TRANSKRİPT 🔗
1
00:00:01,035 --> 00:00:05,725
Welcome to Reinventing Professionals,
a podcast hosted by industry analyst
2
00:00:06,165 --> 00:00:10,685
Ari Kaplan, which shares ideas,
guidance, and perspectives from market
3
00:00:10,685 --> 00:00:14,685
leaders shaping the next generation
of legal and professional services.
4
00:00:18,674 --> 00:00:24,324
This is Ari Kaplan, and I'm speaking today
with Jim Sullivan and Tom Palladino, the
5
00:00:24,324 --> 00:00:29,604
Founder and President, respectively, at
eDiscovery AI, which leverages artificial
6
00:00:29,604 --> 00:00:34,564
intelligence for document review, privacy
response, investigations, and litigation.
7
00:00:35,224 --> 00:00:36,404
Jim, Tom, great to see you.
8
00:00:36,744 --> 00:00:37,564
Thanks for having us.
9
00:00:39,434 --> 00:00:44,134
Jim, tell us about your background
and the genesis of e-Discovery ai.
10
00:00:45,399 --> 00:00:46,619
I'm a computer nerd at heart.
11
00:00:46,789 --> 00:00:47,879
But I'm also an attorney.
12
00:00:47,949 --> 00:00:51,879
I've been in the discovery space for
about 10 years, and I really was that
13
00:00:51,879 --> 00:00:55,249
traditional predictive coding nerd,
consulting on predictive coding using
14
00:00:55,249 --> 00:00:58,484
technology to solve eDiscovery problems.
15
00:00:58,654 --> 00:01:02,444
And this came from my experience
working with e-discovery cases, using
16
00:01:02,444 --> 00:01:05,284
analytics, using predictive coding, and
seeing a lot of the shortcomings and
17
00:01:05,284 --> 00:01:09,614
problems that we struggled to solve 10
years ago, five years ago, and, really
18
00:01:09,614 --> 00:01:14,284
understanding how well AI is at solving
a lot of those problems, solving the new
19
00:01:14,284 --> 00:01:17,244
things that we couldn't even touch before
that there just wasn't a solution for.
20
00:01:17,344 --> 00:01:21,484
So, a lot of what we're doing is really
just working on building solutions to
21
00:01:21,844 --> 00:01:25,084
problems that we've had for a long time
that we can now solve that we couldn't
22
00:01:25,084 --> 00:01:28,919
before, as well as trying to move the in a
new direction and really taking advantage
23
00:01:28,979 --> 00:01:32,679
of generative AI to change things overall
and to make the entire process better.
24
00:01:33,498 --> 00:01:36,518
Tom, tell us about your background
and how your experience in the
25
00:01:36,518 --> 00:01:41,278
legal industry has helped shape the
vision for applying AI to discovery.
26
00:01:42,918 --> 00:01:47,038
I was one of the co-founders with
Andrea Wallack of NightOwl Global.
27
00:01:47,058 --> 00:01:51,758
We got acquired in 2020 by HaystackID,
and I stayed with Haystack for
28
00:01:51,758 --> 00:01:52,678
two years, which was great.
29
00:01:52,698 --> 00:01:56,228
And then I was out of the industry for a
year and I kept going to Legalweek just
30
00:01:56,228 --> 00:01:58,828
because it's, I see a lot of friends
there and it's always interesting
31
00:01:58,848
Load full transcript
--> 00:02:00,288
and I was really asking about ai.
32
00:02:00,288 --> 00:02:03,178
I was very interested, you know, I was
thinking about how it might change the
33
00:02:03,178 --> 00:02:05,898
industry and I was introduced to Jim.
34
00:02:06,478 --> 00:02:10,328
Through a mutual friend and I was
absolutely blown away by what he had
35
00:02:10,328 --> 00:02:11,448
already built, what he was doing.
36
00:02:11,948 --> 00:02:13,428
I wanted to be part of it immediately.
37
00:02:13,598 --> 00:02:16,558
In the year and a half that I've been
working with Jim, I just see more and
38
00:02:16,558 --> 00:02:18,108
more potential for this on, in the future.
39
00:02:20,388 --> 00:02:26,008
Jim, what should legal professionals
be doing today to prepare for and adapt
40
00:02:26,108 --> 00:02:28,808
to an AI centric eDiscovery landscape?
41
00:02:30,428 --> 00:02:30,808
Use it.
42
00:02:31,468 --> 00:02:32,228
Just use it.
43
00:02:32,608 --> 00:02:34,388
Use it for fun.
44
00:02:34,568 --> 00:02:35,268
Use it for work.
45
00:02:35,488 --> 00:02:36,948
Use it for whatever it
is that you're doing.
46
00:02:37,778 --> 00:02:40,118
Generative AI is incredibly powerful.
47
00:02:40,398 --> 00:02:43,768
Most people have access to something
simple like ChatGPT to help
48
00:02:44,658 --> 00:02:46,688
brainstorm ideas, answer questions.
49
00:02:47,118 --> 00:02:50,668
One of my primary ways of using it early
on was I used to work with my daughter.
50
00:02:50,808 --> 00:02:51,988
She used to like drawing pictures.
51
00:02:52,288 --> 00:02:54,948
So we would just describe
things and AI would draw it.
52
00:02:54,948 --> 00:02:56,428
And she would come up with another thing.
53
00:02:56,428 --> 00:02:59,238
An AI would draw and it would be
a cat riding a unicorn, jumping
54
00:02:59,238 --> 00:03:01,938
over a rainbow, and we'd make
coloring books and things like that.
55
00:03:02,438 --> 00:03:05,988
You just get experience from using it,
seeing how it works, how it answers
56
00:03:05,988 --> 00:03:08,628
some questions, what things it's good
at, what things it's not good at.
57
00:03:09,008 --> 00:03:13,458
And then just to take that a step further
is, use it in your job, use it, with.
58
00:03:14,248 --> 00:03:15,248
What we're trying to accomplish.
59
00:03:15,588 --> 00:03:18,328
One thing I saw too often with predictive
coding, and we're seeing a little
60
00:03:18,328 --> 00:03:21,768
bit now, but luckily not nearly as
much, is people that are pushed into
61
00:03:21,768 --> 00:03:25,888
a situation where their first use
of technology is out of necessity or
62
00:03:25,888 --> 00:03:27,538
panic, where, they don't have time.
63
00:03:27,848 --> 00:03:32,358
They have to use a technology
solution because there's
64
00:03:32,358 --> 00:03:33,318
no other options available.
65
00:03:34,843 --> 00:03:37,733
They're using it for the first time
on the biggest projects of their life.
66
00:03:38,003 --> 00:03:41,523
The number of times where we saw someone
use predictive coding for the first time
67
00:03:42,063 --> 00:03:45,763
on the massive second request where they
had no other choices, don't do that.
68
00:03:46,133 --> 00:03:47,663
Test it out on small projects.
69
00:03:47,883 --> 00:03:48,783
Use it on small matters.
70
00:03:49,768 --> 00:03:52,108
Run it side by side with
your existing process.
71
00:03:52,888 --> 00:03:55,908
Do it exactly like you're doing
now, but then do it the same
72
00:03:55,908 --> 00:03:59,508
way with AI and see how AI works
and see how much better AI is.
73
00:03:59,888 --> 00:04:03,108
So when you come across those big
projects and you need to use it, you
74
00:04:03,108 --> 00:04:05,948
already have the familiarity you need
and understanding of how it works.
75
00:04:08,568 --> 00:04:12,948
Tom, what is early case intelligence
and how does it affect the way legal
76
00:04:12,948 --> 00:04:15,508
teams prepare for complex litigation?
77
00:04:18,153 --> 00:04:20,593
I was an early tech adopter
throughout my career.
78
00:04:21,073 --> 00:04:23,553
I really liked the ECA
tools that had come out.
79
00:04:23,603 --> 00:04:26,323
They were successively better
at what they were doing.
80
00:04:26,813 --> 00:04:33,318
They really allowed people to address
data sets and understand them more quickly
81
00:04:33,618 --> 00:04:35,318
in ways that were more sophisticated.
82
00:04:36,058 --> 00:04:39,628
Jim led many, advanced text analytics
operations, for years and, and
83
00:04:39,628 --> 00:04:41,248
was a real proponent of that.
84
00:04:42,388 --> 00:04:47,238
When I started to understand what we could
do with GenAI, I immediately thought of
85
00:04:47,238 --> 00:04:51,998
that early part of the process, but I
didn't think early case assessment really
86
00:04:51,998 --> 00:04:56,208
said it well, because while those tools
can get their arms around what the data
87
00:04:56,208 --> 00:05:00,113
looks like and give you some concepts and
cluster things, they couldn't give you
88
00:05:00,113 --> 00:05:04,103
the insights that you could literally get
right at the beginning of a case now, so I
89
00:05:04,163 --> 00:05:07,623
do think that things are rapidly changing,
and I think this is one of the areas
90
00:05:07,623 --> 00:05:14,068
that will change the most is that within
the first week, the amount of insights
91
00:05:14,068 --> 00:05:18,948
that we can gain from a data set having
to do with a case, an investigation, a
92
00:05:19,148 --> 00:05:23,258
possible litigation coming up, a breach
matter is amazing and I didn't think that
93
00:05:23,338 --> 00:05:25,218
ECA was the right container for that.
94
00:05:25,268 --> 00:05:27,348
I thought early case
intelligence made more sense.
95
00:05:27,648 --> 00:05:31,038
And I really do think that this is
the beginnings of what's gonna be
96
00:05:31,078 --> 00:05:35,428
a very sticky go forward process
that, and people use our tools, but
97
00:05:35,428 --> 00:05:36,188
I think there'll be others as well.
98
00:05:36,188 --> 00:05:37,228
That'll be really good.
99
00:05:37,328 --> 00:05:40,708
But I just think the amount of
information that you're going to be
100
00:05:40,708 --> 00:05:45,048
able to get from small and large data
sets within three days, five days.
101
00:05:45,538 --> 00:05:51,178
This is compressing what it took 12
weeks to understand, in 2022 or 2023
102
00:05:51,718 --> 00:05:54,998
I don't know what that ripple effect
is to change, our industry on, maybe
103
00:05:54,998 --> 00:05:57,678
it's just a time compressor right
now and other things will come along.
104
00:05:57,748 --> 00:06:01,168
But I do think if you're that early in
a litigation and you understand so much
105
00:06:01,168 --> 00:06:04,588
about your data that fast, what does
that relationship look like now between
106
00:06:05,173 --> 00:06:08,033
corporation, law firm, provider, software.
107
00:06:08,103 --> 00:06:09,623
I think it's gonna
reshuffle that a little bit.
108
00:06:11,073 --> 00:06:16,533
Jim, how does servicing key documents,
people, events before a document review
109
00:06:17,193 --> 00:06:19,133
begins in power litigation support teams?
110
00:06:20,818 --> 00:06:22,098
It just gives you more knowledge.
111
00:06:22,098 --> 00:06:25,378
It gives you more insights into your
data, so now you can move forward with
112
00:06:25,378 --> 00:06:29,268
that information to make your process
more efficient, better, across the board.
113
00:06:29,768 --> 00:06:33,118
We're always trying to find ways
to give you information that helps
114
00:06:33,118 --> 00:06:37,488
your case as early on as possible
with as little effort as possible.
115
00:06:37,858 --> 00:06:40,558
We don't want it to be that you
have to go through a tremendous,
116
00:06:40,608 --> 00:06:44,368
number of hoops to jump through , to
get pieces of information or,
117
00:06:44,458 --> 00:06:45,978
have to do a ton of upfront work.
118
00:06:46,048 --> 00:06:49,348
People really don't like going
through long drawn out processes
119
00:06:49,368 --> 00:06:50,668
to try to train a model.
120
00:06:50,768 --> 00:06:54,318
We're just trying to get you the
information that you need as quickly as
121
00:06:54,318 --> 00:06:56,588
you can with as little effort as possible.
122
00:06:56,848 --> 00:06:59,828
So then when you face that first
case and you're on your first week.
123
00:07:00,253 --> 00:07:03,453
Whether you're starting to do a
document review in a traditional
124
00:07:03,453 --> 00:07:06,933
fashion, whether you're gonna use
technology in some way, having all
125
00:07:06,933 --> 00:07:10,573
that information at your fingertips
really makes it a smoother process.
126
00:07:11,113 --> 00:07:15,043
I think we've all been in cases, at some
time, where we had a review, a protocol,
127
00:07:15,183 --> 00:07:19,443
and we started working on a review with
a team of people, and as we learned more
128
00:07:19,443 --> 00:07:21,978
about that data, that protocol evolved.
129
00:07:22,278 --> 00:07:25,538
We learned who the different parties
are, we learned what some of the
130
00:07:25,538 --> 00:07:28,858
events are, what some things that we
didn't know upfront, and that changed
131
00:07:29,158 --> 00:07:31,418
how the document review happened.
132
00:07:31,478 --> 00:07:32,618
And it evolves over time.
133
00:07:32,618 --> 00:07:34,918
And I think most human doc
reviews evolve over time.
134
00:07:35,288 --> 00:07:38,788
So we're trying to just get
you that information before
135
00:07:39,008 --> 00:07:42,818
you start so you already know
everything that's gonna come in.
136
00:07:42,818 --> 00:07:45,418
And now obviously some things are
still gonna pop up, but we can try
137
00:07:45,418 --> 00:07:46,498
to minimize that as much as we can.
138
00:07:49,473 --> 00:07:53,053
Tom, where do you see the most
significant opportunities for legal
139
00:07:53,053 --> 00:07:57,733
teams to improve outcomes in e-discovery
and litigation by embracing AI?
140
00:07:59,351 --> 00:08:00,791
In a risk assessment business, right?
141
00:08:00,901 --> 00:08:04,941
In the end, if somebody's doing an
investigation or somebody's responding
142
00:08:04,961 --> 00:08:10,081
to, litigation, we are trying to
understand where we sit in that,
143
00:08:10,101 --> 00:08:11,501
what is the risk that they want?
144
00:08:11,501 --> 00:08:14,851
Even the spends and the proportionality,
all stems from that equation.
145
00:08:15,311 --> 00:08:19,051
And AI is going to, accelerate
this risk assessment.
146
00:08:19,231 --> 00:08:22,441
Am I investing in this case, knowing
what I know now in two weeks.
147
00:08:22,886 --> 00:08:26,566
Along with outside counsel who is
rapidly adapting to this shortened
148
00:08:26,566 --> 00:08:29,766
timeframe and knowing so much, I
think we're going to see things
149
00:08:29,876 --> 00:08:32,376
move faster to conclusions with AI.
150
00:08:32,376 --> 00:08:33,416
I think that's good for everybody.
151
00:08:33,596 --> 00:08:36,596
There's a growing mountain of
data that can only be managed
152
00:08:36,656 --> 00:08:38,056
by tools like this today.
153
00:08:38,056 --> 00:08:41,416
We can't use older methods to do
this and by the way, it's interesting
154
00:08:41,416 --> 00:08:44,966
because I'm talking about more
criminal matters than I ever have.
155
00:08:45,026 --> 00:08:48,306
My career has only been civil
litigation for 25 years.
156
00:08:48,806 --> 00:08:52,226
But criminals also buried under a mountain
of data that they can't get through.
157
00:08:52,226 --> 00:08:54,346
They don't have the budgets
to get through, and it just
158
00:08:54,346 --> 00:08:55,526
gets left behind a lot.
159
00:08:55,586 --> 00:08:58,266
And I think these tools are
going to accelerate that
160
00:08:58,266 --> 00:08:59,586
process, make it more manageable.
161
00:08:59,956 --> 00:09:03,341
I think they're playing field
levelers in a way because plaintiffs
162
00:09:03,341 --> 00:09:05,701
can use this as well as defense
can now to
get through data.
163
00:09:05,801 --> 00:09:09,921
So I do think that, this is going
to change how this plays out.
164
00:09:10,021 --> 00:09:14,451
I don't think big productions and
long, case cycles are necessarily
165
00:09:14,541 --> 00:09:16,451
going to be an advantage anymore.
166
00:09:16,481 --> 00:09:19,161
I do think there's gonna be some
shifts, but with what the effect
167
00:09:19,161 --> 00:09:23,146
that AI has on, what happens in
this whole risk assessment cycle.
168
00:09:24,836 --> 00:09:26,856
Jim, how do you see e-discovery evolving?
169
00:09:28,246 --> 00:09:32,941
I think we're gonna be able to identify
risks earlier on and, hopefully, at
170
00:09:32,941 --> 00:09:36,221
a point where we can identify risks
before they become litigation, identify
171
00:09:36,221 --> 00:09:39,461
the issues before they become civil
cases, identify the problems before
172
00:09:39,461 --> 00:09:43,301
they become bigger issues to help
those corporations save a lot of the
173
00:09:43,301 --> 00:09:44,861
money associated with those risks.
174
00:09:45,321 --> 00:09:49,511
So if we could identify a risk earlier
on there's just a lot of advantages
175
00:09:49,511 --> 00:09:51,271
to how that can save time and money.
176
00:09:51,491 --> 00:09:57,216
And shrinking the impact of the discovery
phase or the review phase, in a case
177
00:09:57,386 --> 00:10:00,736
where it's now gonna be a lot easier,
it's gonna be a lot faster, it's
178
00:10:00,736 --> 00:10:04,136
gonna be a lot cheaper to get to the
information you need to get to so you
179
00:10:04,136 --> 00:10:07,176
can focus on the other parts of the
case that matter and hopefully being
180
00:10:07,176 --> 00:10:10,796
addressing, what we wanna know is,
are we gonna win or lose that case as
181
00:10:10,956 --> 00:10:13,836
quickly as possible so we can determine
is this something that we wanna settle?
182
00:10:13,976 --> 00:10:15,196
Is this something we need to work out?
183
00:10:15,276 --> 00:10:18,486
A lot of that case strategy
comes from the knowledge and the
184
00:10:18,486 --> 00:10:19,756
information that we can gain.
185
00:10:20,166 --> 00:10:25,246
This is gonna give us knowledge and
information earlier, faster, better, and
186
00:10:26,136 --> 00:10:30,126
going to only improve to a point where
at some point we're gonna know nearly
187
00:10:30,126 --> 00:10:31,486
everything at the tip of our fingers.
188
00:10:34,091 --> 00:10:39,901
This is Ari Kaplan speaking with Jim
Sullivan and Tom Paladino, the founder and
189
00:10:39,901 --> 00:10:45,021
president respectively at eDiscovery ai,
which leverages artificial intelligence
190
00:10:45,041 --> 00:10:48,861
for document review, privacy response,
investigations, and litigation.
191
00:10:49,401 --> 00:10:50,621
Jim, Tom, thanks so very much.
192
00:10:51,911 --> 00:10:52,331
Thanks.
193
00:10:52,331 --> 00:10:52,691
Thank you.
194
00:10:55,151 --> 00:10:58,541
Thank you for listening to the
Reinventing Professionals Podcast.
195
00:10:59,271 --> 00:11:05,661
Visit reinventing professionals.com or
ari kaplan advisors.com to learn more.