Latest / The Jon and Marc Podcast / AI for Law Firms: What Should Stay Human and What Shouldn't

Transcript

AI for Law Firms: What Should Stay Human and What Shouldn't · The Jon and Marc Podcast

0:00
You see a future where we have Eve in the courtroom with the
0:05
lawyers in trial. I mean, I can't name the firm,
0:08
but this is already happening. It's just too good.
0:10
But the ability to get that information immediately is just
0:14
so useful, right? If you hear something that sound
0:16
a little bit off but your brain can't quite jar it, you can
0:20
basically just give you what to dig into.
0:23
And that's a game thing. Right.
0:35
Start a fire over here man. It's cold.
0:37
Speaking of starting a fire, we have the CEO and Co founder of
0:41
EVE here with us live at NTL. We are very excited to get this
0:45
on the books. You guys are everywhere right
0:47
now. You are blowing up an incredibly
0:50
powerful AI technology platform for legal practices.
0:54
We have been waiting anxiously to sit down and talk to you.
0:57
Jay, thank you for joining us. We're very excited for you to be
1:00
a guest on the John Mark Podcast.
1:01
Thank you so much for having me. I'm going to take a shot.
1:03
Jay Madeshwaran. We'll give you half credit, yes.
1:10
I saw you avoiding it. Yeah.
1:12
Thank you so much for joining us.
1:14
We're at National Trial Lawyers here in Miami.
1:17
You guys have giant booth downstairs.
1:20
Why don't you tell us what EVE actually does?
1:22
I want to hear your elevator pitch on the Eve platform.
1:25
Absolutely. So EVE is a leading AI platform
1:28
for plaintiff lawyers to help with casework.
1:31
So whether it is taking in a new potential new client, analyzing
1:34
whether they're a good fit for your firm and writing demands,
1:38
doing medical chronologies all the way to settlement and
1:41
discovery request response in between, Eve helps you do all of
1:45
that. And that's been part of our
1:47
secret sauces. We do it purely with AI and that
1:50
was what took us so long to do it.
1:52
We've been in business technically since 2020 and we
1:55
launched our first product end of 23 early 24.
1:59
And since then, it's been thank thanks to our customers like a
2:02
rocket ship. So how many firms are you
2:04
working with across the nation, right?
2:05
Now we're working with close to 1000 firms right now.
2:08
Wow. It's a lot, all Pi for most
2:10
part. It's about, I would say 60% Pi
2:13
and 40% labour and employment. So we actually have a good
2:17
amount of labour and employment firms.
2:18
Did you find success first with Pi firms or with labour and
2:21
employment? Yeah.
2:22
So people ask this a lot and we started caring totally with
2:24
labour and employment, which is harder for technology.
2:27
So what we wanted to get right early on was does this work at
2:31
all right like that? That is the hardest, that is the
2:34
hardest question to answer as a technology company, especially
2:38
in a risk averse place like legal, where if you mess up a
2:42
fact that is potential sanction, you know that that is you lost
2:46
that client and outcome they deserved.
2:48
And the risks are just so high that we really wanted to see if
2:53
all this tech that we built getting really, really accurate
2:56
results out of AI could be utilized to help accelerate
2:59
casework. So why we picked labor and
3:01
employment was a larger percentage of Labor and
3:04
employment cases actually moved to litigation.
3:06
So vast majority did not settle pre lit unlike personal injury
3:11
and as a result it forces us to go further into the case life
3:14
cycle and see more settlements end to end.
3:17
And until we saw a couple of those happening with very little
3:20
hallucination, if at all, that was kind of our aha moment.
3:23
And then we moved to personal injury in about a month.
3:26
And since coming to personal injury, things have been really
3:29
exploding. Does the platform layer on top
3:33
of specific case management systems or are you guys kind of
3:36
agnostic? We work with everyone that lets
3:38
us work with them, so we. We understand.
3:40
That answer so that that's maybe the politically correct way of
3:46
saying it, but honestly, one of the things that's really
3:49
important to us as we innovate, you know, we're constantly
3:51
releasing new products and features.
3:53
Last year, we were kind of the first to market with a pure AI
3:57
way of handling cases. Now we launched Eve 2 point O,
4:00
which where Eve goes from prompting and and kind of
4:02
talking to it to Eve just doing the work for you more
4:06
proactively. And we've been innovating out of
4:08
all the challenges, right. So we understand your data is
4:11
yours as a firm owner. If you think about your IP, it's
4:14
your people, process, data, money, eventually these are kind
4:19
of what you have to grow with and what you have to kind of put
4:22
a stamp on the world with. We found many clever ways of
4:27
using technologies to help you get the most out of your data,
4:29
right. But there's a CMS, whether
4:31
there's an DMS, we work with customers on a Windows version
4:34
of Prevail, that's a local database, and we found a way to
4:38
work with that. So we will innovate our way out
4:40
of any system you have. If they allow them, yes.
4:46
We're not throwing shade today. You know who you are.
4:50
When it comes to AI generally, right, not specific to the E
4:55
platform, but generally in the legal process, either in Pi or
4:59
labor and employment. What are the areas that you
5:01
believe are OK using AI for the most part?
5:05
And what are the areas where you feel like need to be AI assisted
5:08
but still need to be human driven?
5:09
And what are the areas you feel need to be 100% human?
5:11
So we're going to put things in three buckets.
5:13
Give me a few things that fall in each.
5:15
This is quite tricky to handle because there's a number of
5:18
legal tasks you do in a case is in the many hundreds right?
5:22
If not thousands. You know everyone is writing
5:24
slightly different briefs. Even demand letters don't look
5:27
the same across States and different jurisdictions.
5:30
So it's really hard for me to give you 1 answer.
5:32
My general advice is is to train yourself and your staff to work
5:37
with AI such that you are still owning the output it comes out
5:41
with. I think there's no getting
5:42
around it. No matter of how good AI gets,
5:45
eventually it's your name on the paper.
5:47
If you're submitting A complaint to A to a court, it's your name
5:50
that goes on it, not in a is. So I think we're still at the
5:54
point where vast majority of things require humans to review,
5:57
but the amount of carefulness through which review has been
5:59
decreasing, right? So back in the day, even a year
6:02
ago, chances were pretty high if you've wrote anything longer
6:06
than two pages with AI, there were going to be some mistakes.
6:09
It's not a lot. It's it's less than what, 300
6:12
words, 400 words? And nowadays we're getting to
6:14
the point where we're spitting out a medical chronology and
6:17
overview that summarizes literally 20,000 pages of
6:21
documents into a 20 page report, you know, completely accurately.
6:27
And that is that took a lot of effort.
6:30
And the challenge there is it's not easy to sum up as one thing
6:33
to make that happen. You have to break down for those
6:35
20 pages, what were the issues related to OCR errors?
6:40
You know, like how many people have these ancient medical
6:42
records that, you know, look like smudges?
6:44
And those are also AI errors, right?
6:47
Because if, if AI gets it wrong, then everything goes slack down
6:52
downstream. So we put a lot of effort into
6:54
optimizing each one of those things.
6:56
I would say today, if it's anything less than a page, AI is
7:00
pretty safe to use and you can review it very quickly and you
7:04
can submit it for anything longer than a than a couple of
7:07
pages. You still need to utilize like a
7:09
purpose built platform like Eve or others to kind of get that
7:13
output you want. But if you do, you actually can
7:15
use it for things like medical or views, chronologies, drafting
7:18
all sorts of documents. And it's it's made for that.
7:21
How concerned are are you with having client data within a
7:28
platform that has the ability to kind of act on its own based off
7:32
of inputs that are not necessarily going to be in your
7:35
control, right? You now have medical records,
7:38
you now have treatment histories, you now have Social
7:42
Security numbers, You now have all this personal information.
7:45
How concerned are you that that information is now within this
7:48
any system? Yeah, I mean, I think that's a
7:51
tricky question to answer because technically it's in a
7:54
lot of systems already. You know, it's in Dropbox,
7:56
whatever DMS you use, it's in your CMS, it's in your e-mail.
8:00
The question it comes down to is like, what can you rely on AI to
8:04
do autonomously for you and what can't you, right.
8:08
Because now all that information is in there.
8:10
You don't want it to go wild and start emailing people SSN
8:13
numbers that, you know, that shouldn't get it.
8:16
And I think those are the things you want to put limitations on.
8:19
At EVE, architecturally, we took a hard stance that we actually
8:22
don't train on your data to train any central models.
8:26
And the reason for that is the way these models still work is
8:29
they're predicting the next word.
8:32
So when you put client attorney conference show information in
8:35
the training data set, there's actually pretty decent risk that
8:38
it hallucinates something real, right, which isn't safe.
8:42
So we've taken a hard stance. We don't do that.
8:45
The second thing that we've done, which has been really
8:48
resonating with clients is it's a Eve is a closed system, right?
8:51
So it doesn't go out of the Internet, it doesn't come back
8:53
from the Internet. Whatever you upload into Eve is
8:56
what it sees and it stays there. So that puts like a pretty good
8:59
guardrail on Eve and how wild it can go.
9:03
But eventually the controls that people still have are the
9:06
eventual outputs in the form of documents, right?
9:08
That's kind of what moves the case forward step by step.
9:11
And pretty soon it'll start doing things like closing tasks
9:15
for you on your CMS or going and collecting medical records for
9:18
you. And those are kind of coming
9:20
next. Is a lot of the future of E
9:22
building more agentic functionality to do things first
9:26
just document generation? We are trying to figure out how
9:29
to utilize technology to help you grow, manage your firm, grow
9:34
and manage your firm. That's that's pretty much it,
9:36
right. So, you know, I was going to
9:38
talk earlier today and I said if I had a magic wand I would
9:40
delete the word AI from everyone's mind because I think.
9:43
You probably like also have a microphone and everyone can hear
9:46
me. Yeah, for for those that don't
9:49
know, there was a power cut and you know, I had to yell.
9:52
And we are one of the only rooms with power and for some reason
9:56
it seems like all the air conditioning is pumping directly
9:58
into this room so. Sorry, I digress.
10:00
Keep going. So like, basically, technology
10:04
is how we all should be viewing it.
10:06
You know what? What can this technology do now
10:09
that it couldn't do before? Where are the right place to use
10:12
it? Where are firms seeing the most
10:14
value in terms of outcomes, right?
10:17
And you have to drive from there.
10:19
So some of the things that we do, you know, when you sign up
10:21
with us as a client is we actually help you set your North
10:24
Star metric. You know, do you want your North
10:26
Star metric to be revenue for employee?
10:29
Do you want it to be time on desk?
10:30
Do you want it to be average time to settle a demand or
10:34
average time to get the first demand ready?
10:36
And these are all different metrics firms choose to shoot
10:39
for that indirectly are proxies towards revenue, profitability,
10:43
basically EBITDA metrics. And that is incredibly important
10:47
because eventually you have to start developing as an owner,
10:51
developing an intuitive understanding of how to take
10:54
those metrics, utilize technology, which is changing
10:57
and very empowering on a rapid basis, and marry together your
11:02
firm with it. Which means how do you get your
11:04
people to see the new technology, adopt it, change the
11:08
process and to be changed and identify what are the right
11:11
places to use it today? There's a lot of very obvious
11:14
places, right? So the low hanging fruit places
11:17
are very simple like extraction or drafting work.
11:21
It's pretty, pretty much there for that today.
11:24
A clever use case that I think a lot of people aren't doing, but
11:28
firms are doing it right now are seeing about 20 to 30% increased
11:33
value of a case when they settle is actually they asked
11:36
themselves like, what would I do if I have 100 of me?
11:39
What are some things you would be doing if you had the time?
11:43
And this is adding so much value.
11:45
An example of this is like one of our firms of worker Jeffrey
11:48
Glassman. He ended up using our auditor
11:51
functionality to go audit automatically, you know, as
11:54
though he was looking at every single case file that thousands
11:57
upon thousands of cases he had and he found undiagnosed TBI,
12:01
you know, quite a few of them, right?
12:03
And now these are things that again, if you had some genius in
12:07
your firm or firm knowledge you've incorporated over time,
12:10
you rely on just happening in case review round table meetings
12:14
or or what have you, now you can have that happening continuously
12:17
every night done for you. And TBI is just one example.
12:21
It can literally be anything. How did you, and you can use the
12:23
the Glassman example or or you can, you know, generalize this
12:26
more, but how do you actually get that tribal knowledge into
12:31
the system to be able to come to those conclusions semi
12:35
autonomously, right? Because you're saying you don't
12:37
train on the data, but yet there has to be some training that's
12:40
involved in order to come to those conclusions.
12:42
Yes and no. There's a lot of, I mean, this
12:45
gets into the heart of how Eve builds his product stack.
12:48
Ultimately, what we really focused on is also similar to
12:51
what we advise to others, such as how do you operationally set
12:54
yourself up the operation, set ourselves up to execute our
12:58
product velocity at a rapid pace.
13:01
Because today even software engineering has the same problem
13:04
lawyers are dealing with, which is now, yeah, I can write
13:07
software code and they're they're having the same
13:10
existential crisis. I think, you know, a lot of the
13:12
lawyers here have. And we're at the forefront of
13:14
using it properly to give you a better, higher quality product.
13:18
What we've learned is at the heart of it is actually
13:21
evaluations. So how do you actually know if
13:25
AI is doing a good job? And that's where we put a lot of
13:28
effort, right? So to understand things like,
13:31
you know, how do you know if your average client outcome is
13:35
going to get higher? There's been a lot of math and
13:37
different ways to kind of broken down the problem into smaller
13:40
sub component problems, which we can work on and train without
13:44
any model training. And, and the reason for that is
13:47
to these days you get a lot without training a chachi PT
13:51
like model. You can just prompt it, you can
13:54
do fine tuning, feature engineering.
13:56
There's many ways of steering the model to do what you want to
14:00
do without training. And we kind of rely on those
14:03
techniques. So you can layer additional code
14:05
and functionality on top of the AI model to achieve new tasks
14:08
without creating a whole new one.
14:09
Exactly. And this is This is why we can
14:11
move so fast, right? So we launched Voice AI for
14:13
intake and pretty soon Voice AI because of the way we built the
14:16
platform, that technology is going to be used throughout the
14:19
case life cycle. You know, now when you get a
14:21
discovery request proponent on you and you're missing some
14:24
information, that can just be a call, right?
14:27
And the call is now smart enough to actually go understand and
14:30
and pull out all the information you'd wanted to do.
14:32
And this is just, I think, some of the ways in which we're going
14:34
to see technology affecting your firms going forward.
14:38
What is the dream scenario for the use of, I'm not going to say
14:43
AI for the use of technology? Thank you.
14:46
What do you feel is like we've reached the the limit and this
14:50
is where like everything else has to be done by humans?
14:53
Where does that benchmark? I think we ultimately have to
14:56
play to our strengths. What I mean by that is no matter
15:00
what output AI comes with, if you just accept it, you left
15:04
something on the table, right? Like I, I don't think that's
15:07
ever changing. Exceptions being really, really
15:10
wrote auto, you know, dumb repeatable things like pulling
15:13
out name, date of birth, stuff like that.
15:15
Like maybe that can that can go away.
15:17
But if you're looking at a demand, for example, the best
15:22
person in the firm would slice a demand in so many different
15:26
ways. And that extra level of effort
15:28
you used to put into just pulling information out of
15:31
documents can now be put into making that demand even better
15:34
or that complaint even better. And we're seeing that already
15:38
some of our firms like a clever tactic they've been using is
15:41
because EVE has features to help you respond to discovery
15:44
requests. Pretty, I mean, it's technology
15:46
response right away. They've actually been responding
15:49
back to discovery the same day they get it just to spook out
15:52
defense. It's like we knew you were going
15:55
to ask this and we're ready, you know, so like these are some of
15:58
the clever ways of putting a lot of pressure back on defense to
16:01
their benefit, you know, like they're settling faster.
16:03
And I think those are just the beginning stages of how he uses
16:06
extra time to actually advocate for your clients better.
16:09
The depth of advocating for clients is immense.
16:13
And, and the reason why I say that is if you look at the
16:15
evolution of accounting or really any industry that has
16:19
gone gone through similar evolution in the past, you've
16:23
always utilized getting insights out of data and driving
16:27
processes, technology changes as a result.
16:31
And I think law firms are going down that same track today.
16:34
A lot of this data is in tribal knowledge, right?
16:36
So a senior attorney might know a lot like they they understand
16:40
like here's how it litigated it show the experience.
16:42
I have a trial. Here's how it bake it in further
16:45
in the in the case if I could, and this is just stuck in their
16:48
head. Some have done a better job of
16:51
taking that out. It might live in complaints have
16:53
written, you know, it might live accidentally in in different
16:56
documents, but that information is lost and it's stuck with the
16:59
attorney. But now with technology, you can
17:02
actually take that and teach a system how to think like that,
17:07
right. And now you're going from
17:09
operating on one case at a time to work it collectively on
17:13
making the system better and better.
17:15
And that has, I think, an infinite level of depth to go
17:19
because eventually it's a competitive environment and
17:21
clients are going to pick whoever advocates for them the
17:24
best. Do you see a future where we
17:26
have Eve in the courtroom with the lawyers in trial?
17:31
I mean, I can't name the firm, but this is already happening,
17:35
right? So because it's just too good,
17:37
right? I mean, when defense comes back
17:39
with 300 pages of, you know, really well prepped docs and
17:42
like a team of 10 ready and willing helping at the trial.
17:46
We have firms right now that actually do use EVE during like
17:49
an active trial. They typically have someone else
17:52
use EVE while while they're doing it.
17:54
But the ability to get that information immediately is just
17:57
so useful, right? If you hear something that sound
18:00
a little bit off but your brain can't quite jar it, you can
18:03
basically just give you what to dig into.
18:06
And that's a game changer, right?
18:08
Does it listen? Not yet.
18:10
Not yet. You have to feed it the
18:11
information. The only place where it listens
18:13
today is for people that don't have a RingCentral or some sort
18:16
of call recording software for intake.
18:18
We have a software that actually records the call for you and UMS
18:22
it into EVE or ECM Server. Have you?
18:25
One of my my favorite hacks that Mark showed me is we went to go
18:29
visit a firm, he takes out his phone and he puts it on the
18:32
table and we use a tool called the wave Wave AI, right?
18:35
We turn it, he turned it on and just sat there and it recorded
18:39
for like two or three hours our entire conversation.
18:41
And it knew the different people because at various times like we
18:44
introduced ourselves to each other.
18:46
We put wireless mics on, so when we were like in the room with
18:48
different people, that catches all of our conversation.
18:50
It's great. And then it just summarized the
18:52
whole thing and gave us an action plan.
18:54
And like, these are the things and like the timelines that set
18:56
and that type of technology that allows you to have a
18:59
conversation with another human and synthesize the aspects of
19:03
that conversation that actually need to move on to, to do acts
19:07
or have something else happen like that.
19:10
That to me is like incredible. It's infiltrating like personal
19:14
relationships and like how you engage with people.
19:18
Incredible technology. To do that in the courtroom
19:20
would be insane. I'm not saying who it was, but
19:22
one of our good friends built a tool where when they're in court
19:27
for a specific type of case, when the defense used like a
19:29
certain example, it was listening and they pull up
19:32
immediate case laws examples against it.
19:34
And I'd be like, Nope. And this one is right.
19:36
Immediately pull it up like and the defense is like how did you
19:39
it's? All it's all over.
19:42
Here, that's the future of where we're going.
19:44
You know, like we're raising the bar and I think the bar can be
19:48
continuously raised. I think that's the thing.
19:51
Is this going to get incrementally harder to raise
19:52
the bar? But that's where our time will
19:54
go. So are you going to allow your
19:56
technology to go into the dark side, into the Defence Council's
19:59
hands? We have no plans to go to
20:01
defence because I, I think right now we see a lot of benefit and
20:04
the challenge with AI is adoption eventually, right.
20:07
So it's not just the technology, but obviously very important how
20:10
accurate it is you got. To implement it, you got to
20:12
train people. That's a non trivial part of it.
20:14
So by focusing on plaintiffs and also the other thing is, I think
20:18
from a philosophical lens, like I'm more attracted to plaintiff
20:24
work, like it's, it helps people and there's something really
20:26
nice about that. And on top of that, this is all
20:29
underdog versus people with more money fight, which, you know, I
20:32
come from a pretty poor family. So I think that also is like
20:35
very motivating for me to work for.
20:37
But what I found that's really important is the fact that we
20:41
work around a case and we only get paid when we win the case is
20:45
massively important for AI because that's basically a
20:48
business forcing function getting us to utilize AI
20:53
technology, what have you, to create efficiencies where
20:55
possible at the benefit of client outcomes.
20:58
You're just using the best tools available to increase your
21:01
chances of winning. Exactly.
21:02
And that's beautiful. Defense does not have that
21:04
benefit. In fact, they have to be very
21:05
quite careful how they roll out any improvements to you know
21:10
efficiencies call it. And that's an advantage we have
21:13
right now. And we believe by doubling down
21:15
even more on making plaintiffs incredibly successful, we have a
21:20
chance to really flip the tables and that's why we're launching
21:24
intake this it doesn't have as much use in defence, right.
21:27
We do not really in taking hundreds of new clients every
21:30
day in defence. Now, once you intake something
21:33
in defence, all the different workflows all the way to
21:35
settlement can now be excited by the intelligence you've kind of
21:39
built into EVE. What is your feedback to people
21:43
who say they don't want to talk to an AI agent, even if it's
21:47
like a like a doctor's office, right?
21:48
And like the system's making the call to like to find out
21:51
information. Is someone treating like getting
21:53
information on what's happening there?
21:55
Like, do you think that that's inevitably where we're going or
22:00
is there the pushback actually going to?
22:02
I think the other trip has sailed because like, it's one of
22:05
those things where once you start using it, it's very hard
22:08
to go back. I mean, it sounds like a drug
22:10
when I say it that way, but it's it's somewhat true, right?
22:14
Like, once you realized, oh, I don't have to spend all my time
22:17
just chasing information. I get to work on things that
22:21
actually could matter, right? Once that shift happens, it's
22:24
really, really hard to go back. What I will say is, I mean, I
22:27
ask this question a lot when I come to conferences, meet
22:29
people, you know, show of hands, how many people have used AI?
22:33
It's remarkably different. John and Mark, like last year,
22:37
barely anyone has tried AI, right?
22:40
Or very scared to try AI this year.
22:43
Like literally today I asked this question, what was it?
22:46
Like a hundred 200 people and pretty much everyone or at least
22:50
using it in one function in their law firm, which is
22:53
remarkable. Like that type of technology
22:56
adoption is like never happened before.
22:58
The ship has sailed. Now what we're getting into is
23:00
questions around, hey, I bought Set AI tool, I have X&Y, but I'm
23:04
not seeing the impact, right? Like people say they're faster,
23:09
but I don't really see the impact on a firm wide basis or
23:13
in my revenues or what have you. And I think that's the next
23:17
level which we're heading into now, which is how do you
23:20
actually implement it? What are the use cases you go
23:22
after? How do you see change positive
23:25
change? These lawyers, they're starting
23:27
to use AI. There's we're at the conference.
23:30
If you look at the platinum sponsors, wherever the highest
23:32
amount, it's you guys. It's even up.
23:34
There's a ton of foundation AI. They're a little different than
23:38
you guys, bunch of different AI tools.
23:40
How should a lawyer that's here go about evaluating which
23:45
platform is the right one for them?
23:47
I tried out, you know, that's that's the basics.
23:51
Do you try before you buy a program?
23:53
So that's actually one of the things we almost started in the
23:57
AI space. Like when we, when we started,
23:59
there was no concept of try because most AI products back
24:03
then were actually people pretending to be AI, right?
24:07
Because we actually can respond in 5 minutes really get go try
24:11
it. Like nothing like seeing it work
24:13
on your own case file to tell you if it's going to work or
24:16
not. And we also don't want to sell a
24:18
product that you're not going to use or not not going to get
24:22
value from. Like that's something we've been
24:23
working for years building and that's not fun to us either.
24:27
So we actually do have a try before you buy policy.
24:29
Look at that. Yep.
24:31
Well, if you got to implement it too, I'm going to double down
24:33
and go a little deeper. Besides just you guys, how
24:36
should they evaluate which platform is the best?
24:38
There's 3 or 4 you guys leading the charge here on the space.
24:41
What's the differentiator? Yeah.
24:44
For us, I don't know the difference.
24:45
I've interviewed many of the CE OS at this point.
24:47
You guys always are very smart and very kind and the products
24:50
look great and I know most of the salespeople.
24:53
So what's the differentiator for you all?
24:55
What are things they should consider when going through
24:58
these evaluations? Yeah, I mean, funnily enough,
25:01
we're still the only provider I think on one of the large
25:04
providers that still is pure a high.
25:07
And the reason why that's important is eventually you
25:09
still have to review the documents at the end of the day
25:11
anyway. So we'd rather teach your firm
25:14
and give them tools for how do you do that quickly, effectively
25:17
and then force ourselves to make AI as good as possibly can.
25:21
So compared to everyone else, we actually spend almost all our
25:24
money on AI, which which is why like it's, it's better, right?
25:28
Second thing is velocity. So one of the things that we
25:30
took in is the eval methodology. I kind of talked to you about
25:34
earlier before we put a lot of effort into launching features
25:37
as quickly as possible because one of the challenges we see
25:40
happening is AI is changing so rapidly, no one can keep up with
25:44
it. You know, like what happened
25:46
last week on AI. You know who knows?
25:48
Personal intelligence with Google.
25:51
OK, nice, nice. Yep.
25:52
Gmail calendar photos, it integrates into Nest.
25:55
You can have it like set a daily basis, give you a summary of
25:59
like all the the things that your Nest camera caught like the
26:02
Google ecosystem. You probably don't know this,
26:04
but we've been talking about this for the past year.
26:06
Like Gemini is growing at a very fast rate and ChatGPT, it's
26:13
market share is declining. And that is because Google is
26:17
going to win the whole AI race. They have the ecosystem already.
26:20
They have the adoption. Strong opinions.
26:23
We're all in on the Gemini. I think Google has a lot of
26:26
things going for them. They have the, they have the
26:28
money, they have the talent, they have the data, the.
26:33
User adoption. They have the chips as well,
26:35
which is unique and the user adoption.
26:38
But personally, like, I hope for competition, like I think, you
26:41
know, that's good for all of us. If you have competition, it
26:43
forces, forces all of us to make better products.
26:46
And I, I think there's still a shot.
26:49
I think, you know, we, we work closely with all three teams, of
26:52
course, and they're all doing different things and I think
26:55
they're going to have the way you say specialized to.
26:58
When you say 3 you're like open AI Google what anthropic?
27:01
Yeah, see, people forget anthropic, but they're sneakily
27:04
one of the, I think better ones in the legal space.
27:07
The differentiator is between the LLMS.
27:10
It's going to become more apparent like Anthropic is
27:13
moving more towards like coders and vibe coding, like being a
27:17
platform for developers. Mark Lanier talks about how he
27:20
uses clod and then clod. Seems to be very good at the
27:23
written word. Gemini seems to be like more of
27:26
like the personal use platform, right?
27:29
It's integrating like into your house.
27:31
It's your Google Home devices. It's just all like their
27:34
subscription now is Google Premium now includes your Nest
27:39
subscription. It includes like the extra
27:41
storage. So like they're forcing you into
27:43
like their ecosystem. I don't know where Open AI fits
27:45
like Open AI was. The ChatGPT was the sort of the
27:49
first big player and now I don't know what their differentiator
27:53
is anymore. The fun part I also see is like
27:56
weird unlock you get from these different tools, for example,
27:59
with with clock, you know, you talked about cloud code.
28:02
We see actually weirdly, some of our law firms, marketing teams
28:06
use cloud code and here's what they do with it, right.
28:08
So if you ever have, let's say you want to make a landing page
28:11
for a, a mass court case of some kind or or just a landing page
28:14
in general to track some particular segment of your
28:17
marketing dollars. That page old days, maybe you
28:19
hired a contractor, maybe you set up web flow or something
28:23
similar to go do it yourself. But now you can actually just
28:26
tell cloud code, hey, make this website for me.
28:28
It does a pretty good job. Was it spit out like an HTML
28:31
file, acss file, everything. It'll write the JavaScript.
28:34
It writes the JavaScript and you can tell it go make this
28:37
animation for me. If you wanted to, you can add a
28:39
web form to it to collect. And these are ways in which the
28:43
law firm owners are using AI in unexpected ways, right?
28:48
Like you can actually use it today, not just EVE, but just
28:51
technology overall to do things in your different functions to
28:55
grow that you just couldn't do before.
28:58
You know, like who would have thought you could put an
29:00
engineer next to a marketer and that'd be cost effective, You
29:03
know, like that just doesn't make sense.
29:05
But now it does. And because that does, maybe
29:08
your marketing automation looks different.
29:10
Maybe your personalization looks different, maybe your ad
29:12
targeting looks different. For example, one of the
29:15
successful cases of AI has been Bayer.
29:17
The medical company actually like predicts outbreaks before
29:21
they happen using AI and aggressively spends marking
29:24
dollars there when it's cheap to go get mindshare, right?
29:28
So we have similar things happening with like property and
29:31
casualty cases where potentially you could see them coming
29:34
beforehand. And these are all different ways
29:36
in which I anticipate AI to be used for the entrepreneurial law
29:40
firms out there even today. We're starting to notice a very
29:44
clear split between those who use AI, which as you said, like
29:49
everybody raise your hand who uses AI, right?
29:51
Like pretty much everyone uses AI right now and those who are
29:55
power users who actually know how to harness the power of what
30:00
is truly at their fingertips, right?
30:01
There's a big difference between opening up Gemini and like
30:04
asking it a question and it giving you like a response from.
30:07
It's like grounding versus like leveraging Agenta capabilities
30:12
and setting up an MCP to like some other platform and like
30:16
doing all this stuff, coding things for you.
30:18
Building landing pages like that I think is going to happen this
30:21
year because AI usage is going to, or if it hasn't already
30:25
become so commonplace. That AI is a skill set as a
30:29
trained skill set is I think where we are going as just an
30:34
overall workforce and like in in the business world, not just
30:37
even in law firms. John and I've been talking about
30:39
this. I think we've mentioned at least
30:40
once today we have this dangerous thing we do now where
30:43
we come up with an idea. We like brainstorm how we're
30:46
going to do it using a guy's Gemini's, like asking questions.
30:48
And then I start vibe coding it out and I get stuck.
30:51
And then we have a developer that all she does is work on our
30:53
projects and I just throw over the fence.
30:55
I go figure it out. I'm like, I got 80% of the way
30:58
there. Make it work and then she just
30:59
spends the day and gets it working and something that would
31:01
have taken us like a month to build.
31:03
The next day she has it ready for me to do.
31:05
Oh man. So I'll share an interesting
31:07
stat with you guys on her own usage data that I think you'll
31:10
find interesting and aligned with this theme, which is if you
31:13
look at the engaged cohorts. So like, let's take the top 10%
31:17
of users right in the in the firm and compare them to the non
31:21
engaged ones, people that use AI for just one task or or two
31:25
tasks and they've been trained to use it.
31:27
You know, like you said, like go prompt Gemini.
31:30
Guess what the difference in usage is between the two cohorts
31:32
in usage usage. How am I answering this?
31:35
It's percentage of percentage. You can say percentage, you can
31:38
do times whatever, whatever you want.
31:40
I would say, I mean, if we're talking like one end of the
31:44
spectrum versus the other end of the spectrum, it's got to be
31:47
like 10 to 20 times. Yeah.
31:48
So I was going to guess. It nailed it.
31:50
So it's it's on average it's about 10 * 10 times more for I
31:53
was. Gonna say I was gonna say 1200%.
31:57
That's very accurate. I I was like how we answer this
32:00
question Pass fail. But isn't that isn't that crazy?
32:03
So what you said is very true. Like what we're seeing are these
32:06
patterns of for some people it clicks and they realize this is
32:11
what it's good at, this is what it's bad at, here's how to use
32:13
it to actually change something. And some that still need to be
32:16
kind of told and coached what to do.
32:18
And I think we're going through this face for our workforce and
32:21
everyone really is learning how to use these tools properly.
32:25
The really clever things we've seen firms do is they all handle
32:29
it differently to get this adoption going.
32:31
One of our firms actually, Jeffrey Glassman, since we, you
32:33
know, we'd mentioned them. Before Jeffrey, we'd love to
32:35
have you on the pod, talk about how you.
32:37
Yeah, I think it's worth it if you're available with it.
32:39
Not to pick on them too much, but he did something quite risky
32:43
with adoption, which is he gave his most risk averse most, you
32:48
know, the guys that said no to every technology.
32:51
He was the one that had access to Eve 1st and he had to prove
32:54
it. And this is the guy that you
32:56
know, crashes down even the junior attorneys demands like
32:59
he's super hard to everyone and he loved Eve and then suddenly
33:03
that made adoption in that firm like blow up, right.
33:08
And others have taken a very different approaches.
33:10
They they hire almost an evangelist.
33:11
So you have typically this would be kind of like your CIO person,
33:15
but someone champions, they understand technology and they
33:19
understand how all your different functions work.
33:21
And they're just constantly pushing for change, right?
33:24
And at some point it starts clicking in the organization's
33:27
like, oh, OK, this is what happens.
33:30
There's all sorts of clever tricks that people come up with.
33:33
One of the firms we work with, they actually set a goal company
33:36
wide for their firm. That is, I want everyone to make
33:39
100K or above this year, right? Which is an interesting goal.
33:43
And they're using that to kind of really reimagine all their
33:46
processes. AI and they're doing things
33:48
that, you know, we haven't really seen before.
33:50
All of that is just super fascinating to me.
33:52
Besides Jeffrey Glassman, which apparently you're doing a great
33:55
job. Good job, Jeff.
33:56
Which firms are using Eve the best or the most right now?
34:00
Like top three? Mike Moore's law firm, they're
34:03
very active users using it quite a bit.
34:06
And I think Dudley Du Boisier also quite a.
34:09
Very not surprising names. They're quite forward thinking
34:11
and innovative. I think Dudley also, they have a
34:13
lot of different law firms as well.
34:15
And that's a good testament and I think pressure test to the
34:18
system ourselves. So is it actually easy to use
34:21
and adopt across different cultures, different different
34:24
teams? One labor and employment firm 2,
34:27
two of them a shout out like Frontier Law, which is 1, where
34:30
he actually made his own software to work alongside Eve
34:35
to reimagine client coms. So not only is he a firm owner
34:38
that reimagine his whole practice with Eve, but he's also
34:42
making another tool to kind of reimagine client coms and work
34:46
alongside Eve to basically take the whole thing for the very
34:49
first time a client gets touched anywhere at all to settlement.
34:53
When you're looking at pricing for the firms, the traditional
34:57
SaaS model is like on a per user basis, right.
35:00
But with with AI, it's very token based.
35:04
How do you price it? You don't have to talk about
35:06
what the price actually is, but like there's this this big shift
35:09
right from like I've seen it per.
35:10
Per document or per demand or I've seen a bunch, I've seen
35:14
some wild pricing models out there.
35:15
We keep things simple and we price per case.
35:17
So we we actually for the most part don't change it drastically
35:20
for all these different criterias and we just take the
35:23
hit and on it. And the main reason is we
35:26
believe it should be indirectly related to value, right?
35:29
Eventually we want to be driving value to you, right?
35:32
Either it's time saved or more likely like increase settlement
35:36
rates and settlement dollars, we want to drive that value to you.
35:40
And eventually we want to get really good understanding what
35:43
is the type of work that is required to drive that value.
35:47
And eventually that amount is going to be a well known amount.
35:50
So this is some of the ways in which why we've been AI focused
35:53
and why we've raised so much money has helped us is.
35:56
You guys have raised a lot of money.
35:57
We have raised a lot of money. Congratulations.
36:00
Thank you. That's, that's helped us because
36:03
we can actually pass that along to you guys.
36:05
And together we're, I mean, honestly, you're doing a lot as
36:09
well. You're putting a sweat equity in
36:11
reimagining how to run your law firm with EVE and we want to
36:15
make that as easy as possible to you.
36:16
And we found token usage to be incredibly unpredictable.
36:20
Like for example, if you're dealing with a 10,000 page
36:23
Medcron that can cost 3 or $50.00 in tokens once right to
36:28
run it once. You don't want to be dealing
36:30
with that like you could, you could easily just run into a
36:32
situation where you're spending millions on tokens.
36:36
And we want to be the kind of that safeguard.
36:38
For you are the law firms passing the.
36:40
You were asking my question, Is it an expense that they add to
36:44
the end of the case? So I would say more than half
36:45
are doing so look. At this, we're on the same
36:47
wavelength, yeah, and. Many are actually doing that
36:50
along with disclosures. So they disclose, they use AI,
36:53
and here's how I use AI, here's what I'm using it for.
36:57
It gets you all communicate the value to them, right?
36:59
It gets them higher settlement rate, faster response times,
37:02
faster time to settle, better and better outcomes in general.
37:06
And people don't say no. This might change state to
37:10
state. There might be some statewide
37:12
changes, but in California, people are largely up for it.
37:16
And in many other states as well, especially the younger
37:18
generation, they would rather talk to a, to an AI weirdly,
37:22
than a person. You know, that's, that's already
37:26
happening. It's, it's a bit scary in some
37:28
cases. It makes sense because we deal
37:30
with labor and employment as well.
37:31
You know, sometimes you run into cases where they're dealing with
37:34
very sensitive things like sexual harassment or, or very
37:37
personal things. They don't, they might not even
37:38
feel comfortable talking to a person yet about maybe they
37:41
never talk to anyone about it. They feel safer talking to AI
37:44
because oddly AI is better active listener and then most
37:50
most people out there. I'm scared for the next
37:52
generation, like all these skills that you develop over
37:55
time, just interacting with people and like active
37:58
listening, right? They're they're slowly going to
38:01
go. Away I was watching a video and
38:02
they're talking about the biggest issue that companies are
38:05
having interviewing the newest generation of kids coming out of
38:08
college and they like ranked them.
38:10
I forgot it was the two biggest items during interviews is
38:13
they're unable to maintain eye contact during conversation on a
38:17
remote interview and the second one was they cover their camera
38:19
up during a zoom interview. Like they don't.
38:22
Why? Like you're being interviewed
38:24
for a job, why would they cover their camera?
38:25
I I agree, that was my thought as well.
38:27
Skills you do not learn, they just are unable to have.
38:30
They struggle with like intense face to face conversations that
38:34
require icon. Probably because like, social
38:35
media has become so polished where like, you can record a
38:38
reel or story if you don't like it.
38:39
Like, all right, I'm just going to do it again.
38:41
Like you can't do that and. Quick question, are you a robot?
38:46
Not yet, not yet, not yet. Out of left field.
38:52
Are you a robot, Jay? This has been fantastic.
38:56
Like loved having you on the podcast.
38:58
I think we learned some stuff. Learned a ton.
39:00
And it's great to always talk to technical founders who
39:04
understand both the legal side and the technical side.
39:07
For the law firms who are watching this.
39:09
Don't know what Eve is? How can people find out more
39:12
about what your platform does? Eve dot legal go to it and like
39:16
I mentioned, you can demo it on your actual cases before before
39:21
you check it out. The second thing is our team has
39:23
gotten pretty good at understanding how to take you
39:26
through the AI transformation journey as well.
39:28
So the more you tell them, typically they can give you
39:30
pointers on what are the first places you should start.
39:33
Things we've been learning over time across thousand law firms
39:35
we work with use us to at least at minimum, learn about what the
39:39
initial places to do it do it should be.
39:41
Jazz, a fantastic conversation. We know you're a busy man.
39:44
We appreciate you taking the time of your schedule to meet
39:46
with us and I would love to sit down with you in like a year to
39:50
see where the product's been and get an update.
39:52
Absolutely would love that. Thank you so much for having me
39:54
on again guys. Yeah, absolutely.
39:56
Thank you very much.