Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
This podcast is for informational purposes only
0:02
and does not constitute an advertisement. Views
0:05
expressed are those of the individuals and not necessarily
0:07
the views of Toma Bravo or its affiliates. Toma
0:10
Bravo funds generally hold interest in the companies discussed.
0:12
This podcast should not be construed as an offer
0:14
to solicit the purchase of any interest in any
0:16
Toma Bravo fund. Hello
0:22
and welcome to a very special live
0:24
recording of Toma Bravo's podcast, Behind the
0:26
Deal. I'm Toma Bravo managing
0:28
partner Holden Spate. We're
0:31
coming off an amazing second season of Behind
0:33
the Deal. Well, if the auto companies want
0:35
you to be bigger than you are, you've
0:38
done them a service because you've got $300
0:40
million in revenue when we bought J.D. Power.
0:42
Today we're approaching $900 million. How'd
0:45
you pull that off? Well, you guys were helpful. Thank
0:47
you for that advice and capital. We
0:49
initially created this podcast to give an
0:52
inside look into the transactions and partnerships
0:54
that shape our portfolio. Over the last
0:56
year, our partners sat down with our
0:58
portfolio company CEOs to explore the stories
1:00
and lessons behind some of our most
1:02
dynamic software deals. Orlando never flinched. He
1:05
got into his car, drove
1:07
south down the 101 and pipped
1:09
the deal from this other private
1:11
equity firm. The rest
1:13
was history after that. I remember that
1:16
very well and it was an
1:18
exhilarating evening of deal making and
1:20
it was also the first step in our partnership.
1:24
But perhaps the best deal we made was with our
1:26
listeners. Behind the Deal has
1:28
received over half a million downloads and topped the
1:30
US business charts on Apple Podcasts. Firms
1:33
change, firms evolve, partners leave, there's
1:35
spin outs, people start new firms.
1:38
And at Toma Bravo, we've been at this together now
1:40
for over 25 years. Nothing
1:43
has changed. We're still here doing
1:45
the same thing. And frankly, as
1:47
we say it all the time, I feel
1:49
we're only getting started. Thank
1:52
you for listening. We appreciate your support. And
1:54
now it is my pleasure to introduce Toma
1:56
Bravo managing partner, my
1:58
friend and more important. Certainly behind the
2:00
deal co-host Seth Bora. This
2:03
is Behind the Deal Live from
2:05
Miami. Well,
2:10
that was awesome. I'm
2:12
a little nervous here.
2:14
I've not done a live
2:16
video podcast before in
2:19
this format, but I think we're going to be okay.
2:21
We have an awesome, awesome topic today
2:23
that I know a lot of people have
2:25
a lot of interest in because we get
2:28
this question constantly about AI and
2:30
what's going on in our portfolio. Here
2:33
today joining me from three of
2:35
our flagship companies in fund 14 and fund 15
2:37
are Sumit Dhawan
2:40
from Proofpoint, Mike Capone
2:42
from Click, and Charlie
2:44
Gottdiner from Aniplan. It's a
2:47
nice representation of applications
2:49
with Aniplan, security in
2:51
the email cyberworld with
2:53
Proofpoint, and then data
2:55
and analytics with Click. Maybe
2:58
just to level set, AI
3:00
now is in
3:02
the vernacular constantly. It's
3:06
something that a couple of years ago, of course,
3:09
we knew about, but we weren't talking about it as much. It
3:11
would be great just to get each of your
3:13
perspectives in terms of how
3:16
we got here. Why is it today
3:18
that we're talking about
3:20
it? It's hard to read a
3:23
business article or a company article
3:25
without talking or thinking about AI.
3:27
What's changed, Sumit? Maybe we'll start with you.
3:31
Good afternoon, everyone. I sort
3:33
of say Proofpoint has always
3:35
had AI. We did cybersecurity.
3:37
You can't have people defending all these
3:40
other attackers. You have to use technology,
3:42
and the technology needs to keep learning.
3:45
By definition, that's classical machine learning.
3:48
I jokingly say Proofpoint did AI
3:50
before AI was cool, and I
3:52
call it the before cool era
3:54
or BC era of AI. We
3:59
Have done all forms of AI. Machine Learning and
4:01
a i might have to
4:03
essentially prevents attackers that are
4:05
constantly attacking through email and
4:07
other forms of communication. That.
4:10
People have. And a eyes
4:12
the technology of the models as the way
4:14
you beat all of you. You do pattern
4:16
detection and then you'd you detect. This is
4:18
a potential attacks and you don't let it
4:20
go. To. The to the employee
4:23
your the and uses as we
4:25
call it in the tech job
4:27
and and now what's happening is
4:29
with the generative A I. Basically
4:31
it's consumer rights where everyone can
4:33
experience the power of a I
4:36
because you can actually talk to
4:38
it is can generate responses, images
4:40
all forms of move models that
4:42
are available in terms of how
4:44
you experience Ai so it's readily
4:46
available. And. It's readily available
4:49
for both. Good. Guys like
4:51
us to use as well as in
4:53
the world of cyber. The bad guys
4:55
they are. You know, a I can
4:57
be used to potentially create new forms
5:00
of attacks and a I can be
5:02
used Now. For. Preventing those new
5:04
forms of attacks. So but in
5:06
the world of cyber as we
5:09
are certainly seeing more and more.
5:11
How. This sensitive a I both in
5:13
terms of. Threats. As well
5:16
as how you prevent that sets to
5:18
become more and more relevant and important
5:20
going forward. Mike What
5:23
some. Yeah. You've obviously in
5:25
class been delivering insights your
5:27
customers in various ways overtime.
5:30
How has this? Shift.
5:33
To Generative enabled you to deliver more about
5:35
now I'd actually be curious. Also, Charlie Tatty
5:38
hear from you when make his gun in
5:40
terms of what kind of business value
5:42
you're able to generate today with this with
5:44
this major platforms chef that damn that's
5:46
taking place really over the over the last
5:49
year, right? right? Well suited, right? First of
5:51
all, this is not new. I mean, we're
5:53
we're writing. I'm a I programming when I
5:55
was in college. You know, some twenty thirty
5:58
years ago. This kind of breakthrough. In
6:00
to consumer with Judge should be T and
6:02
L A Lamb's I'm has been able by
6:04
the masses computer power this out there to
6:06
it's available to us and cloud computing and
6:09
just more sophistication in Canada techniques. Are we
6:11
a cliff? We've been getting raped his moments
6:13
the last five years. We. Were
6:15
a analytics company and yeah we had
6:17
a I built into our platform. We've
6:19
been to an L P, Naturalize Processing
6:21
and our platform you will interact with.
6:23
Data. Using modern Ai techniques for long
6:25
time. Or what changes with his
6:28
massive compute his ability or for companies
6:30
out a harness all the information as
6:32
a finger tips in this kind of
6:34
explosion of ai in the scanner. Renaissance.
6:37
Had happened last year was a huge wake
6:40
up call to companies. You suddenly realize I
6:42
have to build a foundation. take advantage of
6:44
a i can just plug and and el
6:46
amor chatty bee Jay's at the and yeah
6:49
to do the work is what we see
6:51
this country's release. I'm start to scramble to
6:53
build and infrastructure in a foundation that secures
6:56
and governs their data. For. Ai
6:58
and make sure that you know run into
7:00
kind of problems I you're seeing out there
7:02
today our are false positives, hallucinations and things
7:04
and so the series of acquisitions I we
7:06
did including leading up their talents and was
7:09
really to get ready for this moment swimming
7:11
and races mom for last five years below
7:13
harness data for enterprises actually effectively use a
7:15
I as part of a corporate structure vs
7:17
on consumer. Riding High School term
7:20
papers for example is not Weber for China Do.
7:23
Yeah so I'm before. Talk about the
7:25
value points. Iceman One thing agree with
7:27
both of us. Everything that buses and
7:29
said. I think the other thing that's
7:32
important to recognize is that Ai and
7:34
Ml was really domain of. Data.
7:36
Science experts right and that expertise really
7:38
isn't needed as much as it still
7:40
needed Said to make advances when not
7:42
meet needed to engage in a I
7:44
like the ones was because it hasn't
7:46
and summarized by the alarms and and
7:48
the of the open A I architecture
7:51
so I think that was also a
7:53
breakthrough right where he just don't need
7:55
the decides expertise to engage with a
7:57
ice that we did. Two. Three
7:59
years ago. I'm. No
8:01
better and plan So we think about
8:04
creating value for customers with they are
8:06
in three different ways. The first is
8:08
driving more insight for them. So.
8:11
In a Plan in what customers do
8:13
on and a plan is they forecast
8:15
and plan their businesses and what a
8:18
eyes. And we've had a I products
8:20
now for five years. When Ai allows
8:22
them to do is to make more
8:24
accurate forecasts because we're we're running Mls
8:27
algorithms cause is mostly machine languish models
8:29
across much larger datasets, much deeper datasets,
8:31
much larger datasets both first party and
8:34
third party denim and that improves forecasting.
8:36
so that's kind of the first doctor
8:38
a value. The second is access. So.
8:41
This is where we get into Jenner
8:43
and right where where Release allowing customers
8:45
and all different levels of the organization
8:47
to access their models in a much
8:50
simpler way. So today primarily people access
8:52
it during a plan. Models or model
8:54
builders are experts that have to go
8:56
into the platform and dugout the inside
8:58
stuff. Hard to dig them out, but
9:01
you have to know something about the
9:03
platform. Tomorrow. We're going to
9:05
leverage network or really web prototype now.
9:07
Where. People can build to ask
9:10
their models questions using natural language, so
9:12
a Cfl could ask for a forecast
9:14
right on his way to work and
9:16
get that answer back when I haven't
9:18
got to go through an interim intermediary
9:20
that's expert in the platform. So.
9:22
That's that's today in the nuts. Actually getting
9:24
tied to work flocks so they can actually
9:26
asked. Sending the question to somebody has a
9:28
follow. So. That's really second
9:30
point of value. The third is really
9:33
efficiency. So. We think about developing a
9:35
copilot for modeling so said he asked again
9:37
you have to be an expert to build
9:39
a model and and a plan that takes
9:42
a lot of times. You start from scratch
9:44
and what the copilot will do is will
9:46
be a little old, build a model for
9:48
you and then model blurs become others as
9:51
opposed to craters that that's much more productive
9:53
and obsession. Sumit
9:55
In in cyber there is some
9:58
so many threats sectors. And
10:00
you know clearly. The
10:02
advent of generative ai has created yet another
10:04
one uses how are your customers managing those
10:06
threats today? Maybe walk through an example of
10:08
what you've seen in the feel just in
10:10
terms of some of the more modern day
10:12
threats that are coming out of that. The
10:14
use of as you said that you know
10:17
that they're bad guys has engendered of a
10:19
I of course it helps you know that
10:21
group as well and it's very well funded
10:23
and in what are you doing? On the
10:25
flipside. To. Help combat some of
10:27
these new threats and how quickly
10:29
have you had to react to
10:31
that? Yeah, Firstly, I think every
10:33
one of the was probably experienced
10:35
an email. What? a text message
10:37
that comes in either trying to
10:39
make you click on something. Or.
10:42
Sometimes. Even. Some. Something
10:45
coming from your Cfr Ceo for
10:47
in on discard that you want
10:49
a by. It's pretty common attacks
10:51
both for. Fraud. As
10:53
well as far as basically in
10:55
of planting the little malware into
10:58
your network or your computer which
11:00
can then eventually become ransom with
11:02
attacks and it i'm many of
11:05
your phones have either been through
11:07
it or in always prevent some
11:09
doing it. turns out more than
11:12
three quarters of those attacks stock
11:14
from email. And. They continue
11:16
to be that way. So
11:18
if you think about cyber
11:20
cyber already started from network
11:22
you know he added events
11:25
my computer network and makes
11:27
your bad guys. Can't. Get
11:29
In! But social engineering which is
11:31
people attacking people is the biggest
11:33
surface area for most attacks coming
11:36
in and that's what we focus
11:38
on. We have three zillion. Emails
11:41
that me that go through our
11:43
system. About one and a half
11:45
trillion Sms messages that go through
11:47
our system which give us a
11:49
mechanism to build. Models.
11:52
Models that can predict. Which
11:54
is a good email And what's a bad
11:56
email? So. It. At. Any given point
11:58
of time that able to see the
12:01
as a company. probably the only company
12:03
in the world who can really see
12:05
ahead of the curve on what is
12:08
the emergence of new types of attacks
12:10
which is what sets you would asking.
12:12
So using all this data we have
12:15
it set research and said intelligence team
12:17
that continually continuously publish published to our
12:19
customers and the world how the attacks
12:22
and evolving. what we're seeing now is
12:24
that generate of Ai at this point
12:26
of time. Is. Being used
12:28
more and more by attack as
12:30
which we can tell we and
12:32
tell when and the tax which
12:34
is coming in through an email
12:36
or a spurious looking website that
12:39
users that being induced to click
12:41
on because they looked familiar that
12:43
all been created. using. Sensitive
12:45
Ai. So in other words the
12:47
same attacks but are being created
12:49
by gender to me I so
12:51
what does that mean? That means
12:54
often times these attacks are created
12:56
with non native language speakers. That's
12:58
no longer an issue so we
13:00
can just sort of make these
13:02
emails and website through language assessment.
13:05
If. They are as threats or not,
13:07
that's no longer the case. We have
13:09
to throw that out and we have
13:11
to use other indicators to tell if
13:14
it's upset or not. Secondly, there is
13:16
more contacts You can bet is often
13:18
times now generative ai based attacks are
13:20
not going to be read this and
13:22
one. Email. To
13:24
a bunch of people as a
13:27
campaign. Instead. Because they're
13:29
being robotically generated, Generative A
13:31
I enables effectively a robot
13:33
to have a conversation with
13:35
several individuals in this room
13:38
which is more conceptual to
13:40
you as individuals, which makes
13:42
the models that we have
13:44
to have to protect against
13:46
those kinds of threats. Even.
13:49
More sophisticated it needs to have
13:51
more information on if I'm as
13:53
a send us sending you those
13:55
kinds of emails and trying to
13:57
induce you to click on some
13:59
things. Then using our technology now
14:01
which is also based on these
14:04
laws language models we can detect
14:06
you know what summit never really
14:08
sends. An email to
14:10
Orlando asking for credit card numbers
14:13
or. Just.
14:16
A difference but they have
14:18
such but so so that
14:20
would basically created them sexual.
14:23
Model. To. Say this.
14:25
Senders. Is not than
14:28
this is not the right context for
14:30
this type of female and that's the
14:32
kind of sophistication both said actors are
14:34
using and v at evolving an our
14:36
models to make sure we can protect
14:39
and that's a good thing in my
14:41
assessment for incumbents. Such. As us
14:43
because I started this by saying we
14:45
have three trillion emails a year that
14:48
we process one point three trillion you
14:50
know Sms seventy million euros That me
14:52
see what does. That mean that means we
14:54
have the data. Leak Insane. These models
14:56
better than anyone else. It's hard
14:58
for an incumbent like us to
15:00
be disrupted by someone because building
15:02
the tests and the code for
15:04
these martin is not the hard
15:07
thing. Training that's with the right
15:09
morning is a hard thing and
15:11
that's what incumbents. Benefit
15:13
from. Clique
15:15
Clearly the importance of data.
15:18
As. Is something and we've talked about for a while.
15:20
but if you it feels like we're now in a new realm.
15:22
Of. The value and only mad
15:25
and make I know you're neighboring customers
15:27
in this journey Can you help us
15:29
understand what your customers you're asking for.
15:31
Today's they get ready. To.
15:34
Use these large language models
15:36
term. Cradle. You for their
15:38
customers or internally and what does that look like?
15:40
It clicked. a customer is ready today. You do
15:42
their budget to the of the towel and have
15:45
a what What does that? You know what is
15:47
a process look like read now so are So
15:49
you're a year ago were all sorry mound this
15:51
explosion. And. There is almost a panic
15:53
out in the market rate as and boards
15:55
are the only a C O saying do
15:57
something, do anything and people are announcing an
15:59
issue the sending money. but really it was
16:01
just to show that they were paying attention
16:03
and and not paying for a been disrupted.
16:06
The good news is a lot of that
16:08
is settled down right now. The. Hype is
16:10
sort of settled in our into this phase of
16:12
people being a lot more thoughtful which is great
16:14
and what they're saying is and it's build a
16:16
foundation A double the right foundation. For.
16:18
My future of Ai and really comes down to three. Thanks
16:20
for says how can I get. The. Maximum value.
16:22
The second is how can I government and
16:24
be secure site on land in. In.
16:26
Jail he saw the each past really
16:29
sweeping legislation. Recently. Arm
16:31
around. What can happen if you miss use
16:33
a ice and then thirty cost right? there's
16:35
There's caustic up there is so say we're
16:37
building up his infrastructures. I'm on a date
16:39
aside. What? What we've been building where we
16:41
do is really simple which is. Data.
16:44
From anywhere any source cloud on
16:46
promise or wherever at high velocity
16:48
in real time and scam arm
16:50
cloud data like structures but also
16:52
governed catalogs with veracity proven data
16:54
lineage. I'm on on placing garden
16:56
aware not only way that is
16:58
where it came from and then
17:01
again allies. Easy modern kind of
17:03
techniques like analytics, an ai and
17:05
that's long as processing and then
17:07
something try some for is really
17:09
important. Can. Accidents and once you
17:11
get the inside is see Ozil Tommy I
17:13
had a got the answer the analytics Me:
17:15
I gave me the inside the now
17:17
my com. he didn't do anything with their
17:20
rights to the village actually take that
17:22
and put it into a workflow, put it
17:24
into an automation, send an alert. I'm integrate
17:26
to an Rp a platform supercritical otherwise
17:28
you'll get a value on that's what happened
17:31
so that that's Canada. The holistic view
17:33
of things and nematode say is on. On
17:35
the on the cassada things people are starting
17:37
to get the first anniversary of their
17:39
their cloud. Data like belay the starting understand
17:41
what's happening and how much money they're spending.
17:44
The mood date around. And they're freaking
17:46
out either saying I can't This isn't sustainable as
17:48
someone harness all my data. so we did
17:50
was. We built a series of. Analysts
17:52
have season a Powwow! It's platform a
17:54
top of our data platform so customers
17:57
manage putting data in the most cost
17:59
effective place. Don't drive the right outcome
18:01
but make sure that you know when you're
18:03
paying for computer you're paying for state I'm
18:05
I'm restores as you putting these in the
18:07
most cost optimal place and still getting me
18:09
I'll com as really how people are thinking
18:11
about in the going is is is moving
18:13
and a more thoughtful pace. Now scientists do
18:15
anything to say that you doing something am
18:17
there are budgets people are spending I'm a
18:19
lot of a soft down. C O sang
18:22
an earmark lot of money to get on
18:24
a I but now that money's being deployed
18:26
very very thoughtfully. Charlie
18:28
Obviously you need internal towel to
18:30
work with all these new. Technologies.
18:33
An Attorney You consider this a
18:35
platform shifter enabling technology Bad. Maybe
18:37
help us understand. You know
18:40
what that is look like as you're
18:42
running the company him in terms of
18:44
what sort of internal resources you need
18:46
actually turn this into a product that
18:48
revenue generating for you and has up
18:50
in a challenge given how quickly moved
18:52
are or how how the matter said
18:54
so far. Yeah so I'll go back
18:56
to a someday I talked about earlier
18:58
so you know this all started west.
19:00
Really a data science. Sort. Of
19:02
academic experiment right? That's where machine
19:04
learning really started. It became more
19:07
commercial impractical over time as we
19:09
the hardest more compute. larger.
19:11
Datasets And so the system
19:13
we've seen Intel on is.
19:16
Assessing whether our data scientists can make
19:18
that chef to be more commercial. To.
19:21
Drive commercial outcomes,
19:23
From. Their. Of machine learning
19:25
models and really drive a I
19:27
that's going to drive the business.
19:29
So I talked earlier about. You.
19:31
Know us are leveraging generative
19:34
in. In. I'd tell the of
19:36
nib the prior to ask Anna plan
19:38
South Pacific could actually query miles through
19:40
natural language. That's. A
19:42
harder shift for. Academic
19:45
data scientists. They don't think about the
19:47
world that way because they want to
19:49
run the next experiment. They. want to
19:51
build the next in a large language
19:53
model that may or may not have
19:56
a practical applications so that's really our
19:58
challenge right as turnout figuring out who
20:00
can really make the leap to commercial
20:02
because we're not an academic institution, right?
20:04
We want to leverage the knowledge of
20:07
academic institutions and we want to commercialize
20:09
it. So that's the journey that we're
20:11
on. We've got a fairly big team
20:13
that's trying to make that journey that's in Israel
20:15
that came through in acquisition for us. Mike,
20:18
you mentioned the EU legislation
20:20
that came out last week. Maybe
20:23
walk people through because you're very
20:25
deep in that. You've formed an AI
20:27
council around governance and ethics. I think
20:29
it'd be great just to get your
20:31
perspective and then Sumit and Charlie, your
20:34
perspective on just from a governance and
20:36
ethics perspective, how this is
20:38
being managed internally at your companies because it is
20:40
a big issue. It's
20:42
a huge issue. Just like every
20:45
other technology, the innovation is
20:47
going to outpace society's ability to kind of
20:49
regulate it. It's just how the
20:51
world works in today's world. We
20:54
in the corporate world actually have an
20:57
even greater obligation to govern this and
21:00
to make sure that what we're deploying, the tools we're
21:02
deploying and the technology is used in
21:04
the right way and can be governed in a way that you can
21:06
control the ethics of it. So
21:08
we're really thoughtful about that at Clique. For
21:10
the most part, I'd say most companies in
21:13
our industry are behaving that way. Some
21:15
aren't, but most are. My
21:19
favorite kind of paradigm is if you think about
21:21
social media 20 years ago and how that started
21:25
taking off and what that's done to teenagers
21:27
in today's world, would we have approached that
21:30
the same way? Would we have let that go
21:32
the way it did, completely ungoverned for so long?
21:34
Would we have been more thoughtful about it? I
21:36
think that's where we are with AI today. We
21:39
at Clique spend a lot of time
21:42
thinking about our responsibility to the ethics and
21:44
the morality of AI. What
21:46
we know is that we don't know. We
21:48
haven't figured everything out. We formed a council of
21:51
people mostly outside of industry, people
21:53
from public sector, people from academia. These
21:55
are our four very respected individuals, come from
21:58
Cambridge and things like that, who actually... advise
22:01
us and our customers on the
22:03
ethical deployment of AI and how we can do
22:05
this in the most, you know,
22:08
capture all the value, certainly we're in the
22:10
value creation business, but also do it in
22:12
a thoughtful way where people view us as
22:14
a trusted partner that helps them be successful
22:17
the right way rather than the wrong way.
22:19
The EU legislation is a great example where
22:21
if you've read it they've gone quite deep
22:23
and they've categorized different uses of AI, some
22:26
of it being completely prohibited. So using
22:28
AI to predict criminal behavior, even
22:30
to predict ethnicity or gender,
22:33
some things you might not be able to do anymore in Europe
22:36
with your platform if this really takes
22:38
hold and they've categorized it from ways
22:40
and it puts a huge burden on companies to make
22:42
sure they're using AI the right way and govern their
22:44
data the right way and like, you know, the EU,
22:46
they never do anything like light, right, the EU Parliament,
22:48
so the penalty is up to 7%
22:51
of turnover, 7% of revenue if you
22:53
screwed up, right. So you really need to be
22:55
very thoughtful, very reminiscent of GDPR and so, you
22:57
know, we have to help customers, all of us
22:59
have to help customers make sure that they're doing
23:01
this not only fast and capturing a
23:04
lot of value but the right way morally,
23:06
ethically and in compliance with the law. Yeah,
23:09
I think given with proof
23:11
point I mentioned we have
23:13
data from customer data, so
23:16
one of the big things in addition
23:18
to obviously our models being trained
23:20
the right way and are doing the right things
23:23
that are compliant, we also have to be extremely
23:25
careful with data residency laws
23:28
that are there, so data residency
23:30
laws are different across the globe
23:32
and they become even more important
23:34
and critical when you're applying these
23:36
models on them. You know,
23:39
it becomes quite complex like for example,
23:42
I think everyone here has heard
23:44
of or maybe even familiar with
23:46
or experienced co-pilots from Microsoft. One
23:49
of the banks that maybe we have
23:51
folks here as well, I
23:53
heard from them that they were
23:55
doing a pilot and co-pilot was
23:58
linked into certain business applications. as
24:00
well as, and I don't know if
24:02
it was Copilot or some other generative
24:05
AI chatbot that was prepared, that was
24:07
connecting to different business applications and they
24:09
were running a pilot and they enabled
24:11
a set of users to start doing
24:13
queries. And once you
24:15
enable business users to have
24:17
that form of free form
24:19
access to information that's in
24:22
multiple business applications,
24:25
the type of queries that people
24:27
start making can expose
24:29
the information that sometimes regulate
24:31
it and protect it in
24:33
certain ways which you
24:36
never thought anyone
24:38
would sort of be able to access. So
24:40
all of a sudden it is opening up
24:43
a whole new dimension of
24:45
data security protection, governance, requirements
24:48
for customers. And so I do
24:50
see in terms of adoption of
24:52
the technology, even from the customer
24:54
end, where they are all going
24:56
through this sort of phase of
24:59
how they are gonna tackle cost and governance
25:01
elements. And for our end, we have to
25:03
be much more thoughtful in terms of what
25:05
we do with our own data governance for
25:07
all kinds of regulations
25:09
that exist from sovereignty
25:11
perspective, residency perspective, or in
25:13
general compliance because of the new rules that
25:16
are coming in. Yeah,
25:18
I agree with that. I think many of
25:20
you may be asking, so what now? What
25:23
do companies do? I think the
25:25
news is actually a little better than it was five,
25:28
10 years ago from a data
25:31
privacy perspective because there
25:33
are regulations and there are models
25:35
that companies have followed for data
25:38
privacy and data governance. And so
25:40
there are techniques like privacy by
25:42
design that really govern the
25:45
use of data in product innovation in
25:47
most companies. Not every company is great
25:49
at it, but at least these kind
25:51
of concepts exist. And I
25:53
suspect AI will leverage those to begin
25:55
with, but we'll need, as Mike, you
25:57
point out, a lot more governance because...
26:00
it can be a free-for-all and
26:02
privacy by design is actually a pretty
26:04
rigid process that forces data governance rules
26:07
but if AI is in everybody's hands
26:09
then it would be hard to really
26:11
manage that and police it. So
26:13
but I do think the news is a little
26:15
better there's a starting point and most companies have
26:18
data privacy and data governance committees
26:21
set up that as a starting point
26:23
and that's where I where I see
26:25
a lot of companies starting is
26:28
in those committees. You
26:30
hear talk about the fact that generative
26:32
AI could be an issue as it
26:35
relates to IP and
26:37
you know the the competitive
26:40
landscape within certain
26:42
industries especially within software where maybe
26:45
there's an ability to spin up a new company
26:47
quickly. Can you share your thoughts
26:49
on you know on that how big
26:51
a risk is it to your businesses
26:53
what should we all be thinking about as
26:56
investors in software as it relates to that
26:58
maybe submit. Yeah I mean first of
27:00
all I think as I mentioned
27:02
in the world of cyber broadly
27:04
speaking there is infrastructure and human
27:06
level protection we focus on the
27:08
ladder that's the biggest surface
27:11
area of risk or threats and
27:14
the generative AI is actually to
27:16
some extent bit of
27:18
a tailwind for us because it
27:20
increases the potential of attacks makes
27:22
it creating an effective attack easier
27:25
and more available and
27:28
accessible to bad actors at
27:30
lower cost than them trying to do
27:32
it themselves. Okay so we are
27:34
seeing that as potentially
27:36
a tailwind and as
27:38
long as we have the right technology
27:40
to defend against those types of attacks
27:42
in general that sort of grows the
27:44
need for a sort of robust solution
27:46
like proof point. Secondly as
27:49
I mentioned at the end
27:51
of the day you need more sophisticated
27:53
models to prevent against those types of
27:55
attacks and those models can
27:57
only be sure you can have coders
28:00
write code faster, but without access
28:03
to that data, you're not gonna
28:05
be able to develop those models.
28:07
And so there is a inherent
28:09
incumbency benefit. Can you get data
28:12
elsewhere? Maybe possibly, but nowhere the
28:14
same quality of data that we
28:16
have as an incumbent when we
28:19
are already serving enterprise customers and
28:21
delivering their critical asset like mail
28:23
and other data protection solutions. So
28:26
I think second is that it
28:28
does give benefit to incumbents. I'd
28:30
say third, I don't think we are sitting
28:33
still. We are
28:35
leveraging the power of generative
28:37
AI and these large language models
28:39
to create more upsell and cross-sell
28:41
opportunities. For example, you
28:44
may have all experienced it. You may
28:46
have fat-fingered yourself, a
28:48
wrong person and sent, not
28:51
just an attack, but just information
28:53
that was not meant
28:55
to be sent to someone that
28:57
you sent an email to, and then you sort of
29:00
sent a note, oh, can you please delete it? That
29:02
was not for you. It's happened to all of us.
29:04
Because we are in the mail flow, we've built a
29:06
solution and we already have the context,
29:09
the example that I gave where I
29:11
sent that attack to Orlando. Instead of
29:13
an attack, it may have been just
29:15
an email that I accidentally was sending
29:17
to Orlando. And our technology now would
29:19
prompt me to say, hey, do
29:22
you really intend to send it to
29:24
Orlando? This doesn't look like the type
29:26
of information you usually share to him.
29:28
So it's a nudge as
29:30
AI technology. And that's extremely helpful if
29:32
you can think about even in your
29:35
businesses, in large banks, there are teams
29:37
of 40 to 50 people that are
29:39
simply checking outbound emails to all sorts
29:41
of clients if that email is the
29:43
right one to be sent or not.
29:46
We can take all of
29:48
that cost out and create that
29:50
incremental monetization all because of the
29:52
power of Gen AI. So enabling new
29:54
use cases because of the technology. So
29:57
A, it's a tailwind for us in
29:59
the cyber. B, incumbency helps us,
30:02
doesn't hurt us in any ways. And
30:04
C, with us standing still, we're building
30:06
new monetization and new solutions that are
30:08
serving real problems for our customers, which
30:10
are much, much easier for us to
30:12
bring to market than anyone else. Charlie,
30:15
do you worry about somebody building an Anna
30:18
Plan competitor quicker
30:20
and easier now
30:22
that code's easier to generate and you don't need
30:24
to be a specialist anymore? I
30:27
don't, primarily because if I'm
30:29
exposing my customer's data to
30:31
public generative AI models, I'm
30:34
done, right? Because they
30:36
now own my customer's data. And
30:38
so we have a proprietary platform that
30:40
really protects that data and the usage
30:42
of that data. So to
30:45
me, a lot of this comes back to data
30:47
at its core, right? Generative AI
30:49
is gonna allow us to get a lot more insight
30:51
out of data. And so what
30:55
we're working on to advance the insight
30:57
dimension that
30:59
I talked about earlier is really
31:01
leveraging generative AI to do
31:04
more dynamic forecasting. So really
31:06
early warning systems. So
31:08
you can imagine a company, let's say it's
31:10
a manufacturer and they've got distribution,
31:12
they've got big retail customers, think
31:14
Walmart or Target, really big customers.
31:16
And now they've got external events
31:18
that they actually wanna do, put
31:21
into a early warning system
31:23
like weather, right? Would be the simple example.
31:26
So if my big customers are running a
31:28
massive promotion on my product and I've got
31:30
a weather event, I could be
31:33
pretty out of sync with where inventory
31:35
is in my distribution centers. And I
31:37
could really risk irritating them because they
31:39
have a lot of stockouts when they've
31:41
got a big promotion coming, right? So
31:43
we're gonna use AI to actually create
31:45
early warning systems. And that was a
31:47
simple example Across multiple data
31:49
sets. So We can help our
31:51
customers forecast even more accurately and
31:53
actually get ahead, really get in
31:55
the predictive business so they can
31:57
avoid stockouts in my example. Rather
32:00
than you know what, what are we going to do
32:02
about a forecast? Will we know we're going to miss
32:04
inventory levels? That kind of
32:06
where we're headed out with Gen Vi. We
32:11
talked a lot about our our
32:13
portfolio today and driving efficiency. And.
32:16
You know? currently running at and. Forty. Percent
32:18
margin, sometimes more. It would be
32:20
great if you could walk through
32:23
examples internally so we talked to
32:25
Buy your product. We talked him
32:27
away during the customers the market.
32:30
How can you leverage this next
32:32
generation of Ai technology to across
32:34
your organization to drive operational efficiencies.
32:37
In all aspects or and when and where do you
32:39
see today when you see a going. On
32:42
today today our our called the low
32:44
hanging fruit exercises so where we are
32:46
left so for example cogeneration I were
32:48
at in the day. We write software
32:51
The ability actually I generate code that's
32:53
fairly standard. We use wall. Is.
32:55
With a eyes gotten much much better with
32:57
gone so far beyond you know can open
32:59
source in our those those those a public
33:01
of Alwan and that's how are a little
33:03
rhyme. Or. Indie percentage lower than
33:05
lot of our competitors raid. In addition, a
33:07
low cost locations here there everywhere. I'm a
33:10
big impact right now isn't so the sales
33:12
and marketing area where we're able to do
33:14
a lot more. Targeted are
33:16
selling targeted marketing. I'm.
33:19
Using our own capabilities inside o'clock in the
33:21
uses some third party tools like six as
33:23
exaggerate actually looking kind of buying patterns. Who's
33:25
in the mark Airless target them on. make
33:28
sure your our Santa Me is is not
33:30
knowing rights. I'd I'd data and I'll be
33:32
ideal coming out before we see them. Now
33:35
it's kind of the tools out there. We
33:37
can look out in the market with his
33:39
during certain searches and predict who's about the
33:41
likely by. Our software. Computer software
33:44
an axe ago. Target? That's so. be
33:46
adding that. See ya, That's the. As
33:48
I was it easy stuff doesn't have it's easy
33:51
but it's is right in front of us rightists?
33:53
it is right there and then you know going
33:55
forward. I think I'm the own for us. We're
33:57
We're pretty simple business right? We buy software. Me.
34:00
All software. But. There are opportunity
34:02
to help our customers who got massive
34:04
massive data, who got a lot of
34:06
bomb. Yeah. I'd say kind of
34:08
money processes and noom son of fight
34:10
through. stop is gonna be is going
34:12
to be almost unlimited and enter the
34:14
real. The the real kind
34:17
of differentiation we have on on. Really sad
34:19
about the town business because you're a lot
34:21
of people think that having more data is
34:23
better and more is not always better. some
34:25
has better is better. I said ability to
34:27
actually take the data the customer has Were
34:29
talking earlier sad about a large automotive manufacturers
34:31
to build their own. Infrastructure.
34:33
Their own L am their own, their own
34:36
ecosystem and will be helping them do now
34:38
is harness all the data inside of their
34:40
four walls. I'm including the data comes off
34:42
the cars and the sensors and everything in
34:45
real time and make decisions about him and
34:47
to worry about arm of where put the
34:49
next charging station for example and as a
34:51
sluggish use productivity I left for them and
34:54
then to protect Ip nothing goes out. And
34:56
when they bring. Data. From
34:58
the outside bacon. Approved.
35:00
Or acid of it. They can carry that day and
35:03
make sure that it's not going to inject bad
35:05
date or bias into their dataset. So for me, that's
35:07
the most exciting part of all. This is what we
35:09
can do for all of our customers. Semi.
35:13
To be great to hear about how
35:15
you. Today. You're in mid thirties
35:17
margin would you see operationally internally day
35:20
move that to and in mid forties
35:22
and fifty percent for be recorded to
35:24
see now. It's
35:26
a commitment as a commitment. Yeah, yeah.
35:29
wedding. I think. First of all, we're
35:31
We're We're in a similar state. starting
35:33
with you can call it low hanging
35:35
fruit. although the results of those have
35:37
been astonishing li positive in the code
35:40
development. Ah, Since developers have
35:42
embraced it, the success that we're
35:44
seeing in terms of code that
35:46
comes in as a suggestion that
35:49
people except it's basically is another
35:51
proxy to show and that is
35:53
productivity intrinsically being improved in terms
35:55
of how fast week been new
35:58
code so that certainly there's. So
36:00
how are you managing the security on that code,
36:02
which I know is a big issue? No, we,
36:05
you know, code right now sort
36:07
of goes into your own instance
36:09
of GitHub largely. So these are
36:11
appropriately sort of governed private
36:13
instances of code that is
36:15
running. So the right security for
36:18
IP protection is being put in place. But
36:20
I do, we do think that is improved
36:23
productivity without any compromise of IP
36:25
leakage that's been done. I
36:27
think where I'm most excited
36:29
about where we are just starting to
36:31
do is more predictive
36:33
analytics on churn. Because
36:35
we have a lot of indicators
36:37
of information from customers
36:40
on usage of the product and
36:42
how through that we can
36:45
create patterns and very quickly then
36:47
have the right reactive or
36:49
quickly corrective measures in a more
36:52
proactive manner put in place. Today,
36:55
a lot of that
36:57
is done somewhat manual, somewhat reactive
36:59
and just, you know, you can
37:01
imagine resource allocation is not
37:04
sort of really been put in
37:06
place as proactively as we could.
37:08
So using
37:11
models, you can think of them as
37:13
still fairly simplistic AI models, but these
37:15
days those models can greatly improve your
37:18
gross retention rate. And all of a
37:20
certain, just a single point
37:23
of gross retention rate can give a
37:25
significant boost to the earnings flows through.
37:27
So I think that's the one in
37:29
addition to sales and marketing I'm excited
37:32
about. We're starting on that. And then
37:34
there is some education,
37:36
tech support, again, public information
37:38
that's available, how we can
37:41
just streamline it so that there are less calls
37:44
coming in and improves the productivity
37:46
of when we actually serve The
37:48
customers calling in, we are doing so in
37:50
a faster way and whatnot.. Because Usually there's
37:52
a lot of time taken by tech support
37:54
engineers to train them and whatnot.. And these
37:57
days, no one really wants to call. So
37:59
If there is..., Agent that can serve
38:01
large volume of support support requests that
38:03
the another investment we're making to make
38:05
sure we can be more efficient. Straight
38:08
Charlie and any saying the using your
38:10
is so very similar theme so lives
38:12
I want we're working with these guys
38:15
said But what one thing I'll add
38:17
as we're. We're. Starting to leverage
38:19
ai pretty deeply and cells productivity
38:21
in that scenario that we're working
38:23
on improving in the business michigan
38:25
a small margin expansion. And
38:28
we've lovers a third party or a I
38:30
told to do this. but one of things
38:32
that we discovered is that I'm We went
38:34
into Loesser thinking that your head of a
38:37
A meetings a week to reach her father
38:39
as a salesperson at that activity level would
38:41
drive it. We. Learned was that
38:43
says she not a meetings especially for
38:45
meetings with director level or above that
38:48
generates I shall hundred and twenty five
38:50
percent of one in shipments. And
38:52
so then the question becomes well who
38:54
has the relationships right? What's the relationship
38:57
map look like Of course we find
38:59
that we're not where we need to
39:01
be to generate those for high quality
39:03
meetings. a weird across some buddies portfolio
39:06
and so that allows a sad than
39:08
as the next question which is how
39:10
we're going to leverage Ai to access
39:12
the right relationships whether it's third party
39:14
data. Be a linked in
39:16
or something else and that's really
39:19
what we're working on now because we
39:21
know that that relationship exists, that you
39:23
know better meetings, words more senior people
39:26
allows us to focus the activity
39:28
and be more productive at the same
39:30
time. I'm and in a when we
39:32
marry that with our product and we
39:35
actually have an Ai product called Predictive
39:37
Insights And what their product does is
39:39
it helps our customers are identified
39:41
the ideal customer profile. So now you've
39:44
got the right customer profile. You got
39:46
much more targeted meetings and that should
39:48
be a much more productive formula for
39:50
driving new bookings. So.
39:53
All of that leverage their. Maybe with
39:55
the last question? How are
39:57
you balancing? Where. We're at
39:59
today and the pace of innovation just in
40:02
this area going forward in at you know
40:04
how much time are you spending on the
40:06
future versed in the present because. I
40:09
think of this and this may be
40:11
this platform Sept is moving out of.
40:13
the platform shifted as generate this. You
40:15
know the technology innovation here is moving
40:17
as fast as anything we've seen. And
40:20
the absorption rate has been really high. and
40:22
the fact that you're using it today we're
40:24
talking about it in a way that very
40:27
tangible is impressive given we're really just a
40:29
couple years and it has generated a i
40:31
say but simeon have had. how do you
40:33
think about that a manner sad, just that
40:36
piece of innovation and any sauce and what's
40:38
next day? I think first of all this
40:40
technology as many of the have club probably
40:42
dead in terms of just adoption from filled
40:45
with population all that it's off the charts
40:47
in terms of sexy pity adoption. so intense
40:49
for us. We have to be
40:52
as a cyber defense company,
40:54
a step ahead of the
40:56
bad actors at this point
40:58
the time. the internals machinery
41:00
at proof point is assuming
41:02
that the. Bad actors
41:04
have access freeform access to this
41:06
technology that sets will be created
41:08
using this and be are leveraging
41:10
the technology to ist fullest when
41:13
it comes to building the defense.
41:15
Against it that is no this.
41:17
that said topic as our city
41:20
offices sort of squarely responsible for
41:22
it or we did an acquisition
41:24
of the company. contests in there's
41:26
a lot of effort was as
41:28
going in to integrate. It brought
41:31
in allies language model of our
41:33
own. With. The right sort
41:35
of costs up to that can
41:37
build evades it's contextual grass. You
41:39
may have heard this term which
41:42
essentially created relationship between large number
41:44
of objects so that you can
41:46
very quickly you know identify patterns
41:49
that eyes off the you know
41:51
the norm and so both through
41:53
organic efforts as well as odd
41:55
investments via Amman A V R's
41:58
on that front yard. No,
42:00
we're not assuming this is something that's
42:02
that's gonna happen in the future. We
42:04
have to stay a step ahead and
42:06
that sort of gives us credibility. With.
42:09
Our customers that creates momentum with
42:11
customers. a drive somewhat of growth
42:13
as well for us. So that's
42:15
sort of autism sin. Or. When
42:17
it comes to a cold business, Us
42:19
Cyber. For. Areas that
42:21
we have: Data Protection. Portfolio.
42:24
And Desist as well which is a
42:26
growth business in addition to email security
42:29
or data protection business where we are
42:31
closely monitoring. There's what. Tangible.
42:34
Innovations we could build so
42:36
that when customers start adopting
42:38
a I for their critic
42:40
the lines of business. Applications
42:42
use cases. The example that I
42:45
was giving bad there was a
42:47
chalkboard talking to various applications. Power
42:49
data protection engine can just how
42:52
we can ride that wave at
42:54
that adoption happens. So for Descend
42:56
solutions we're assuming it's years for
42:59
our data protection solutions. We have
43:01
a closely monitoring how and when
43:03
the customers adopted so that odd
43:05
innovations and on behind of But
43:08
it's it's A it's a topic
43:10
we're discussing on an ongoing basis.
43:12
Because his can't afford not to. Heading
43:16
for us the a week we built a
43:18
lot of structure to make sure that we're
43:20
intern or sorts ottoman or Ai council which
43:22
is really important rather gives is one perspective.
43:24
We have us a large customer council or
43:27
that advises on a i'm what they're doing,
43:29
what they're seeing in the markets and we
43:31
have representation from all industries on that that
43:33
council to help us go. I'll tell you
43:35
though, the the biggest secret weapon we've got
43:37
actually is our I'm In A Strategy and
43:40
our partnership with Summer Bravo. We got me
43:42
on my top in In and Money's and
43:44
we're constantly scanning the market looking. For. Cool.
43:47
A item is where the during a lot of
43:49
times the best source of innovation is to look
43:51
at what's out there and the market. Where's the
43:53
metro money flowing and what's going on out there
43:55
to predict how know what the next wave of
43:57
innovations gonna be and that is super helpful and
43:59
and understanding. What we need to get ready for
44:01
a while we we will. We should be buying. And
44:03
you know as B C Young dries up in some
44:05
of these opportunities were going to be all over him.
44:08
China to buy you know, pennies on the dollar for
44:10
went into these things. Now.
44:12
And I would say in a we're
44:15
going at a measured pace right word
44:17
center earlier in the whole than I
44:19
think as some other companies have just
44:22
gone through a big transformation of the
44:24
early stages of transformation. So we're we're
44:26
gonna in continued of take Arrayah legacy.
44:29
Leverage it into the products in the platform
44:31
that we have. I can imagine a fusion.
44:33
I've started talking to my. I had
44:36
a product and technology about this where we
44:38
have a leader of a I products I'm
44:40
in that will help us with an with
44:42
they are and M and m in a
44:44
strategy to us to augment what we do
44:46
their and that's a conversation which we just
44:48
started but I suspect we'll get there are
44:50
some time this year because it is such
44:52
a big topic and some big topic for
44:54
innovation will we are of the shot clock
44:56
is almost out. We this
44:59
has been great. I think the thing
45:01
that's most interesting is often times you
45:03
know you you hear a lot about
45:06
in the venture industry innovation taking place
45:08
in in areas like A I. One
45:10
of the big take away today that
45:12
we would like to born here to
45:15
understand is that are companies are innovating
45:17
in and around these areas in big
45:19
ways with massive are d budgets not
45:22
only delivering innovation to customers and to
45:24
the industry, also creating policy. you know,
45:26
thinking about ethics in. Which
45:28
is incredibly important but then also taking
45:31
this new technology and driving it into
45:33
internal operations to produce better business results.
45:35
So thank you for the great work
45:38
as and fascinating and that we we
45:40
really appreciate or times you being here
45:42
tonight and a surplus of. This
45:49
is our final episode of season since I
45:51
wanted to take a moment to say thank
45:54
you to all of you for coming along
45:56
on this journey with us. We've
45:58
done to share something from those
46:01
stories of hundred partnerships and we
46:03
appreciate of illicit. Will be taking
46:05
a brief break to record more
46:07
episodes but we'll be back again
46:09
to take you behind the deal
46:11
later this year. Hum Orlando Bravo!
46:13
Thanks from is. Certain.
46:26
Statements about tell Him Babbel made
46:28
by portfolio companies exodus are intended.
46:30
To illustrate Summer Bibles business relationship
46:32
with such persons, Rather than tell
46:34
my brothers capabilities. Are expertise with respect.
46:36
To advisory services portfolio company
46:39
executives were not compensated in
46:41
connection with their podcast participation,
46:43
although they generally receive compensation
46:45
and investment opportunities in connection
46:47
with their portfolio company. Roles
46:49
and in certain cases are also
46:51
owners of portfolio company securities and
46:53
or investors and Tell the Bravo
46:55
funds sense compensation and investments subject
46:57
podcast. Participants to potential conflicts. Of
47:00
interest.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More