WEBVTT

NOTE
This file was generated by Descript 

00:00:09.012 --> 00:00:09.202
Raju: Okay.

00:00:09.212 --> 00:00:10.542
Hello, listeners and viewers.

00:00:10.942 --> 00:00:13.532
Welcome to another episode of RRE POV.

00:00:14.072 --> 00:00:16.832
I'm Roger Ricci, and I'm joined by my partner, Will Porteous.

00:00:17.121 --> 00:00:17.822
Will: Hey, everybody.

00:00:18.272 --> 00:00:22.601
Raju: At RRE, we're a well-established venture capital firm, and we've raised

00:00:22.621 --> 00:00:27.261
eight venture funds and two opportunity funds over the span of 32-plus years.

00:00:27.581 --> 00:00:30.821
And as such, Will and I know this really well.

00:00:31.071 --> 00:00:33.881
He actually spends a lot of time talking to LPs.

00:00:34.422 --> 00:00:37.401
We've had the absolute pleasure of having worked with some of the best

00:00:37.401 --> 00:00:42.641
limited partners across institutions, corporations, and family offices.

00:00:43.622 --> 00:00:48.988
And As Will and I both know, over the past year or so, we've been asked by

00:00:49.007 --> 00:00:53.447
three or four of our largest LPs to help them think through their AI strategies.

00:00:54.158 --> 00:00:57.337
Um, so this is what this episode is.

00:00:57.447 --> 00:01:00.937
This one goes out to the ones we love, our LPs.

00:01:02.608 --> 00:01:06.587
Which reminds me of that R.E.M. song, and I kind of

00:01:06.588 --> 00:01:09.137
feel like I need to acapella a short riff of this.

00:01:09.157 --> 00:01:10.118
Will: Are you gonna sing for us, Raju?

00:01:10.727 --> 00:01:11.448
Raju: I think I am.

00:01:11.678 --> 00:01:12.078
I am.

00:01:12.377 --> 00:01:13.567
Uh, just totally acapella.

00:01:13.937 --> 00:01:14.187
All right.

00:01:14.787 --> 00:01:19.137
This one goes out to the one I love.

00:01:20.367 --> 00:01:26.048
This one goes out to the one I've left behind.

00:01:26.917 --> 00:01:31.538
A simple prop to occupy my time.

00:01:31.847 --> 00:01:36.608
This one goes out to the one I love.

00:01:36.907 --> 00:01:37.387
Fire!

00:01:39.737 --> 00:01:40.047
Anyway.

00:01:40.167 --> 00:01:40.597
Okay.

00:01:40.657 --> 00:01:41.017
All right.

00:01:41.077 --> 00:01:41.797
I had to do it.

00:01:42.397 --> 00:01:43.047
Will: Absolutely.

00:01:43.047 --> 00:01:48.417
With gratitude to, uh, to Michael Stipe and to our LPs out

00:01:48.417 --> 00:01:48.507
Raju: there.

00:01:48.507 --> 00:01:49.407
Absolutely.

00:01:49.438 --> 00:01:50.727
Sorry for that diversion.

00:01:50.727 --> 00:01:51.788
You guys know I love music.

00:01:51.837 --> 00:01:52.507
I just can't.

00:01:52.578 --> 00:01:54.978
I get… Okay, w- back to the main programming.

00:01:55.618 --> 00:01:58.507
So what do LPs do about incorporating AI into their workflow?

00:01:59.258 --> 00:02:02.907
This is… You know, recently, I've had intimate one-on-one sessions

00:02:02.907 --> 00:02:07.687
with several of our LPs, and I usually walk them through the following.

00:02:08.428 --> 00:02:12.767
And by the way, every one of these guys is overworked, everyone.

00:02:12.837 --> 00:02:13.328
I mean-

00:02:13.497 --> 00:02:16.557
Will: Yeah, I think that's the, the most important starting point.

00:02:16.597 --> 00:02:20.928
I, I got a lot of sympathy for our LPs sitting on the other side of the table

00:02:20.937 --> 00:02:25.607
trying to manage all the inflow of new managers they're looking at, trying

00:02:25.608 --> 00:02:28.827
to stay on top of their existing fund managers, trying to think about their

00:02:28.828 --> 00:02:33.437
portfolio construction model, trying to do it all with a really small team.

00:02:33.917 --> 00:02:34.977
It's, it's a lot.

00:02:35.307 --> 00:02:36.537
They're unsung heroes.

00:02:36.788 --> 00:02:39.457
Raju: Yeah, and, and you know, like, we feel that as well.

00:02:39.468 --> 00:02:47.137
Like, there's kind of a, like, small divide between limited LPs and GPs.

00:02:47.137 --> 00:02:47.717
Small divide.

00:02:47.747 --> 00:02:51.287
I mean, we're both in flux with tons and

00:02:51.288 --> 00:02:54.067
tons of, you know, sort of opportunities.

00:02:54.387 --> 00:02:55.697
We have to evaluate them.

00:02:55.697 --> 00:03:01.457
We've got to go deep into diligence, and it's just gets overwhelming.

00:03:01.468 --> 00:03:07.197
And a lot of these organizations just aren't like, you know, kind of flush with

00:03:07.197 --> 00:03:12.397
analysts and associates and, you know, principals to help do all that work.

00:03:12.397 --> 00:03:15.188
So I think AI is gonna be a really important tool.

00:03:15.218 --> 00:03:20.348
So, um, what I want to start with, Will, is just, um, a bit for

00:03:20.377 --> 00:03:25.732
our LPs for them to understand today, just today What are the

00:03:25.742 --> 00:03:30.762
LLMs actually good at, and what are they really overhyped at?

00:03:31.252 --> 00:03:32.512
So yeah.

00:03:32.932 --> 00:03:38.422
Yeah, maybe we can just like, um, go into some context around those pieces

00:03:38.442 --> 00:03:43.211
and, and I know you u-used it as well, but like what it's good at, you

00:03:43.211 --> 00:03:49.261
know, if you take multiple files, including long text and, you know, just

00:03:49.262 --> 00:03:53.852
a bunch of PowerPoints, spreadsheets, whatever, and you need to summarize

00:03:53.852 --> 00:03:58.181
it, or you need to draft initial content, it's not so bad at that.

00:03:58.361 --> 00:04:01.311
You know, you can think about just the

00:04:01.311 --> 00:04:04.401
analysis of a particular company or a space.

00:04:04.411 --> 00:04:06.942
Like you're getting into a space, you're, you're looking at a sector

00:04:07.271 --> 00:04:11.351
and you're, you know, copied like five or six key documents and you

00:04:11.352 --> 00:04:14.292
can throw it into one of these LLMs and it can give you some feedback.

00:04:14.771 --> 00:04:16.362
Um, I think it's pretty good at that.

00:04:16.372 --> 00:04:17.171
What do you, what do you think?

00:04:17.212 --> 00:04:18.462
What's, what's your feedback on that?

00:04:18.471 --> 00:04:19.381
Will: Oh, absolutely.

00:04:19.381 --> 00:04:19.741
Yeah.

00:04:19.791 --> 00:04:23.152
Uh, summary and synthesis and kind of k- helping you

00:04:23.152 --> 00:04:25.721
get to the heart of the matter and learn quickly.

00:04:26.181 --> 00:04:26.551
Yeah.

00:04:26.582 --> 00:04:29.932
I mean, I, I think, I think we get leverage out of it that way.

00:04:29.932 --> 00:04:31.331
I think our LPs can too.

00:04:31.782 --> 00:04:32.132
Raju: Yeah.

00:04:32.331 --> 00:04:34.982
And I think most people are starting to experiment

00:04:34.982 --> 00:04:38.031
with that or use that in a, a more rigorous way.

00:04:38.971 --> 00:04:44.461
The second that it's good at that, you know, people are le- a little bit more

00:04:44.462 --> 00:04:50.852
reluctant to use, I would say, but maybe they should start, right, is, and

00:04:50.852 --> 00:04:55.292
this is on a personal level, is just, you know, AI-based workflow automation.

00:04:55.841 --> 00:04:58.846
Simple stuff like AC NAC.

00:04:58.986 --> 00:05:01.486
You know, read an email, draft a response.

00:05:01.596 --> 00:05:02.426
Don't send it.

00:05:02.796 --> 00:05:03.455
Don't, don't send it.

00:05:04.226 --> 00:05:05.536
Do- don't have it send.

00:05:05.856 --> 00:05:07.975
Um, although I, I know people who've tried this.

00:05:07.996 --> 00:05:10.315
You know, they just have, like, automated, like,

00:05:10.345 --> 00:05:13.916
the response function, and it will bite them.

00:05:15.376 --> 00:05:17.006
Every once in a while it bites them.

00:05:17.626 --> 00:05:20.595
So I, I do think that's a tool.

00:05:20.616 --> 00:05:21.455
I, I've tried it.

00:05:21.485 --> 00:05:22.035
Have you tried it?

00:05:22.565 --> 00:05:24.686
Will: I have tried it, and, you know, the

00:05:25.056 --> 00:05:30.425
I, I, I tend to get these really long responses, responses

00:05:30.725 --> 00:05:35.216
that go kind of ramble and go, go kind of all through all

00:05:35.225 --> 00:05:38.346
the permutations without necessarily getting to the point.

00:05:38.346 --> 00:05:40.885
So it has, it has to be fine-tuned.

00:05:40.955 --> 00:05:44.655
Um, you have to invest the time to make it work for you.

00:05:44.665 --> 00:05:47.985
But I, I agree, and, and, you know, email workflow

00:05:48.275 --> 00:05:51.335
is such a time sink, uh, for all professionals.

00:05:51.345 --> 00:05:55.055
So it- there's a gift there in terms of productivity, to be sure.

00:05:55.335 --> 00:05:55.595
Raju: Yeah.

00:05:55.606 --> 00:05:59.825
I, I think so too, and, you know, I, I think it's better over time.

00:05:59.845 --> 00:06:05.046
Um, but I, I don't think it's ready to actually make decisions on your behalf.

00:06:05.065 --> 00:06:07.625
It can draft, and drafting is actually quite good.

00:06:07.945 --> 00:06:10.995
And so that's something that I think it's good at.

00:06:11.315 --> 00:06:16.595
The last piece is something that is, uh, really important in the enterprise,

00:06:16.605 --> 00:06:20.755
which is something called RAG architecture, and that's connecting to

00:06:21.195 --> 00:06:26.735
private business data so that, you know, you can get more, I guess, targeted

00:06:26.736 --> 00:06:31.515
answers, um, that are focused on data that's sort of proprietary to you.

00:06:31.986 --> 00:06:35.255
Uh, and I, I think the architectures are getting better at that.

00:06:35.705 --> 00:06:38.695
Um, and so I think you would agree with those three, you know,

00:06:38.695 --> 00:06:43.115
sort of core business-oriented tools that could be useful.

00:06:43.455 --> 00:06:43.965
Will: 100%.

00:06:44.306 --> 00:06:46.735
Raju: The overhyped part is actually making

00:06:46.736 --> 00:06:51.490
business decisions And this is really important.

00:06:51.900 --> 00:06:55.789
Uh, I think if you sit there and say, "Allow it to make business

00:06:55.789 --> 00:07:00.750
decisions," a lot of what we do is relatively unique decision-making.

00:07:00.890 --> 00:07:04.529
It's like a little bit of an art form, um, because if

00:07:04.530 --> 00:07:09.999
you just focus AI on helping you make a decision based

00:07:10.020 --> 00:07:13.799
on prior results, you don't get breakthrough innovations.

00:07:14.159 --> 00:07:17.389
And this is why, you know, there's a lot of controversy around whether

00:07:17.969 --> 00:07:24.110
AI is going to be effective at art because art kind of skims the edges.

00:07:24.470 --> 00:07:28.089
And so you gotta think outside the box.

00:07:28.090 --> 00:07:32.969
You can't train the entire model based on like, "Hey, give me

00:07:32.970 --> 00:07:35.030
the norm, give me the median, give me the median, give me the

00:07:35.030 --> 00:07:38.349
median," and have it, you know, kind of push the envelope.

00:07:38.729 --> 00:07:43.069
Um, and so, you know, kind of making d- business decisions, at least in our

00:07:43.069 --> 00:07:49.329
world, it's not like, you know, I've done this task 1,000 times and 99.9% of

00:07:49.329 --> 00:07:52.499
the time it's always the same answer and, you know, like there's an outlier.

00:07:53.439 --> 00:07:55.270
Th- there's y- you, there's sort of like

00:07:55.280 --> 00:07:58.059
esoteric factors that go into decision-making.

00:07:58.060 --> 00:08:00.539
So, uh, you know, I don't recommend that.

00:08:00.909 --> 00:08:04.030
Um, and you know, I, I'm sure you've seen the same,

00:08:04.109 --> 00:08:06.699
same in, in a lot of the thinking that you've had.

00:08:06.729 --> 00:08:07.269
Will: Sure.

00:08:07.270 --> 00:08:11.039
I, you know, the thing that I, I struggle with is the, the, the

00:08:11.039 --> 00:08:15.600
results that you get with AI-based decision-making are often, uh,

00:08:15.600 --> 00:08:20.240
uh, over-weighted equally in the sense that y- you, there, there's

00:08:20.259 --> 00:08:24.239
not the, the appropriate weight being given to decisive factors.

00:08:24.279 --> 00:08:29.650
And so you may get nice sounding middle of the road answers in a lot of

00:08:29.650 --> 00:08:33.469
situations where there's, there needs to be a clear cut yes or no, and

00:08:33.469 --> 00:08:37.109
there, there are h- there are high stakes associated with that yes and no.

00:08:37.899 --> 00:08:41.699
And, uh, we, we have yet to sort of make that, that leap to

00:08:41.719 --> 00:08:46.930
critical decision-making, um, threshold with the AI tools that I

00:08:46.939 --> 00:08:50.329
think most, most businesses and most individuals have access to.

00:08:50.799 --> 00:08:52.289
Raju: And you know, just keep in mind, this

00:08:52.299 --> 00:08:55.960
podcast is definitely sort of like time-bound.

00:08:56.309 --> 00:08:59.390
We're gonna fix a lot of this stuff over time and, you know,

00:08:59.449 --> 00:09:02.769
maybe has a lifespan of a year, maybe has a lifetime, half a year.

00:09:02.770 --> 00:09:05.880
But certainly for now you don't wanna making business decisions.

00:09:06.180 --> 00:09:08.000
And the, and the last two over-hyped things is,

00:09:08.009 --> 00:09:10.120
you know, factual accuracy and complex math.

00:09:10.369 --> 00:09:12.699
We know there's limitations around that.

00:09:13.199 --> 00:09:16.619
You know, you gotta … It's almost like you gotta talk to your teenager.

00:09:17.520 --> 00:09:17.629
You know?

00:09:18.449 --> 00:09:21.559
It j- uh, factual accuracy is just a little bit off.

00:09:22.059 --> 00:09:25.559
You know, it tries, but you know, w- and we've talked about this in prior

00:09:25.559 --> 00:09:30.900
podcasts, Will, which is it's not programmed to say, "I don't know."

00:09:31.630 --> 00:09:32.170
Will: Indeed.

00:09:32.490 --> 00:09:35.604
Raju: It's programmed to give you an answer It, it's

00:09:35.864 --> 00:09:39.934
really desirous for the AI to give you an answer

00:09:40.054 --> 00:09:44.604
Will: And to exaggerate its, its confidence, its, its certainty in that answer.

00:09:44.674 --> 00:09:45.134
Raju: Exactly.

00:09:45.164 --> 00:09:45.744
Exactly.

00:09:45.754 --> 00:09:49.434
You know, so they're all kind of, like, focused around giving you that answer.

00:09:49.434 --> 00:09:53.753
And, um, you know, even if it's a math answer or it's

00:09:53.773 --> 00:09:55.513
doesn't have enough data, it doesn't say, "I don't

00:09:55.513 --> 00:09:58.323
know," and we gotta retrain them to do things like that.

00:09:58.734 --> 00:10:02.583
And then there's bigger limitations that we've talked about in the just prior,

00:10:02.663 --> 00:10:06.604
well, two podcasts ago, which is, you know, we have context window limitations.

00:10:06.604 --> 00:10:09.264
That means if you've got massive amounts of data

00:10:09.274 --> 00:10:11.944
and you ask a difficult question, like a complicated

00:10:11.944 --> 00:10:16.234
question, there's just not enough tokens that can- Yeah

00:10:16.243 --> 00:10:17.434
get you that answer.

00:10:17.734 --> 00:10:20.523
Um, so you can either ask a very simple question in a large

00:10:20.523 --> 00:10:23.093
data lake, or you can ask a larger question in a small data

00:10:23.094 --> 00:10:27.053
lake, but, you know, as soon as you get big data lakes and you

00:10:27.054 --> 00:10:29.974
ask a complex question, you're-- it's just gonna barf on that.

00:10:30.223 --> 00:10:31.353
And then cost, right?

00:10:31.374 --> 00:10:36.603
Those are the things that we understand as limitations, you know, across.

00:10:36.603 --> 00:10:38.793
So that's kind of a grounding function for everyone.

00:10:39.224 --> 00:10:39.563
You know?

00:10:39.754 --> 00:10:45.843
It does a good job today summarizing files, you know, drafting initial content,

00:10:46.353 --> 00:10:52.413
um, doing some automation of a workflow, like drafting an email response.

00:10:52.974 --> 00:10:56.853
It does a good job in connecting with private, you know, sort of business data.

00:10:57.364 --> 00:11:01.853
What it's over-hyped at is making actual decisions, giving you factual

00:11:01.873 --> 00:11:06.853
accuracy across the board that you can trust every time, and complex math.

00:11:07.144 --> 00:11:11.053
And then there's bigger questions about cost and the types of

00:11:11.053 --> 00:11:14.933
questions that you can answer, um, depending on the size of the lake.

00:11:14.934 --> 00:11:21.773
So what I'm gonna do to divide this next piece up is, um, how to get started.

00:11:22.133 --> 00:11:25.464
You know, what are the areas where we think all LPs should

00:11:25.473 --> 00:11:29.053
be playing around with given that they're incredibly busy?

00:11:29.453 --> 00:11:31.884
And then we're gonna talk a little bit about

00:11:31.884 --> 00:11:34.933
the specific nuances around corporate LPs.

00:11:35.473 --> 00:11:39.414
Um, and we have a fair number of corporate LPs and, and,

00:11:39.414 --> 00:11:43.654
um, for those of that you that aren't yet LPs, you know,

00:11:43.654 --> 00:11:45.993
we've got a very sophisticated program for them, you know.

00:11:45.993 --> 00:11:50.514
And that program involves, you know, giving them a feel for

00:11:50.894 --> 00:11:55.413
technology trends that are pertinent to their sector, um, but

00:11:55.423 --> 00:11:59.543
also these kinds of, you know, touchpoints where we'll spend

00:11:59.543 --> 00:12:03.273
a couple hours and we'll go through their line of business.

00:12:03.283 --> 00:12:06.793
And if you're a manufacturer, that means how do we get

00:12:06.793 --> 00:12:10.254
started with AI in manufacturing, and how do we get started

00:12:10.254 --> 00:12:12.923
in finance, and how do we get started in other sectors?

00:12:13.423 --> 00:12:20.213
So ground floor, all LPs, family offices, corporates, institutions

00:12:21.038 --> 00:12:24.038
You know, first thing I always tell them is you got to start

00:12:24.178 --> 00:12:27.718
playing with the tools, and you got to start taking some training.

00:12:28.158 --> 00:12:31.488
And, you know, this kind of goes into one of the other podcasts that we've

00:12:31.488 --> 00:12:36.747
had, Will, where we talked about sort of where, what's that layer that

00:12:36.768 --> 00:12:41.747
actually is the starting point for every technology inflection shift?

00:12:42.288 --> 00:12:44.717
And, you know, I've always said it's been services, and we

00:12:44.718 --> 00:12:47.358
tried to, the world tried to poo-poo that and say it was,

00:12:47.358 --> 00:12:49.147
"Oh, no, we're going to go right to platform companies."

00:12:49.167 --> 00:12:54.798
But the reality is you got to start playing with the tools, and that

00:12:54.798 --> 00:12:59.188
may mean that you bring some outside resources in, um, to train.

00:12:59.448 --> 00:13:02.828
And I'm sure you're coaching your companies that are sort of thinking this

00:13:02.927 --> 00:13:07.988
through as well that aren't native AI, you know, like how to get started.

00:13:07.988 --> 00:13:10.848
But, you know, that's kind of one of the things I've been talking about

00:13:11.520 --> 00:13:11.920
Will: Sure.

00:13:12.410 --> 00:13:16.380
Raju: The second area is maybe just bringing in some AI skills in-house.

00:13:16.770 --> 00:13:20.060
Consultants, maybe an analyst that you're gonna hire.

00:13:20.310 --> 00:13:24.830
Make AI skill sets a kind of a requirement, you know,

00:13:24.859 --> 00:13:31.009
across the board for, uh, those individuals, and start

00:13:31.019 --> 00:13:33.560
doing the things that we talked about earlier, right?

00:13:33.560 --> 00:13:37.300
Like, let's look at, you know, basically how do

00:13:37.300 --> 00:13:42.180
I do sort of email or, um, drafting a response.

00:13:42.220 --> 00:13:45.909
Um, and so those things are relatively important.

00:13:46.619 --> 00:13:48.559
What I wanna jump to from there, and those are kind

00:13:48.560 --> 00:13:51.709
of obvious, i- is a couple of our portfolio companies.

00:13:51.950 --> 00:13:52.846
I'd really like to jump in.

00:13:52.846 --> 00:13:54.479
I'd love for you to sort of ta- talk about the

00:13:54.479 --> 00:13:56.029
first one, and I'll talk about the second one.

00:13:56.519 --> 00:14:00.249
And the first one is, you know, one of our GPs, uh,

00:14:00.269 --> 00:14:04.060
Vic, is CEO and founder of this company Originalis.

00:14:04.450 --> 00:14:11.469
And Originalis is an AI-native tool set designed for,

00:14:11.499 --> 00:14:15.050
you know, both venture as well as, you know, our LP set.

00:14:15.419 --> 00:14:21.319
And that is a really fast path for a bunch of people who, on this

00:14:21.319 --> 00:14:25.819
call, who may be sitting there saying like, "How do I kickstart this?

00:14:26.050 --> 00:14:27.260
Yeah, I get it, you know.

00:14:27.279 --> 00:14:29.979
I can go and start playing around with it.

00:14:29.999 --> 00:14:34.769
I can go, you know, replace an analyst with somebody who's got AI skill sets.

00:14:34.779 --> 00:14:39.630
I could start having it, you know, compose some emails, but I

00:14:39.630 --> 00:14:44.329
don't have time." You know, uh, and, and, and so, you know, maybe

00:14:44.419 --> 00:14:47.179
spend a few minutes and kind of go through some of the things

00:14:47.179 --> 00:14:50.359
that Originalis can do, and maybe on a subsequent podcast, we'll

00:14:50.359 --> 00:14:53.779
have Vic back on and have him talk about kind of the capabilities.

00:14:53.779 --> 00:14:56.280
But I think this is an important one for people to understand.

00:14:56.290 --> 00:15:01.559
This applies to any one of our LPs that, you know, independent of

00:15:01.559 --> 00:15:05.710
whether they're institutional, corporate or, you know, family offices.

00:15:05.839 --> 00:15:08.589
Will: It's important to begin by the kind of

00:15:08.639 --> 00:15:10.920
assessing the origin where all this came from.

00:15:10.920 --> 00:15:15.159
Because it began as an initiative internally for, for us to get

00:15:15.169 --> 00:15:18.840
more leverage in our investment decision-making process and our due

00:15:18.840 --> 00:15:22.609
diligence activities, and then in our portfolio management activities.

00:15:23.039 --> 00:15:27.370
And, uh, you- We began our, our commitment to the Originalis

00:15:27.400 --> 00:15:31.440
team about 18 months ago and, and began using the tool

00:15:31.440 --> 00:15:36.140
in earnest here at RE about nine months to a year ago.

00:15:36.140 --> 00:15:42.020
And over time, uh, it's grown into a, a fundamental platform for

00:15:42.020 --> 00:15:45.649
us that supports all new investment decision-making consideration.

00:15:46.110 --> 00:15:52.379
And it solves both the complex and the mundane in, in, in that process.

00:15:52.430 --> 00:16:00.039
Um, we, we use the Originalis platform to ingest, uh, data sets and due

00:16:00.090 --> 00:16:06.099
diligence items and reference call notes and, um, a whole-- all the artifacts

00:16:06.099 --> 00:16:10.390
that emerge in the process of considering, uh, a new investment decision.

00:16:10.390 --> 00:16:15.469
And it's that synthesis properly weighted that I think is created the

00:16:15.469 --> 00:16:20.499
first aha moment for us as a partnership, which is all of a sudden it, it

00:16:20.500 --> 00:16:25.429
has become much, much easier to keep the firm as a whole abreast of what's

00:16:25.439 --> 00:16:29.679
happening in the consideration of a new investment, and to package that

00:16:29.679 --> 00:16:34.319
information i-in the right ways with the right level of, of importance,

00:16:34.619 --> 00:16:39.630
uh, so that people are, are kind of immediately in the flow rather than

00:16:39.639 --> 00:16:44.379
waiting until the end of hearing about something to read a perfectly crafted

00:16:44.380 --> 00:16:48.069
investment memo that they may have waited weeks to get into their hands.

00:16:48.289 --> 00:16:51.309
The process can now move along in a much more dynamic way.

00:16:51.609 --> 00:16:54.679
The wonderful thing we discovered in the process of this is that the

00:16:54.679 --> 00:16:59.809
Originalis tools that are so good at helping us weight various factors

00:16:59.810 --> 00:17:04.359
in considering a new investment for the portfolio are equally good in

00:17:04.359 --> 00:17:08.989
evaluating fund managers, both in the context of a new fund commitment-

00:17:09.139 --> 00:17:10.529
Raju: That's super interesting, yeah

00:17:10.599 --> 00:17:12.589
Will: and even when it comes to monitoring.

00:17:12.709 --> 00:17:16.290
And, and so a lot comes down to that RAG issue, that

00:17:16.290 --> 00:17:20.890
willingness to feed it, uh, li- internal data, legacy data.

00:17:21.160 --> 00:17:27.199
But, but once trained, uh, a, a firm can apply its own rubric and get

00:17:27.200 --> 00:17:30.740
really powerful, really useful output to support their decision making.

00:17:30.740 --> 00:17:37.159
And today we see a great group of early adopters among our LP base

00:17:37.429 --> 00:17:40.390
who are using the Originalis platform for their own decision making.

00:17:40.629 --> 00:17:43.716
Raju: You know, may-maybe just Talk about some

00:17:43.716 --> 00:17:45.666
of the, you know, one of the capabilities.

00:17:45.666 --> 00:17:49.956
Just pick one area that Originalis can sit there and say, "Hey, look,

00:17:50.305 --> 00:17:54.655
you're a limited partner. You see you've got a bunch of managers that

00:17:54.656 --> 00:17:58.445
you have to support today, and then you've got to sort of support,

00:17:58.456 --> 00:18:02.446
you're evaluating potential new managers, or every one of your l-

00:18:02.495 --> 00:18:06.235
your, your GPs is kind of every few years kind of coming back."

00:18:06.485 --> 00:18:12.345
What's like one output that, you know, Originalis takes care of or drafts?

00:18:12.505 --> 00:18:15.635
It doesn't have to be like the end, end game, but just gets you

00:18:15.635 --> 00:18:19.215
far enough along that it's like, oh my God, like it just saved me

00:18:19.745 --> 00:18:22.925
a ton of work, and we all know how hard this is for everyone here.

00:18:22.925 --> 00:18:27.376
Will: So I'll, I'll pick, I'll pick two because one is the obvious one you'd

00:18:27.376 --> 00:18:31.535
expect and one is non-obvious, and they're, they're both critical elements.

00:18:31.745 --> 00:18:36.445
So the obvious one is in, uh, is in performance evaluation.

00:18:36.915 --> 00:18:40.545
When you're looking at an existing team and you're trying to get a sense of

00:18:40.905 --> 00:18:45.355
what, what deals have been the moneymakers and who did them, and allowing

00:18:45.356 --> 00:18:50.262
the Originalis platform to basically digest the typical diligence objects

00:18:50.262 --> 00:18:55.182
you'd find in a, in a data room for a, for a firm that's raising a new fund

00:18:55.482 --> 00:18:59.812
allows, allows you to, to basically grind through that question of who's

00:18:59.812 --> 00:19:04.162
generating winners and who's really generating the alpha in this fund manager,

00:19:04.191 --> 00:19:08.691
uh, really efficiently, uh, without having to dispatch an analyst, without

00:19:08.701 --> 00:19:13.502
having to really unpack the cash flows and, uh, do your own set of weightings.

00:19:13.921 --> 00:19:17.142
But the other one, and the one that I, I think is sort of most

00:19:17.142 --> 00:19:22.012
inspiring, is, um, it will score a partnership on culture.

00:19:22.451 --> 00:19:22.741
Raju: Hmm.

00:19:22.802 --> 00:19:24.041
That's super interesting.

00:19:24.162 --> 00:19:29.231
Will: You can not only incorporate references from, from CEOs

00:19:29.232 --> 00:19:32.262
and others, but it'll look at the tenure of the partners.

00:19:32.281 --> 00:19:35.392
It will look at their relative, uh, track records.

00:19:35.622 --> 00:19:38.142
It will look at the roles that they play within the partnership.

00:19:38.162 --> 00:19:40.301
It will look at their own operating experience.

00:19:40.572 --> 00:19:43.171
And it will, of course, look at the question of turnover within the

00:19:43.171 --> 00:19:47.301
firm, uh, but in an intelligent way, one that's orient or- oriented

00:19:47.302 --> 00:19:51.521
towards understanding the, the, the tenure of major contributors

00:19:51.561 --> 00:19:56.181
to the success of the firm versus perhaps the, um, transitory

00:19:56.182 --> 00:20:00.721
activity of more junior or, or, or non-permanent team members.

00:20:00.972 --> 00:20:05.971
So it's quite intelligent on those two factors, one quantitative, one

00:20:05.981 --> 00:20:10.741
obviously qualitative, both critical to thinking about the future, uh,

00:20:10.801 --> 00:20:15.051
of a partnership that you may be making a 10 to 15-year commitment to.

00:20:15.572 --> 00:20:16.331
Raju: Yeah, I love it.

00:20:16.331 --> 00:20:17.892
I want… I kind of want to sing again.

00:20:18.091 --> 00:20:18.701
I'm not going to.

00:20:18.951 --> 00:20:20.632
I'm not going, I'm not going, I'm not going to do it.

00:20:21.021 --> 00:20:21.221
All right.

00:20:21.331 --> 00:20:24.521
I will talk about Farsight, which is one of our portfolio companies.

00:20:24.521 --> 00:20:28.022
It's one that we invested in, and they've been on the podcast.

00:20:28.031 --> 00:20:30.292
So, you know, if you've been a regular

00:20:30.292 --> 00:20:32.761
listener, you've kind of heard about them.

00:20:33.101 --> 00:20:38.972
Um, and their initial focus was really just large M&A organizations,

00:20:39.011 --> 00:20:43.221
mid to large M&A organizations that were sort of separate on their

00:20:43.221 --> 00:20:47.971
own, like, you know, big firms like, um, Deutsche Bank and, and

00:20:47.971 --> 00:20:51.471
smaller ones, you know, uh, as well, so those M&A banks that exist.

00:20:51.901 --> 00:20:56.001
Uh, because the CEO worked for a few years at, you know, General Atlantic

00:20:56.002 --> 00:21:00.041
and a few years at Evercore, and the whole team went to MIT, they really

00:21:00.041 --> 00:21:05.792
constructed this M&A in a box kind of packaging, which was ama- it's amazing.

00:21:05.832 --> 00:21:07.581
It's a really, really amazing tool.

00:21:07.921 --> 00:21:10.521
And for you corporates that are out there, they

00:21:10.521 --> 00:21:14.092
have extended themselves into that dimension.

00:21:14.281 --> 00:21:19.091
So if you're a corporate VC and you invest in fund to funds, and

00:21:19.091 --> 00:21:24.671
you also do direct investing, um, or you diverse div- you know, like

00:21:24.671 --> 00:21:28.922
you basically like are selling entities inside your corporate, which

00:21:28.922 --> 00:21:33.621
happens all the time in big corporates, this is an amazing tool for you.

00:21:34.002 --> 00:21:34.931
It's an amazing tool.

00:21:35.161 --> 00:21:39.361
And this one is, you know- Ready for prime time.

00:21:39.371 --> 00:21:42.331
It, you know, they've got large, large customers.

00:21:42.341 --> 00:21:46.091
I'm, I'm not gonna say who they are because I think there are br- are

00:21:46.141 --> 00:21:50.911
press releases going out shortly, uh, that are, like, mind-blowing.

00:21:50.962 --> 00:21:53.312
Like, they're the top five firms and- Household

00:21:53.312 --> 00:21:53.812
Will: names.

00:21:53.871 --> 00:21:55.041
Raju: Yeah, household names.

00:21:55.042 --> 00:21:55.671
Household names.

00:21:56.141 --> 00:22:00.302
So they're working with, you know, sort of crazy good, uh, folks already.

00:22:00.601 --> 00:22:05.032
But if you're a corporate VC or a corporate LP and

00:22:05.032 --> 00:22:08.221
you invest in GPs and you invest in companies or your

00:22:08.451 --> 00:22:11.901
companies acquires, um, you have to look at this one.

00:22:12.192 --> 00:22:15.552
You have to look at this one, and we can make an intro.

00:22:15.931 --> 00:22:18.431
The demand is so strong in this one, Will.

00:22:18.431 --> 00:22:22.521
Like, we can move the needle obviously 'cause we're on the board and, um,

00:22:22.562 --> 00:22:26.701
get, you know, any of our LPs that are interested in this software kind of

00:22:26.701 --> 00:22:29.681
ahead of the queue or, you know, move them, move them along a little quicker.

00:22:29.701 --> 00:22:30.591
Let's just put it that way.

00:22:30.911 --> 00:22:33.611
Will: I'd love for you to tell our listeners a little bit about

00:22:33.651 --> 00:22:37.071
your perspective on this sector because they're not the first

00:22:37.101 --> 00:22:41.211
company to come along really delivering capabilities to support

00:22:41.222 --> 00:22:46.021
workflow in this area, and yet they're breaking through and, and

00:22:46.032 --> 00:22:49.302
some of their predecessor competitors are actually being ripped out.

00:22:49.361 --> 00:22:52.241
So, so maybe you could help our listeners understand

00:22:52.242 --> 00:22:55.311
why Farsight is winning in such an important area.

00:22:55.471 --> 00:22:55.861
Raju: Yeah.

00:22:55.861 --> 00:22:57.881
This, uh… Thank you so much for asking that

00:22:57.881 --> 00:22:59.901
question 'cause I did not explain that properly.

00:23:00.242 --> 00:23:03.681
So there are other pieces of software out there, including Claude, right?

00:23:03.681 --> 00:23:07.141
Claude will, you know, help you do sort of

00:23:07.141 --> 00:23:10.221
M&A functions, you know, g- provide data.

00:23:10.422 --> 00:23:14.731
You know, one of the most important things in M&A is, uh, production

00:23:14.731 --> 00:23:17.911
of the CIM, which is the confidential information memorandum.

00:23:17.962 --> 00:23:20.492
So this is an example of how they deal with things.

00:23:21.422 --> 00:23:23.051
CIM production takes a long time.

00:23:23.321 --> 00:23:28.861
You know, in most, uh, M&A shops, it, it can be four to six weeks to create

00:23:28.861 --> 00:23:34.701
that CIM because you're taking this giant data room and you have this PowerPoint

00:23:34.751 --> 00:23:39.820
that you typically use as a big firm And you get an army of analysts to

00:23:39.820 --> 00:23:45.439
go scour the data room and populate each one of those slides one by one.

00:23:46.070 --> 00:23:48.850
Farsight, basically you take their, your

00:23:48.870 --> 00:23:51.219
PowerPoint template embedded into Farsight.

00:23:51.290 --> 00:23:56.609
It hits the data room, and I think it's like fifty thousand, a hundred

00:23:56.610 --> 00:24:01.860
thousand simultaneous queries, and minutes later you have a SIM, a draft SIM.

00:24:02.290 --> 00:24:04.999
Let's just say it's ninety percent accurate to start with.

00:24:05.429 --> 00:24:10.350
Um, that is incredible in the grand scheme of things, and if you

00:24:10.350 --> 00:24:14.810
look at the competitor products out there, they cannot do that.

00:24:14.820 --> 00:24:16.099
They are query-based.

00:24:16.109 --> 00:24:19.639
So basically what they do is they give you access to the

00:24:19.639 --> 00:24:23.780
data room to basically query what's inside the data room, and

00:24:23.780 --> 00:24:27.079
you're still effectively building those slides one at a time.

00:24:27.579 --> 00:24:30.769
Um, and Farsight is giving you the end product.

00:24:31.149 --> 00:24:32.979
Um, so that's a big differentiator.

00:24:32.989 --> 00:24:39.969
The second big differentiator is Farsight actually embeds your methodology

00:24:40.429 --> 00:24:44.959
and incorporates it and gets smarter and smarter each SIM you generate.

00:24:45.270 --> 00:24:49.379
So each time you use the product and you've adjusted it slightly and said, "This

00:24:49.379 --> 00:24:54.310
is the workflow that, you know, kinda… This is the story arc that actually

00:24:54.310 --> 00:24:58.439
matters in getting a company sold or getting a company acquired," or, "I've

00:24:58.439 --> 00:25:02.870
gotta pitch this to the board of directors, and so this is the arc I need."

00:25:03.639 --> 00:25:06.349
Every time you use it, it gets better and better and better.

00:25:06.649 --> 00:25:09.389
And so it is different than all of the other

00:25:09.389 --> 00:25:11.760
competitors 'cause it's focused on the end product.

00:25:11.910 --> 00:25:14.249
It's not focused on giving you a query engine.

00:25:14.759 --> 00:25:21.169
And it incorporates a lot of the thinking of the master organization

00:25:21.199 --> 00:25:25.439
into its small language model, which is dedicated to each player.

00:25:25.849 --> 00:25:27.819
Um, and so they are taking share.

00:25:27.889 --> 00:25:29.679
They are beating the competitors.

00:25:29.709 --> 00:25:35.399
They are ripping out the competitors, and anybody who uses them sees why this

00:25:35.410 --> 00:25:41.779
is a much more sophisticated two-year ahead of everybody else kind of product.

00:25:42.199 --> 00:25:45.189
Um, and you know, like, I'm very proud of it.

00:25:45.189 --> 00:25:46.810
I know that Ari is very proud of it.

00:25:47.299 --> 00:25:50.199
Um, these guys are, are a special, special set of

00:25:50.200 --> 00:25:52.589
founders that have done something pretty remarkable

00:25:53.710 --> 00:25:54.380
Will: Indeed they are.

00:25:54.640 --> 00:25:54.980
Raju: Yeah.

00:25:55.260 --> 00:25:56.460
Well, thank you for that opportunity.

00:25:56.460 --> 00:26:01.060
So, so kind of like let me recap a little bit for all LPs.

00:26:01.090 --> 00:26:02.030
You know, what should you do?

00:26:02.059 --> 00:26:04.640
Well, let's start playing with the tools.

00:26:04.650 --> 00:26:08.970
Let's get some training, maybe bring in some talent, either in the form of

00:26:08.990 --> 00:26:13.639
consultants or analysts with AI skill sets, and start looking at some tools that

00:26:13.640 --> 00:26:20.180
are dedicated to your function, whether it's Originalis or Farsight or both.

00:26:20.569 --> 00:26:24.900
Um, I think those are really excellent tools, but you got to get

00:26:24.900 --> 00:26:29.680
over the hump even though, like, you know, this isn't a world

00:26:29.680 --> 00:26:34.450
where you can wait forever to like… for things to mature.

00:26:34.460 --> 00:26:36.279
You just got to start, you know, getting.

00:26:36.699 --> 00:26:40.150
For corporates, and we have bunch of them, right?

00:26:40.190 --> 00:26:43.500
And, you know, as I said, we have a program where, um,

00:26:43.530 --> 00:26:45.579
I think it's worth mentioning the program a little bit.

00:26:45.579 --> 00:26:51.759
Like for corporates, we actually will spend, you know, a couple days every

00:26:52.750 --> 00:26:58.040
Six months or so, um, more w- walking them through technology shifts and

00:26:58.040 --> 00:27:02.500
kind of what's happening, um, or if the time is better spent for them to

00:27:02.500 --> 00:27:07.290
sort of understand how AI is gonna impact their business, we'll do that.

00:27:07.389 --> 00:27:08.030
We'll spend the time.

00:27:08.290 --> 00:27:12.449
Um, it's a very, uh, mutually beneficial re- relationship.

00:27:12.730 --> 00:27:16.930
Um, we love for our corporates to give us feedback in terms of what

00:27:16.939 --> 00:27:19.610
they're seeing and to take a look at some of the products that we're

00:27:19.610 --> 00:27:23.269
thinking about investing in and see if it would be a fit for them

00:27:23.269 --> 00:27:26.670
and what their feedback is before we make that decisioning loop.

00:27:26.779 --> 00:27:29.970
But, like, one of the things we've done is gone in and talked to them about

00:27:29.970 --> 00:27:35.099
like, "Hey, you're a manufacturer. You're a giant fintech. You know, you are

00:27:35.099 --> 00:27:40.480
a healthcare organization, and you are trying to sit there and say, 'How, how

00:27:40.480 --> 00:27:45.699
is AI gonna change my business?'" And, you know, g- and, and that's beyond

00:27:45.699 --> 00:27:51.579
just, like, the corporate, uh, venture capital slash, you know, LP arm.

00:27:51.809 --> 00:27:54.319
That's like, "What's it gonna do to my business? What's it gonna do

00:27:54.320 --> 00:27:57.540
to my business?" And, you know, we have a methodology here, right?

00:27:57.540 --> 00:28:03.089
So in the ground floor stuff, and I don't care what company

00:28:03.089 --> 00:28:07.499
you are, you have to be doing right now, right now, you

00:28:07.500 --> 00:28:11.069
have to get some AI fluency in your business, you know.

00:28:11.419 --> 00:28:16.639
And, uh, you know, we talked to the corporate VC pieces, but every part of your

00:28:16.639 --> 00:28:20.940
business has got to get some AI fluency, and it has to be directed from the top.

00:28:21.549 --> 00:28:25.409
The, the leadership has to embrace the fact that AI is going to be,

00:28:25.650 --> 00:28:27.899
whether it's now or 10 years from now, it doesn't matter, right?

00:28:27.919 --> 00:28:29.469
Like, you gotta get some AI fluency.

00:28:30.099 --> 00:28:33.299
The second thing is use case identification.

00:28:33.750 --> 00:28:38.040
What are the use cases that actually can make a difference in your business?

00:28:38.519 --> 00:28:44.199
Um, and everybody kind of looks at customer service as a potential use case.

00:28:44.699 --> 00:28:48.410
They look at knowledge management inside the business as a use case.

00:28:48.410 --> 00:28:52.319
They look at basically sales and marketing as a use case.

00:28:52.839 --> 00:28:57.019
Um, but you have to identify the ones that are most meaningful, and

00:28:57.029 --> 00:29:01.140
that requires each of your leaders in your business to sit there

00:29:01.140 --> 00:29:03.809
and say, like, "What's the one thing that would make a difference?"

00:29:03.839 --> 00:29:07.280
And then you can sort of look at it across the business and say, "Of the

00:29:07.280 --> 00:29:11.019
five or six or seven that bubble to the top, which two or three really

00:29:11.020 --> 00:29:14.250
are the ones that we're gonna focus in on?" So that's really important.

00:29:14.779 --> 00:29:22.200
The third, and this is really important, is a dashboard that monitors the

00:29:22.209 --> 00:29:28.735
core AI metrics And if I were to synopsize our last couple of podcasts

00:29:28.976 --> 00:29:34.586
around AI, we've talked about all of these to some degree, like token cost.

00:29:35.335 --> 00:29:42.285
So the reason token cost should be on your dashboard is because people make the

00:29:42.286 --> 00:29:47.576
mistake of thinking that today's token costs are gonna be tomorrow's token cost.

00:29:47.985 --> 00:29:52.656
And, you know, we all know that a lot of the LLMs were subsidizing those tokens,

00:29:53.016 --> 00:29:59.065
and now they're getting more toward, like, 50% subsidization or full price.

00:29:59.585 --> 00:30:03.115
But you gotta monitor the cost because any ROI that you're gonna build

00:30:03.175 --> 00:30:07.705
around your use cases is gonna be dependent on the cost of those tokens.

00:30:07.976 --> 00:30:09.705
So that should be on your dashboard.

00:30:10.186 --> 00:30:13.536
There are a bunch of use cases, Will, and

00:30:13.536 --> 00:30:16.016
you know this, that are real-time in nature.

00:30:16.675 --> 00:30:17.195
They're real-time.

00:30:17.575 --> 00:30:23.055
So that means that the AI has got to think and respond

00:30:23.575 --> 00:30:27.844
in real time, so latency matters In those use cases.

00:30:27.924 --> 00:30:34.544
Uh, because in some instances, you don't want the customer or the prospect

00:30:34.544 --> 00:30:39.124
or the, you know, whoever, business partner to know that they're actually

00:30:39.134 --> 00:30:43.904
chatting with an AI, um, or they're talking with an AI, so latency matters.

00:30:44.154 --> 00:30:46.454
Security should be on your dashboard, right?

00:30:46.464 --> 00:30:51.213
Like, if we have still holes in a bunch of, you know, stuff,

00:30:51.234 --> 00:30:55.123
you may decide that you need that AI in-house versus being able

00:30:55.124 --> 00:30:58.653
to go out and, you know, have a round trip in an outside world.

00:30:59.513 --> 00:31:01.563
Answer and quality satisfaction.

00:31:02.013 --> 00:31:04.024
Um, this is super important, right?

00:31:04.033 --> 00:31:07.013
Like, I mean, when's the last time you called your wireless carrier?

00:31:09.153 --> 00:31:10.023
Oh my God.

00:31:10.964 --> 00:31:13.583
I, if we could just solve this one issue, just like, I

00:31:13.624 --> 00:31:16.333
need to get to the right agent, right person to talk to.

00:31:16.383 --> 00:31:19.713
Oh my God, I… It's like an hour every time.

00:31:20.214 --> 00:31:22.123
Will: But as a B- as a former Bell Labs guy,

00:31:22.124 --> 00:31:24.123
it, it has special meaning for you, right?

00:31:24.124 --> 00:31:26.243
You, you, you get kind of a warm feeling when you

00:31:26.243 --> 00:31:26.413
Raju: do it, right?

00:31:26.413 --> 00:31:26.943
Yeah, I do.

00:31:26.973 --> 00:31:27.334
I do.

00:31:27.334 --> 00:31:31.493
When I, when I call my carrier and I wait an hour on the phone, I'm

00:31:31.493 --> 00:31:36.573
just like, "Yeah, man, this is what it's all about." I feel good.

00:31:36.574 --> 00:31:39.653
It's like it's a war… No, it's just so painful.

00:31:39.663 --> 00:31:43.093
Like, um, if AI can do one thing, it's get

00:31:43.103 --> 00:31:45.333
me to the person that I need to talk to.

00:31:45.913 --> 00:31:46.293
Will: Yes.

00:31:46.663 --> 00:31:48.683
Raju: Even if it doesn't give me the answer on

00:31:48.684 --> 00:31:50.444
how to solve it, I just need to get to the person.

00:31:50.553 --> 00:31:52.753
You know, that's like step one in all of this.

00:31:53.214 --> 00:31:55.323
Uh, but answer quality, satisfaction,

00:31:55.333 --> 00:31:58.484
hallucination rate, context window, all important.

00:31:59.203 --> 00:32:02.923
Token cost, latency, security, answer quality

00:32:02.923 --> 00:32:05.674
satisfaction, hallucination rate, context window.

00:32:06.223 --> 00:32:11.674
Those five or six need to be in your dashboard, and you need to monitor them.

00:32:12.394 --> 00:32:14.653
And so let's say you've done this.

00:32:14.653 --> 00:32:18.793
You got some AI fluency, you d- identified some use cases, and

00:32:18.793 --> 00:32:24.043
now you have some AI metrics, and you know that the ROI for

00:32:24.043 --> 00:32:29.833
your use case makes sense when token costs are X price and when

00:32:29.833 --> 00:32:34.703
latency is Y and then when answer and quality satisfaction is Z.

00:32:35.323 --> 00:32:40.924
Um, then the ROI makes sense, and you can go forward with a full

00:32:40.924 --> 00:32:44.833
deployment and move out of pilot, which is a- what a lot of people in

00:32:44.903 --> 00:32:47.883
corporate land are struggling with, is how to move out a pi- pilot.

00:32:47.883 --> 00:32:52.274
I, I, I think that's like sort of ground floor for all the corporates.

00:32:52.953 --> 00:32:58.963
The items that I think corporations by and large need to consider

00:32:59.023 --> 00:33:06.463
today absolutely leveraging are effectively three areas: sales

00:33:06.463 --> 00:33:10.629
and marketing customer service, and software development.

00:33:11.180 --> 00:33:13.590
Would you agree with that, Will, or you think there's another one out there?

00:33:13.590 --> 00:33:13.620
100%. Yeah.

00:33:13.620 --> 00:33:14.080
Will: 100%.

00:33:14.080 --> 00:33:16.330
There, there are quick wins in all three of those areas.

00:33:16.570 --> 00:33:20.160
Raju: And I'm not talking about taking over the sales and marketing, um, with

00:33:20.160 --> 00:33:26.200
an AI, but facilitating, supplementing your sales and marketing organization

00:33:26.210 --> 00:33:31.680
with AI tools, supplementing your customer service w- uh, with, with,

00:33:31.680 --> 00:33:35.539
you know, some tool sets and, you know, speeding up software development.

00:33:35.909 --> 00:33:38.409
Um, I actually think software development's furthest along,

00:33:38.500 --> 00:33:43.199
right, in terms of its ability to create value out of the gate.

00:33:43.639 --> 00:33:46.180
Um, you know, you still need your senior developers to

00:33:46.180 --> 00:33:49.339
sort of authenticate and make, you know, sort of like,

00:33:49.750 --> 00:33:53.159
uh, get people comfortable, but it is furthest along.

00:33:53.599 --> 00:33:56.269
Uh, I, I think sales and marketing is an interesting one.

00:33:56.659 --> 00:34:01.900
It isn't dissimilar to the ones we talked about for corporate venture or

00:34:01.900 --> 00:34:06.499
for limited partners, um, where we talked about Originalis and Farsight.

00:34:06.959 --> 00:34:10.119
There are tools for sales and marketing today that

00:34:10.469 --> 00:34:16.020
can help you find prospects better and faster.

00:34:16.409 --> 00:34:22.009
They can tell you when your prospect is contract

00:34:22.009 --> 00:34:23.970
is up for renewal with one of your competitors.

00:34:24.359 --> 00:34:26.659
This is unbelievably valuable.

00:34:26.949 --> 00:34:27.470
Think about that.

00:34:28.139 --> 00:34:31.829
You're trying to sell to Will Porteous, and

00:34:31.829 --> 00:34:33.770
Will Porteous is using somebody else's product.

00:34:33.770 --> 00:34:36.030
You've called him up, and Will's like, "Ah, I'm already using

00:34:36.039 --> 00:34:38.499
blah, blah, blah, blah," and you're still trying to sell to Will

00:34:38.509 --> 00:34:41.980
Porteous, and you're like, if you knew when that contract was up

00:34:41.980 --> 00:34:47.929
for renewal, you would ping Will three months before that and say

00:34:50.012 --> 00:34:52.942
"Hey, Will, I know you mentioned you were on this other solution set.

00:34:52.962 --> 00:34:55.391
Well, you know, we've heard in the marketplace it has a gap.

00:34:55.412 --> 00:34:56.722
This is what we do better.

00:34:57.422 --> 00:34:59.611
When's your contract up for renewal?" But you know when the contract's

00:34:59.612 --> 00:35:03.292
up for renewal, but Will, Will is now an interested party, right?

00:35:03.302 --> 00:35:05.982
Now he's ready to listen 'cause, you know,

00:35:05.982 --> 00:35:08.111
six months ago wasn't ready to listen.

00:35:08.111 --> 00:35:10.951
I'm- I got n- nine more months with this, with this player.

00:35:11.451 --> 00:35:14.981
So there are a bunch of sales and marketing tools that help you

00:35:15.332 --> 00:35:19.201
generate drafts of content, generate drafts of outbound emails,

00:35:19.441 --> 00:35:26.641
identify who the best prospects are for you, basically draft emails

00:35:26.722 --> 00:35:32.171
that compare you to that competitor and why you need to be displaced.

00:35:32.562 --> 00:35:35.861
Um, I, I, I will give you one of those just because we're at…

00:35:35.891 --> 00:35:37.732
You know, selfishly it's of our interest.

00:35:38.071 --> 00:35:40.181
Like, we ha- we're investors in a company called Avena,

00:35:40.591 --> 00:35:45.002
and Avena is a sales tool that absolutely does this.

00:35:45.391 --> 00:35:48.161
It tells you, you tell it, "Look, these are the type of

00:35:48.161 --> 00:35:51.541
customers we're looking for." It identifies who they are, who

00:35:51.541 --> 00:35:54.832
the, what the emails are, what the, you know, sort of social

00:35:54.832 --> 00:35:58.931
media links are, the LinkedIn on each of the potential buyers.

00:35:59.472 --> 00:36:00.952
It will draft emails.

00:36:01.021 --> 00:36:04.221
It can even tell you when their contracts are up for

00:36:04.222 --> 00:36:08.331
renewal with your competitors, and you should be using this.

00:36:08.361 --> 00:36:13.921
If you are a manufacturer, if you are a corporate VC and your

00:36:14.091 --> 00:36:17.261
parent company needs this, you should be leveraging this.

00:36:17.611 --> 00:36:20.451
I know from personal feedback, we've introduced Avena to a bunch

00:36:20.451 --> 00:36:24.651
of corporates, and they are getting a ton of value out of it.

00:36:25.052 --> 00:36:30.891
Um, customer service is one where y- I, do, I don't know.

00:36:31.192 --> 00:36:33.531
Like, w- ha- have you called customer

00:36:33.531 --> 00:36:35.511
service and talked to an AI chatbot recently?

00:36:36.692 --> 00:36:37.362
Will: Yeah.

00:36:37.421 --> 00:36:39.811
It's getting better, but there's a long way to go.

00:36:40.101 --> 00:36:41.872
Raju: My experience, I don't know if yours

00:36:41.872 --> 00:36:45.674
are the same It just can do the simple stuff.

00:36:46.114 --> 00:36:46.614
Will: Right.

00:36:46.904 --> 00:36:49.864
Raju: And even IVRs used to do the simple stuff, right?

00:36:49.864 --> 00:36:51.953
You could check your account balance, you can find

00:36:51.954 --> 00:36:54.134
out what, when the order is gonna ship, you know?

00:36:54.134 --> 00:36:56.964
But anything that was complicated, it would just

00:36:56.994 --> 00:37:01.203
like kinda barf, and it still a little bit does that.

00:37:01.444 --> 00:37:04.654
Will: Well, you know, Raju, I, I think it probably because a lot of

00:37:04.654 --> 00:37:10.074
this AI has been, been grafted onto process-driven architectures of

00:37:10.074 --> 00:37:14.023
the past, and, and what's actually needed is a fundamental rebuild.

00:37:14.024 --> 00:37:18.274
I mean, think about all the millions of consulting hours that

00:37:18.274 --> 00:37:22.154
were invested in defining all of those customer service processes

00:37:22.154 --> 00:37:26.164
and, and literally mapping out all of those prompt windows so that

00:37:26.164 --> 00:37:30.253
every process could be tightly defined and tightly responded to.

00:37:30.253 --> 00:37:34.424
And, you know, the AI era is, is about freeing us from all of

00:37:34.424 --> 00:37:38.264
that kind of structured responsibility on the part of customers.

00:37:38.683 --> 00:37:43.243
Um, it's probably gonna take a, a, a literally a, a what you and I

00:37:43.243 --> 00:37:48.934
would call a forklift upgrade from, from, from our enterprise selling

00:37:48.934 --> 00:37:53.893
past, um, for tho- for those big enterprises, and they're… Yeah.

00:37:54.343 --> 00:37:54.673
Raju: Yeah.

00:37:55.004 --> 00:37:57.144
Yeah, I, I, I think there's an intermediate step.

00:37:57.153 --> 00:37:59.323
It's not getting the answer but getting the right person.

00:37:59.574 --> 00:38:00.673
This is what I talked about earlier.

00:38:00.953 --> 00:38:04.873
I, I, I really think that that is what people should be embracing.

00:38:05.523 --> 00:38:13.694
How do I get Not the answer, but to the right person faster and without… And

00:38:13.694 --> 00:38:19.334
so I should be the, the AI should be able to understand the problem that I'm

00:38:19.334 --> 00:38:24.183
having and who the individual is inside the organization that can solve it.

00:38:24.764 --> 00:38:27.634
And sometimes you go into billing and you really, it's

00:38:27.634 --> 00:38:29.913
a technical problem, but you're in billing, right?

00:38:29.924 --> 00:38:33.804
Or you're in the technical workflow, but really it's a billing question.

00:38:33.823 --> 00:38:36.754
Or, you know, you're, you're, you're trying to figure out how

00:38:36.754 --> 00:38:39.814
to cancel and you can't get the right person, and maybe they

00:38:39.814 --> 00:38:43.294
do that intentionally, Not, not getting you the right person.

00:38:43.924 --> 00:38:46.353
But, you know, there, there's a bunch out there.

00:38:46.403 --> 00:38:48.904
Uh, you know, I think voice agents are getting better.

00:38:49.174 --> 00:38:52.404
If you're gonna do voice agents, by the way, a- anybody who's listening,

00:38:52.813 --> 00:38:58.423
um, you should talk to our, um, company, VoiceRun, because they…

00:38:58.433 --> 00:39:00.983
You know, I've, we've talked about this in, in podcasts before.

00:39:01.424 --> 00:39:05.084
You know, doing it alone and trying to figure out all the different

00:39:05.093 --> 00:39:10.074
voice AI engines like natural language processing, how do I answer

00:39:10.074 --> 00:39:14.064
the question, actual voice creation, all of them are different LLMs.

00:39:14.403 --> 00:39:16.834
Trying to do that on your own without a framework and

00:39:16.834 --> 00:39:19.454
a forward deployed engineer is a route to failure.

00:39:19.463 --> 00:39:23.904
These guys help you with all of that, um, and that's why they're

00:39:24.024 --> 00:39:27.714
getting, you know, over the goal line with a bunch of companies.

00:39:28.183 --> 00:39:31.124
But I think that, you know, that's the nirvana of it all.

00:39:31.533 --> 00:39:34.253
Knowledge retrieval is an interesting one in corporates.

00:39:34.993 --> 00:39:39.173
I wish… You know, like I, every single company I've ever worked with and for,

00:39:39.433 --> 00:39:45.213
um, frankly, there's always been this thing where you reinvent the wheel a bit.

00:39:45.613 --> 00:39:49.263
Um, y- you know, the bigger the company, the more people that have had

00:39:49.273 --> 00:39:51.953
to solve that problem before, but you just don't know who they are.

00:39:52.433 --> 00:39:54.343
And you remember the, back in the era of this

00:39:54.353 --> 00:39:57.634
knowledge management, um, you know, CRMs?

00:39:58.153 --> 00:39:59.054
Will: Oh, completely.

00:39:59.063 --> 00:40:02.953
All, all, you know, all, all, all the, all the best approaches, all

00:40:02.953 --> 00:40:07.173
the best expertise was gonna be centrally collected and available

00:40:07.173 --> 00:40:11.133
in a pro- repository that anyone could access and train on, and-

00:40:11.264 --> 00:40:11.554
Raju: Yeah.

00:40:11.713 --> 00:40:12.883
You remember intranets?

00:40:13.243 --> 00:40:13.773
Will: Mm-hmm.

00:40:13.823 --> 00:40:14.133
Yeah.

00:40:14.304 --> 00:40:15.643
Raju: It wasn't, wasn't an internet.

00:40:15.653 --> 00:40:15.703
Inside.

00:40:15.714 --> 00:40:19.373
It was just like, yeah, it was like an internet for inside.

00:40:19.373 --> 00:40:21.344
Will: And, and all, all of corporate life could

00:40:21.354 --> 00:40:23.843
be found somewhere posted on the intranet if you-

00:40:23.884 --> 00:40:24.584
Raju: Exactly … if you needed to.

00:40:24.584 --> 00:40:24.943
Exactly.

00:40:25.134 --> 00:40:25.734
Exactly.

00:40:26.063 --> 00:40:28.453
Um, this problem has been knowledge management,

00:40:28.473 --> 00:40:31.213
knowledge databases, intranets have been tried before.

00:40:31.214 --> 00:40:35.053
I think AI is gonna, um, really make this better.

00:40:35.384 --> 00:40:38.373
Uh, it, I, I think this is an opportunity, and it doesn't

00:40:38.373 --> 00:40:41.504
have to be perfect, and that's why I think it's okay to

00:40:41.504 --> 00:40:46.326
start Using AI for this purpose inside of corporations.

00:40:46.896 --> 00:40:50.366
And, uh, and we, we said this statistic before, one of the things

00:40:50.366 --> 00:40:56.896
that's really crazy to me is that 85% of agentic queries are repeats.

00:40:57.566 --> 00:40:57.936
Will: Yeah.

00:40:58.495 --> 00:41:02.476
Uh, Raju, I think this is a really good insight on your part because so much

00:41:02.476 --> 00:41:06.806
of, so much of the problem around accessing that information historically has

00:41:06.806 --> 00:41:11.215
been around indexing and searching for it and having well-structured taxonomies.

00:41:11.256 --> 00:41:14.125
And if AI… Uh, that's, that's the kind of problem that

00:41:14.125 --> 00:41:18.535
AI's quite good at, uh, synthesis and, and resolution.

00:41:18.926 --> 00:41:21.956
And you could… Y- y- y- we c- we can frankly finally

00:41:21.956 --> 00:41:25.326
start to unlock the value from these old, old corpuses.

00:41:25.536 --> 00:41:28.165
I don't know if, I don't know if the, the knowledge is still

00:41:28.165 --> 00:41:31.185
valuable in all those knowledge management systems, but at

00:41:31.185 --> 00:41:33.665
least now we'll, we'll know how to get it when we need it.

00:41:34.036 --> 00:41:36.175
Raju: I think, you know, by… You, you'd be surprised at how

00:41:36.176 --> 00:41:40.298
valuable it is There's probably like a dude or a gal someplace in the

00:41:40.298 --> 00:41:43.348
company and she's the only one that knows how to fix this, you know?

00:41:43.708 --> 00:41:45.928
And it's like, wow, like, you know, it takes you like a week

00:41:45.958 --> 00:41:48.567
to get it, and meanwhile the manufacturing tool is down for

00:41:48.567 --> 00:41:51.538
a week or, you know, you haven't really addressed an issue.

00:41:51.998 --> 00:41:56.208
Um, so I… There's so much more that we could talk about, like, you know,

00:41:56.258 --> 00:41:59.767
how to… What are the things within, you know, software development?

00:41:59.817 --> 00:42:01.007
What business development?

00:42:01.018 --> 00:42:04.757
Manufacturing should be using vision systems, R&D, data analysis.

00:42:05.008 --> 00:42:07.997
There's all these sectors, lending, finance.

00:42:07.998 --> 00:42:11.448
Like, what tools can we apply from an AI perspective for corporates?

00:42:11.827 --> 00:42:13.907
I don't, I don't think we have time to go into those.

00:42:13.917 --> 00:42:18.058
I think we wa- I wa- what was important to get through, at

00:42:18.058 --> 00:42:22.877
least at the ground level, like if you're a LP that is a family

00:42:22.878 --> 00:42:26.277
office or an institution, you know, here's the starting point.

00:42:26.288 --> 00:42:29.027
Here are some tools out there that you should start experimenting with.

00:42:29.387 --> 00:42:31.708
And if you're a corporate, you know, everybody's got sales,

00:42:31.708 --> 00:42:34.618
marketing, software development, you know, those kind of thing.

00:42:34.967 --> 00:42:37.008
Here's the tools that you should be considering, and

00:42:37.018 --> 00:42:38.937
here's the dashboards that you should be following.

00:42:39.498 --> 00:42:43.428
You know, with our core limited partners, the ones that, you know, sort

00:42:43.428 --> 00:42:46.758
of have been with us for a while, we'll, we'll, we'll spend the time.

00:42:46.758 --> 00:42:49.517
And if they're on the call, you know, or are on this, um,

00:42:49.557 --> 00:42:52.507
podcast listening, we're happy to spend some cycles talking

00:42:52.508 --> 00:42:55.857
to you about, um, you know, what you should be playing with.

00:42:56.197 --> 00:42:58.187
And for the new ones out there, you know, understand

00:42:58.187 --> 00:43:01.107
this is a value proposition that RRE takes seriously.

00:43:01.328 --> 00:43:05.117
Um, that relationship with our limited partners is important.

00:43:05.457 --> 00:43:09.038
And there's no guy inside of RRE that spends more time

00:43:09.038 --> 00:43:12.127
with limited partners than my partner, Will Porteous here.

00:43:13.017 --> 00:43:14.558
Will: Well, this is, this is the kind of

00:43:14.567 --> 00:43:17.597
area where we, we, we, we wanna be helpful.

00:43:17.607 --> 00:43:19.957
We're, we're grateful for the long support we have

00:43:19.957 --> 00:43:22.587
so, from so many great institutions and investors.

00:43:22.607 --> 00:43:25.907
And it is, it is a critical moment in which everyone's

00:43:26.178 --> 00:43:29.868
organizations are being changed by these capabilities.

00:43:29.928 --> 00:43:35.027
And, uh, we wanna share what we use and what we think can benefit our

00:43:35.027 --> 00:43:39.658
partners, uh, so that everybody can get some time back or be more efficient.

00:43:40.047 --> 00:43:40.747
Raju: Yeah, that's amazing.

00:43:40.748 --> 00:43:42.097
I'm gonna let you close this one out.

00:43:42.097 --> 00:43:44.167
And if you wanna do song, you're, you know- … you

00:43:44.167 --> 00:43:46.837
can absolutely close it out with a song.

00:43:46.978 --> 00:43:50.517
Will: You know, I, I think, I think you, you set a very high

00:43:50.517 --> 00:43:53.198
standard earlier, so I'm not gonna sing to our audience.

00:43:53.368 --> 00:43:57.527
But I am gonna thank our audience for their loyalty, for being

00:43:57.527 --> 00:44:01.917
with us week after week, for sharing the RRE POV episodes.

00:44:01.917 --> 00:44:05.277
If you think we've, we've presented something in an interesting

00:44:05.277 --> 00:44:08.347
way, if you got something out of these episodes, we'd be grateful

00:44:08.347 --> 00:44:11.747
if you'd pass it along and share it with people in your network.

00:44:12.017 --> 00:44:15.187
And, um, and if there's something we've said that

00:44:15.187 --> 00:44:17.187
you disagree with, we'd love to hear from you too.

00:44:17.197 --> 00:44:18.218
We're easy to reach.

00:44:18.277 --> 00:44:22.037
Uh, so thank you as always, and we look forward

00:44:22.047 --> 00:44:24.847
to sh- having you with us again before long.

00:44:25.288 --> 00:44:25.648
Raju: Yeah.

00:44:25.648 --> 00:44:27.867
And Will, you and I are gonna do karaoke now.

00:44:27.907 --> 00:44:28.937
I think we should do karaoke.

00:44:29.338 --> 00:44:29.658
Okay.

00:44:30.417 --> 00:44:31.017
Thank you all.

00:44:31.247 --> 00:44:31.957
Will: Thank you, everybody.