The ILTACON Wrap-Up

Episode Notes

Transcript

Joe Patrice:

Welcome to another edition of Thinking Like Lawyer. I’m Joe Patrice from Above the Law. I am joined in this studio by nobody today because I am going a little bit solo. Kathryn was the only one of the regular hosts last week, so this week it’s just going to be me. But this is a special episode brought to us by our sponsor, Smith AI. We’re going to be talking to Maddy Martin, who’s the senior vice president for growth there, who is also someone that I have been going to legal tech conferences with for the last decade or so. So we talk all the time and we kind of though maybe a good idea for a show would be since we often get together at the booth and chat about what we’re seeing, we just bring that to you. So we’re going to let you in on what an insider conversation would be like on the exhibit floor.

So I’m joined by Maddy Martin from Smith AI. How are you?

Maddy Martin:

I’m well, Joe. How are you?

Joe Patrice:

Good, good, good. We both go to a lot of these conferences over the years and we just got back from a big one. You weren’t there as long as for the whole time. I was unfortunately stuck in the Nashville Gaylord bubble for an entire week, but we just were at the International Legal Technology Association annual convention and we thought we should catch up on the legal industry as a whole after having seen a wild event. And yeah, so I guess my first question is, you were there, what were your impressions of the Gaylord Nashville?

Maddy Martin:

Of the Gaylord facility itself, it was beautiful. I though that it was decidedly not industrial or corporate, so that was a pleasant surprise. It was actually like being inside of a Vegas hotel where you expect the little Venice boats going down the little river, right? I mean, almost like Disney-like, but that set the tone, I think, for a more casual conversational nature of the event, which was necessary because there was just an unbelievable amount of money that was being spent on the booth setups and the demos. There was a mini stage almost auditorium-like inside of Harvey. And the amount of expenditure I think is reflective of the amount of expenditure that’s happening within these companies to build really exceptional products that lawyers are not only exploring, but demanding at this point. Now, I normally go to the solo small mid-size firm conferences. So going to a conference like Elticon was a whole new world for me.

I mean, I’ve been in legal tech for 10 years now, and this is the first time that I entered a conference and I didn’t even realize there was software for certain aspects of these big corporate legal teams. Now, when it comes to Harvey, we do see a lot of corollaries in the small firm world. So that was really my lens for the event is where do I see this technology going and how soon will it trickle down into the small firm space? So for example, Pacer Pro and Harvey, their partnership, that feels very similar to me to Clio Docket and what was just released on that front. So we’re already seeing that there’s not so much time between what the big firms are doing and what is being released to the small firms. That whole time horizon is condensing. It’s collapsing quite significantly.

Joe Patrice:

Yeah, I think that’s right. And I think one of the takeaways I had from it was obviously a lot of money because AI is a bubble, but that’s how Harvey built the Harvey wall. And you mentioned them, I’ll go a little bit more. They had 24 booths all together, half of a row, and they built this wall, an homage to the Berlin Wall basically that blocked everything off. It was wild. And of course they had a lot of money there. Had Lady A do their party. Lagora similarly had Cheryl Crow do their party. So there was a lot of money. That said, and big law firms can afford that kind of money, but one of the takeaways I had was I was hearing more rumblings, even though these are still largely talking to big law firms about cost consciousness. A lot of document management systems were talking about ways in which they enrich data such that less powerful models can get the same results.

I heard Thomson Reuters unveiled that it has its own in-house language model now that is cheaper and more efficient. So to what you’re saying about how this stuff trickling down, I felt like it was almost happening partially because the big Law side is also getting cost conscious. They’re starting to build these cheaper and cheaper options that I think in the next six, seven months we might start seeing in the small world.

Maddy Martin:

I think the number one corollary that we see when companies start to get cost conscious is how do they mitigate that with quality? Partnerships are often the path to that. So when we think about what is Harvey invested in, the UI is really nice, the flow between tasks and the overall spaces that they released is really nice. It feels like project management a lot more than we’ve seen in the past with case or matter management, with practice management. A lot of those tools can feel somewhat disjointed, but what stood out to me with Harvey in the demo that I saw was how harmonious the picture is, how complete that picture is within those spaces that they’ve released. And then when they add in the integrations, it also feels like this singular experience, very smooth without having to develop things from scratch. You bring in a tight integration and immediately your world expands and your workflows become simpler and more trustworthy.

Pacer Pro is obviously very trustworthy. So things like that in contrast to the acquisitive path or the development path, we can see how integrations, if they’re aligned, can bring products to market very quickly and efficiently. And then you also expand this market based on both of your client bases. So that was very wise to me. And Harvey then got to spend its time building out a space where the UI is really beautiful, where it’s highly functional. I mean, they want users to actually want to use it. I mean, that’s a big issue with software today is clunky software doesn’t get used or it’s very underused. So when we think about cost consciousness, I immediately look at integrations as the number one pathway to getting the best of both worlds.

Joe Patrice:

Yeah. I mean, one of the more clever catchphrases, and I’m unfortunately not going to remember what the vendor was who had it, but they had the Run DMC logo that was but in Run MCP, that kind of dictated –

Maddy Martin:

I saw their booth.

Joe Patrice:

Yeah. And they weren’t even American. I was like, “Oh, you tapped into my perfect ’90s cooked brain.” But yeah, MCPs was kind of the whole conversation, lots of partnerships, lots of that, which was interesting because six, seven months ago, I felt like everybody was talking about how they’re building their user experience and you’re going to want to access everything through us or the connections would all happen, but we are going to dominate the entry point. And this time around it seemed like with the possible exception of the Harvey Lagores of the world who kind of set atop that world, everybody else was like, “We have our own user interface. You might really like it, but if you don’t, you can go through Harvey, you can go through La Gore, you can go through Thomson Reuters, whatever it is, however you choose. It seemed like almost a surrender of the idea that anybody is going to get exclusive control of how a lawyer gets at stuff and a new focus on owning the stuff underneath.

You can come at it any way you want, but you’re going to need our document management system because it does the best job with the data or our legal research or our eDiscovery platform sort of took off a little bit more, which I thought was interesting from a big Law perspective because I think it speaks to how they right now at least are in a buy it all and let the lawyers sort it out kind of mode. They can buy everything and people will make the choices they make. Small and solo, that’s a different equation. You don’t necessarily have the ability to buy everything and see what each individual lawyer wants. And I do wonder if this move towards this kind of devil may care move how that’s going to trickle down to the smalls who can’t necessarily get Harvey or Lagora because they’re higher priced.

What are they going to lean towards? And is it going to be one of the actual underlying AI people? We have Gemini and OpenAI even and definitely Claude building legal, building it directly.

Maddy Martin:

Yeah. I think that Claude just a few months ago made it very clear that they’re interested in being in the legal market. But I do think that in small firms, there is high friction to switching tools because there is no dedicated CTO, CIO, et cetera. So it’s the onus on the managing partner, maybe on a COO to install new tools and then get adoption. And those relationships are all very close and personal in a small team. Choices like that can feel a lot more personal and people can take it personally if their choices aren’t respected. So it is high friction in a small firm to switch your software. Now in those smaller firms, there is the luxury of month-to-month pricing for a number of those companies. So the ability to switch is there, which I think is good for the market in that small firm space.

But what I’ll also say is there’s this divide I saw at Elticon between those who are really eager for partnerships and others who are more acquisitive or much more heads down in their own space. Like the Litera press briefing where they showed, I don’t know, 30 different products that they had produced in the last six months or whatever it was. I mean, it was an unbelievable set of new features and products and they admitted how large that corpus of work was, but they went through it so quickly that you couldn’t really get a taste for what each thing was capable of. And you do wonder how capable is it if you’re going to flip me through it at 3X speed through the screenshots and sort of the narrative. Now, what they were more focused on is using the data in-house to put forward the best RFPs, to put the most suitable attorneys into the RFP who have the best experience for that particular RFP.

So when we think about their user base, maybe they don’t need to show the how or the UI so much as just letting their users know that this is something you’ll be able to do because it’s been requested for so long and people are so familiar with those requests that they see it flash on the screen and they immediately understand this is what it’s going to do for my process. But what I took away from that talk in contrast to some of the other ones I went to was look at how well we’re tapping into your data and how much intelligence exists there that can make your work easier to win and then work these RFPs that you win.

Joe Patrice:

All right. Well, let’s take a break right there. We’re going to hear a message and be right back with this conversation with Maddy Martin. And welcome back. All right, continuing with our conversation with Mattie. A lot of talk about getting the best out of the data. That was coming up a lot in things that I was hearing too. And sometimes people didn’t want to necessarily say it that way, but that was what they were selling. They were like, “Oh, we capture all these processes and then it’s easier too.” And I was like, “So you’re enriching the data in a way that makes it easier for…” And I think a bit of what was looming over this show from my perspective was Token Gedden, that eventually they’re going to have to pay for this stuff that has been largely subsidized by the people who make it and what happens then.

And getting the best out of data, but I would say Legal Week, which was the last big thing I went to in the spring, the conversation was a lot about getting best answers, most accurate. We get the best, best, best, the best model at all times. Now much more, we get right answers, but we invite you to do more on your end. We enrich the data so that you can get cheaper models. It’s interesting that that was kind of a shift. And I think that shift, what came with it was a renewed focus on we need to clean up the data so that a model doesn’t ping around 15 million things while it’s trying to find the right answer. We build something good and then it knows, oh, I can figure out it’s here. At the end of the day, after all of the promises, it still is about organization underneath everything.

Maddy Martin:

I think it’s also about two other things that are related in a lot of ways to the token usage and billing, which is the audit trail and the attorney’s responsibility for involvement and at which points and at what supervisory level in a particular case. So what I didn’t see is here’s how we’re going to help you inspect the data. What I saw was here’s the output and it’s implied that you review it before you send it, but we’re going to draft the email for you and you’re going to see what changes we made if you choose to dig into it. But there’s a lot of hand waving in my mind right now around what responsibility attorneys actually hold from the point of view of these companies developing the tech in each step of a legal process. So at what point, because just as an aside, I think we all see in the news how involved Anthropic and other companies are in the sort of supervisory nature and governance of AI and policies that are being set forth in the US and in globally.

But it’s not yet coming into the conversation led by the software companies. So what I’m concerned with is that that conversation is going to be brought to them. And I think it would be wise for them to bring the conversation forward first because it’s going to be a less defensive posture if they can get out in front of it much the same way that Anthropic has tried to do in other companies. Now it’s not perfect, but if you can say, here’s the audit trail, here are the ways that you can inspect the work that’s being done, here’s the master prompt that you could influence if you want to adjust the certain tolerances, here’s how much visibility you can have into the work that’s being done. Do you want citations at every line or summarized after a body of work, yada yada. How much control do I have to audit and supervise the AI in the way that I’m comfortable?

What tools do I have to do that and how are you giving me the dials to turn? I didn’t see so much conversation around that. I wasn’t there for the whole time, so maybe I missed some of those conversations. But on the floor of the expo hall and in the sessions that I did attend, there was a lot of talk about implementation and risk mitigation from an operational and CTO standpoint, but not really from a standpoint of mitigating risk of citations.

Joe Patrice:

You didn’t really miss it because it still is the blow off line human in the loop, which as I always say, if you text somebody that you’re going to keep them in the loop, that is the prelude to you then ignoring them later.That’s what

Maddy Martin:

You

Joe Patrice:

Say right before you don’t invite them. And that’s how I think a lot of this human in the loop conversation has gone. They say, oh, well we create an audit trail. And if at the end the partner wants to read 300 pages of logs, they’re welcome to it and that’s not going to happen. I’ve said a few times that I think the AI can keep getting faster, but what’s not getting faster is the human brain. And we have to start thinking about workflows in a way that caters to that. You can’t run a workflow that’s soup to nuts. You put in a complaint and you get the thing at the back end because without seeing individual steps and having a moment to reflect on them, the answer is you don’t read the audit log. And it is still in that mode, although I was a little heartened.

I did talk to one vendor who was showing off the ways in which they’ve kind of built the interface so that you can zero in and jump in and intervene in certain parts of the task that are identified, which I thought was useful and choose not to in other ones, which is also part of the job. And so I thought that was useful to see that people are starting to think about it. I also heard somebody talk about building their model such that they can get from here to here in a few hours, which I know other competitors say they can do that in minutes. And I thought it was interesting. And when I pressed them on it, they said, “Well, yeah, we do it over the course of hours because there needs to be time for human course correction and interaction in the middle of this flow.” And I was like, “Oh, that’s encouraging.” We’ve been driven by a lot of the AI can do everything conversation for a couple of years.

And there was a little bit more of a what is the human’s role conversation, but it was very much the minority view. It still is largely about we have a magic box that gives you legal answers. And worryingly, I worry that’s going to get worse before it gets better just because we’ve seen legal become the target of all these venture capitalists who no longer have a profitable way to get into the underlying AI game. And I feel as though it’s those investors who really drive the we can replace all people conversation. And these legal vendors I think know that that’s not the path, but they’re indebted to people who want them to sell the message that they will replace people.

Maddy Martin:

There is a conflict there, and I think that’s something that we should talk about more as an industry. How do we resolve some of that conflict? One metaphor that I was kind of coming up with while you were talking is, are you in the driver’s seat or the passenger seat of a car and at what speed? If I’m in the driver’s seat, then I can tell you what turns I made. I’m very aware of the path that we’re taking to get from point A to point B. If I’m in the passenger seat and you ask me, well, what turns did we make? I’m probably looking at my phone. I’m probably looking at and dwelling on one storefront that I just saw and I didn’t see the turn you just made or the street name. Part of the risk and the inefficiency right now in using AI is that if we get too far removed from the process because it’s moving and we’re not paying attention to how it’s moving or what it moved past, what way points there were, then to catch up is extremely inefficient.

We have to go backwards. We have to see how it was built. We have to see what assumptions were included. And the less that we’re the driver, the more catch up work that we have to do because we don’t have that muscle memory of having made that right turn. So to me, the biggest concern is where do we find that balance in preserving the muscle memory while also maximizing efficiency? I want a car that can make nice turns at, let’s say I could make a turn at 60 miles an hour and not roll my car, then well, that sounds great. I’ll get to where I’m going faster because I have 20 turns ahead of me and I’m going to make them without having to slow down. To me, that’s a pretty decent metaphor for what can AI do. But if you’re going to teleport me from point A to point B, and I have no idea what happened and then I’m expected to supervise an audit, there’s a massive disconnect there.

If I’m just a passenger, then I have to really pay attention along the way and I need to know at what points do I need to pay the most attention or am I just going down a five mile strip and I don’t need to pay that much attention because it’s really routine and I’ve done it my whole life. How much do I need to pay attention there? So to me, where the software hopefully is going is to make better suggestions about where I need to take a moment to assess what work has been done so far and set direction and confirm the work being done is on the right track, et cetera. So we still need to be in the driver’s seat. We’re not at the point where we can just be asleep at the wheel in self-drive mode.

Joe Patrice:

Yeah. One metaphor that came up in one of the conversations I had was that there’s a lot of pressure in the AI industry to build robots and legal really to be safe needs to be more of a cyborg model. It needs to understand that at best, this is a joint endeavor between the bot and the human rather than something that can be farmed out and supervised later, like having a house cleaning droid or something. Okay, well, let’s take a break right there and hear a little quick message and be right back. All right, we’re back with Maddy Martin of Smith AI. I wanted to transition a little bit to the Small Law world because I think we kind of highlighted at the top that some of the stuff and trickling down. So from your perspective, let’s talk about smith.ai, right? So from your perspective, what are you seeing as far as where you are in firms and integrations and working together and what it takes to kind of build the Small Law workflow?

Maddy Martin:

To me, it’s about keeping people in the loop without having to be involved. And that’s what we’ve always done at Smith AI, from answering to qualification intake scheduling, it’s about end-to-end intake. So you don’t have to get involved to do any screening. The meeting is booked on your calendar and you see it in your system of record, you see all the notes you need, you show up to that call prepared, and the lead hopefully shows up to that call prepared to sign up with you. Now, what we’re seeing in addition to what we’ve always done is there’s a lot more opportunity, again on the integrations front, to have real time access to information relevant to that caller and to take workflows even further based on logic we can obtain on that call. So it could be an existing client calling in, they don’t remember their case manager, pull that data from the system and identify if there’s an existing flag on the account that a document is missing or a next step hasn’t been completed or there’s a deadline that we should take the opportunity to remind you about that.

The first part of what I said is just APIs, right? That’s nothing new. Pull the data from the system when I call on it and relay it on the phone call. Now, if it’s an AI receptionist versus a human receptionist, because we have both, then obviously there’s some AI involved in relaying that information, speech to text, text to speech, all of that with low latency. But then when we think about the discretion of is there adjacent information that is reflected in a somewhat urgent way in that matter on that record, can we bring that forward onto that call to be additionally helpful? That’s where things kind of get interesting. Can we make low risk decisions on the firm’s behalf where they even have settings in the backend to control that decision making? Then we can maximize the use of that conversation so the caller, when it occurs to them, yes, they also have a court date coming up, that they don’t need to call back when they have that recall because we already proactively addressed or anticipated that question on that initial call.

So that’s the existing client example because a lot of existing client calls don’t need to happen or don’t need to happen as often if people are looking for case status updates and things like that. Now on the new lead side, we want to take the conversation further and AI is exceptional at logic, right? I mean, that is an area of strength. So can we say, well, for this motor vehicle accident, let’s back up. There are personal injury cases that Firm ABC takes, and Firm ABC sees not only MVA cases, but they also see dog bite, slip and fall, yada yada. And then they also see some workers’ comp cases that they refer out or whatever. The MVA cases are really the sweet spot, go figure, and they want to actually send that retainer when the lead has been sufficiently qualified so that they can look into the case further.

So in that case, take the workflow even further, and then if they don’t sign that retainer, send them the follow-up, stay involved so that the firm doesn’t have to take over on that routine work that they normally would have to shepherd. That is exactly the kind of stuff that AI can do very well. While we do also have those human receptionists who are available in any moment to jump in the call when the caller says, “Look, at this point in the conversation, I’m interested in just talking to a human. Can I talk to a human?” There are still times with evolving tech where people just say, “Get me to a human.” Not out of frustration, but just because that’s the point in the call where they’ve reached their comfort limit and we can do that. And that human needs to be able to do as much of that workflow as possible, which is actually one of the constraints that we are seeing, which is all of these APIs are fantastic, but what you don’t want to do is give logins to your vendors for your systems so that humans can just go and log in and make changes to your records.

So then we have to say, “Well, what is the human interface that is a proper corollary to the AI workflows and interface?” I mean, the AI doesn’t need an interface. There’s no UI for AI. It’s just all backend, but the humans need an interface. So there is a point at which we’re at right now where we need to have harmonious continuous service for whoever is calling in whatever their preference is for AI or human, and make sure that the humans are still able to be a part of that workflow and process without Having to get that login user account, those security issues that are present when a human is actually accessing the

Joe Patrice:

Software.That’s true. A lot of the conversation we had earlier about ILTA and all that, we talked a lot about AI being used in the practice of law and the limitations there. On the business of law side, very rich and not carrying the massive malpractice implications that some of the other ones do, which is why functions like intake, receptions, intake, populating, controlling all the retainers and stuff like that, that is a real sweet spot that doesn’t carry as many of the same risks and that is really a lifeblood issue when you’re small and solo because you’re doing all that yourself as a lawyer as opposed to some of these other places where you can farm that out.

Maddy Martin:

I mean, if you even look at the mid-size firms, what they don’t want to do is continue to scale their operating costs as their revenue and market grows. I mean, your market share is growing and then you watch your operational costs scale and you’re thinking, “How can I divert these lines from tracking together?” So outsourcing has always been a great path to get economies of scale. And the other thing that firms are telling me is these people who have been intake specialists, they want to move on to more advanced roles. I mean, that was their initial vision for joining the law firm in the first place. The career path can be exceptional. So it’s also, I mean, think about any sales role, it’s tiring to be in that frontline role. Now, AI doesn’t get tired, which is beautiful. It doesn’t take time off. It doesn’t get sick days, yada, yada.

You’ve heard all those things before, but it also scales exceptionally well, especially when you have a partner where you don’t need to hire that CTO, or if you already have one at a midsize firm, you don’t need to have an entire team. That CTO can be more vendor management, working with partners like Smith AI to implement and custom prompt, create these custom prompts that allow the workflows to be tailored to exactly the business needs of that firm. So when I look at small firms, I think, how can they punch above their weight class? They can spend less on marketing. They can be more nimble with these AI and human agents that handle those intake workflows end-to-end, including, I mean, let’s not discount the risks entirely, conflict checks. That’s a great example of a riskier portion of certain practice areas that you need to get right early on.

I mean, that’s another area where we say, “Well, checking the system of record is very important at this exact step.” And that determines what next steps are taken or truncated. So it’s the combination of the APIs that we have known and loved for the last decade that everyone’s pretty comfy cozy with, with this discretionary rule-based layer of what to do with that information, given the conditions on that call or with that lead.

Joe Patrice:

Yeah. Well, cool. All right. I think we probably reached time. I kept you a little longer than usual, but this is always interesting to have a conversation. Routine listeners of the show know that I like to have these legal tech conversations and some of our co-hosts will not be as excited about it. So it’s good to have somebody who’s willing to engage via a conversation like that. So thank

Maddy Martin:

You for – Anytime.

Joe Patrice:

Right. Thanks for joining us for this episode of Thinking Like a Lawyer. I always encourage you to subscribe to the show so you get new episodes. When they come out, you should be writing reviews and giving them stars. That helps us get to a bigger audience always. And that’s the goal here. You should also check out the Jabo, Kathryn’s other podcast. I’m a guest on Legal Tech Week Journalist Roundtable. There are also a number of shows from the Legal Talk Network that you should be listening to. You should be reading Above the Law, so you read about the stories that we put out every week before we talk about them usually on this episode. You should be following us on social media, abovelaw.com. I’m at Joe Patrice. Kathryn’s at Kathryn1, the numeral one over at Blue Sky. We also are at Twitter a bit, but I’m Joseph Patrice over there.

And with all of that said, we will talk to you again next week.