feel pressure to be like, I got to go get that piece because I don’t, can’t offer everything in my.
Marlene Gebauer (26:38)
Even though they don’t solve everything.
Kara Peterson (26:40)
Right.
Greg Lambert (26:41)
Yeah, just try to get Copilot to search your Outlook inbox.
Rich DiBona (26:45)
God.
Marlene Gebauer (26:47)
You
Kara Peterson (26:49)
Yeah, but it’s, it’s, it’s true, but like, is it just actually like the data? again, Rich is the data person has been working like giant, you know, data sets for his whole life or whatever. But it’s like the, the harm, like back to the harm thing. And that’s the thing is if your data isn’t clean, and I think no offense to anyone who’s self-taught or like anything like that, or who’s like, you know, attorney turn program or like the kind of complexity that some of these giant data sets have and just the nuances.
is not for the weak of heart, right? I mean, you could say like the complexity and these things. And it’s very easy just like in statistics, right? You can look at statistics and tell any story you want. It may or may not be right, just depending how good you are at understanding statistics. So I think the data is kind of similar. Like you can mine all kinds of information from it, but if you’re not getting the right outputs back, you’ve just set yourself back even further than if you didn’t have it in the first place.
Marlene Gebauer (27:42)
Hmm.
Kara Peterson (27:42)
I don’t know, Richard, that makes sense. Now you can make fun of me for my marketing answer when I shouldn’t be talking about data.
Rich DiBona (27:47)
I don’t dare make fun of you.
Marlene Gebauer (27:49)
Yeah
Kara Peterson (27:50)
Smart
man.
Greg Lambert (27:51)
Well, I think that transitions well from AI to data and making sure things are clean. But one other thing that we like to talk about a lot is the ethics around it. you guys, in fact, I think this was announced kind of between the time we recorded last time and the time we published that you won the Anthem Award for Ethical AI. And so I want to…
focus in on that and ask about the big issue. Since then has been a lot of hallucinations. what safeguards, and I guess Rich, since you’re the engineer on this one, what safeguards are you putting in to make sure that tools like your citator are actually retrieving
and grounded into the actual documents themselves and are not just making stuff up.
Rich DiBona (28:46)
Yeah, I sometimes I am thankful that we’re not a document generation platform. Like we don’t create any documents and we don’t call the LLM like I mentioned before. So the way we built our thing was just by giving concrete pieces of data and having the AI operate just on that thing. don’t go say augment it with this or go figure out.
things that are similar to this or whatever. It’s like, here’s some data, do this thing to it or figure out this thing about it and then give us the results. So it has managed to keep our hallucinations but we do have a brief checker thing, which you could put your briefs in and it’ll tell it, it’ll find hallucinated citations for you and also the treatment of them because even if you’re
perfect in yours, you might receive one from the other attorney. So you need to check what you receive too.
Kara Peterson (29:45)
Yeah. So the way I was going to say like, and you know this, you talk about this all day long, but if anyone says their tool or platform hasn’t had zero hallucinations, you should say, yeah, yeah, bye bye. Thank you. Right. And walk away as fast as you Exactly. Exactly. And so, but because this is this closed and Rich can explain it better, but this closed system where we’re not like, Hey, ChatGPT you know, or whatever, like calling out into this.
Greg Lambert (29:58)
It’s a feature, not a bug.
Kara Peterson (30:10)
vast universe, like we’re dealing with this closed set of information as it’s queried, right, Rich? I mean, you can explain that better. Like it’s, it does minimize the potential for really weird stuff to happen.
Rich DiBona (30:21)
Yeah, by the time it gets to the point of being in our platform, we’re basically doing database queries against previously generated stuff.
Marlene Gebauer (30:31)
Okay, so this is the answer to this question is what everybody’s been waiting for. know, ILTACon has been all over LinkedIn and the news. that just wrapped up. I know you guys just got home. But you know, what major announcements did you make there? You know, what kind of reception did you get from the attendees while you were there?
Kara Peterson (30:51)
Okay, so ILTACon was a trip. Like that was something. so we’ve only done, and in a good way. So we’ve only exhibited at two conferences. So we did Tech Show in Chicago, and then we did this one. And so in this one, we were in the Startup Hub. And it was really actually a lot of fun because one thing, if any entrepreneurs are out there, like these things are worth doing because not only do you meet like potential clients or like bigger, you know,
partnership opportunities, but you meet the other startups like letter in your cohort. And it’s really kind of fun. You start to make like friends and you know people and you start to talk about like ways you can work together. But ILTACon was, I’ve been to a lot of conferences and other sort of parts of my life. And I have to say, this was like one of the best run things I’ve ever seen. Like at least from a, from a startup hub perspective, like it was so turnkey and easy and we even couldn’t like put our sign. was like, it had to be exactly right. And everything was really good, but it was.
like a lot of food too, like so much food. I didn’t, I know, and this is probably not the part you really cared about as much, but I was impressed how much food there was and that even got, even that drink tickets. I know, I know we paid for it, but yeah, so you’re right. So I’m like, just feed me and I’ll be happy. No, but I was really impressed by just the organization of it. I was impressed by the kind of,
Marlene Gebauer (31:55)
I always
Greg Lambert (31:57)
to the price, by the way.
Kara Peterson (32:10)
just the social interactions, it was very baked into the conference to have, and probably had to do with the venue, which was this place called the Gaylord National Harbor, whatever it was. Yeah. And it was just, it was almost like you’re inside this little space bubble. was saying to Rich one night when we were sitting there having a coffee or something, I said, this is probably what it’s like, we’re going to be like when we live on Mars. Like you’re just inside this giant glass bubble and it’s like a village and you don’t have to go anywhere. So it was really nice for like the social aspects. Obviously all the
the different vendors are there. ⁓ But it was definitely different than TechShow in that it was not a lot of, at least for us, like lot of like end users coming around. was like the big people who were thinking about it in big scale, like either for big tech or excuse me, for big law or things like that. But towards the end of the conference is when it got really interesting for us is that’s when the big guys kind of came off the mountaintop.
and came down and started checking out what all those startups were doing and that got really, really interesting.
Rich DiBona (33:11)
A highlight for me was just having some people from just some of these larger companies come by and see what we’re doing. It was really cool. was no one was acting like, what are you guys doing? And we’re going to stomp you out. They were just really nice about the whole thing.
Kara Peterson (33:32)
I genuinely genuinely interested in every time of course somebody from one of the bigger companies would leave talking after rich gave him a demo idea what they say what they say and it was almost always the same answer is that they they they don’t understand how we’re doing this and like they keep going he said their jaws are kind of like hitting their chest and going how did you do this how did you do this so so we’ll see we’ll see what that means we’ll see how many phone calls we get this week next
Greg Lambert (33:56)
Well, I’m curious now, what is your elevator answer to how are you doing this? How do you give it?
Kara Peterson (34:03)
Okay, Rich, go.
Rich DiBona (34:05)
No, I mean just…
We couldn’t have done it a couple of years ago without AI. you hear about all the vibe coding and extra programming and all that, but if you already knew how to program really well, it could really kind of help accelerate things. And then also some of the things like we talked about before, like analyzing for case treatments you couldn’t have done without AI because you needed, you know, thousand human editors. So that’s how we’ve done it is
Kara Peterson (34:34)
I mean, that’s it. Yeah, it’s Rich D. Like, I mean, it is and he’s going to be all humble because he’s like that. like, the kind of things he built like, blow people blows people’s mind people who are very no, it’s true. And it’s I’m very proud of you because the stuff you build is incredible.
Marlene Gebauer (34:36)
And with secret sauce.
Greg Lambert (34:55)
All right, take the compliment, Rich.
Kara Peterson (34:57)
Yeah, just take it.
Marlene Gebauer (34:58)
Thank you. Thank you, honey.
Kara Peterson (35:01)
That’s right.
Greg Lambert (35:02)
All right, well, I looked back and somehow you escaped our previous interview without us asking you the question.
Marlene Gebauer (35:11)
Do the crystal ball? What?
Rich DiBona (35:12)
asked us to
at the end, I forget what it was.
Greg Lambert (35:16)
We don’t have we can’t look back and say, know, you were right or wrong in this. So Next time you by in 10 more So I want to you know, we ask everybody our crystal ball question usually So what change or challenge in the industry? Do you think we’re going to be facing that we need to start preparing for now?
Marlene Gebauer (35:24)
We gotta set the baseline.
Kara Peterson (35:39)
my facetious and obnoxious answer is we don’t know if the law will even exist in 10 months. So we’ll have to just, you know, adapt to that. If we have no law, then I guess we won’t need legal tech anymore. That’s not what you really wanted me to say. I would say that, I mean, and we were joking about it at the top of the call, but like that, you know, two years ago felt like a thousand years ago and two years from now, or 10 months from now is going to feel like a hundred years from, you know,
Marlene Gebauer (35:51)
Thank
Kara Peterson (36:04)
So I think the pace of change is just going to continue to accelerate to an extent that we have never seen and that that’s going to be very hard for all industries, but it’s going to be particularly difficult for legal because quick change and rapid innovation is not necessarily, I would say, a core strength for many in the legal industry. And feel free to disagree with me.
And also just the pace of change means there’s talent shortage and it’s hard to hire people to come into your firm, especially if it’s a smaller firm or even a mid-sized firm and help you with these transitions because labor is scarce. So I think we’re entering a rapidly increasing change with ⁓ some very serious competition for resources. I think that’s going to be happening.
Greg Lambert (36:59)
Rich, anything to add? does your crystal ball say?
Rich DiBona (37:02)
Well, what I think about a lot of times, and this is why we started, like, we’re always trying to think ahead of what we can work on next because…
You know, these models, even though people are criticizing GPT-5 and all that, it is super powerful, at least for some of the analysis stuff. Maybe it’s not as good at being a boyfriend or girlfriend as people wanted, like the chat part, but the programming part and the analysis part is super powerful on this high reasoning. So I think about things getting commoditized a bit and like, can you be sure that’s something that you put in to some of these advanced models?
will be that much worse or different than what you put into some of these advanced platforms. So that’s what I think about.
Greg Lambert (37:49)
Nice. ⁓ Kara Peterson and Rich BiBona, I want to thank both of you for coming back. I know you’re exhausted from your trip down to National Harbor, but thanks for taking the time to talk to us.
Rich DiBona (38:01)
Thank you. You always get a little bit of spice when you have us on because you never know what going to say.
Kara Peterson (38:06)
That’s right.
Marlene Gebauer (38:06)
Absolutely. And we like bring the spice. We that. In Texas, we like spice. So that’s right.
Kara Peterson (38:10)
Yeah, I feel like.
Well, thank you. It’s so good to see you guys.
Marlene Gebauer (38:16)
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 (38:26)
All right, and Kara and Rich, if listeners want to learn more about what you’re doing at Descrybe, ⁓ where is the best place to point them?
Kara Peterson (38:35)
Sure, Descrybe.ai and that’s with a Y. Or you can always find me on LinkedIn, which is apparently my second home or maybe my first home. So always look for me there and I really love to connect with everybody. So please reach out.
Marlene Gebauer (38:49)
Okay, so Descrybe with a Y. And as always, the music you hear is from Jerry David DeCicca Thank you very much, Jerry.
Greg Lambert (38:57)
All right, thanks, Jerry. Thanks, everybody.
Kara Peterson (38:59)
Bye.
Marlene Gebauer (39:00)
Bye.