WEBVTT
00:00:00.080 --> 00:00:05.759
I would say from a VC perspective, there's a trade-off between market size and differentiation.
00:00:06.320 --> 00:00:24.559
And you can have all the differentiation you want, but as long as your SAM if your SAM is less than a billion dollars, it becomes very hard from a VC perspective to be able to invest in those companies because those venture fundable returns aren't necessarily there.
00:00:31.920 --> 00:00:38.880
Welcome to Partner Path, a podcast that unpacks the venture capital and growth equity ecosystem from a junior perspective.
00:00:39.039 --> 00:00:45.359
Young entrepreneurs and investors have already had a massive impact on the industry, having started unicorns and launched billion-dollar funds.
00:00:45.759 --> 00:00:52.079
We discuss these success stories and more by sharing perspectives and advice from some of the industry's most prominent role models.
00:00:52.240 --> 00:00:58.640
This podcast expresses our views as of the date published and does not represent the views of and is not endorsed by any company for which we work.
00:01:05.280 --> 00:01:09.760
Manu started his career at Dallas VC before heading to business school in Chicago.
00:01:09.920 --> 00:01:19.519
While in business school, Manu did early stage investing at B Capital, then landed at TVP, an early stage VC fund focused on investing in startups operating in highly regulated industries.
00:01:19.680 --> 00:01:20.560
Welcome, Manu.
00:01:21.040 --> 00:01:21.680
Thank you, Laura.
00:01:21.760 --> 00:01:22.319
Thank you, Will.
00:01:22.400 --> 00:01:23.120
Thanks for having me.
00:01:23.359 --> 00:01:28.319
To start, could you walk us through your background and what led you to come to the US, specifically Texas?
00:01:28.640 --> 00:01:29.680
Yeah, absolutely.
00:01:29.920 --> 00:01:35.120
So I was born in India, but my dad worked for IBM all throughout the 90s.
00:01:35.280 --> 00:01:40.480
And so I lived in a bunch of different areas and had a very interesting childhood growing up.
00:01:40.719 --> 00:01:46.640
I lived in Thailand, Singapore, UK, and the UAE before the age of seven.
00:01:46.799 --> 00:01:50.239
And then ended up moving to Texas when I was eight.
00:01:50.400 --> 00:01:56.560
So it was a lot of hopping, and then all of a sudden in one place for 15, 20 years.
00:01:56.959 --> 00:02:07.599
And you started your VC career in Texas too, specifically working for a solo GP and building the fund essentially from the ground up, which I can imagine was very challenging.
00:02:07.760 --> 00:02:13.759
So it would be great if you could touch on kind of your role in building Dallas VC and what you learned from that experience.
00:02:14.080 --> 00:02:15.120
Yeah, absolutely.
00:02:15.280 --> 00:02:20.960
Uh was very, very lucky to have that opportunity and definitely walked away with a tons of learning.
00:02:21.120 --> 00:02:31.599
So I sort of joined as the first non-partner employee, almost you could think of it at that time as a solo GP uh fund that was sort of nearing the end of its first fund.
00:02:32.000 --> 00:02:35.199
And my first job was to really just help out with the portfolio.
00:02:35.280 --> 00:02:42.400
And I stepped into a bunch of different roles and did anything from PM or BizDev and just learned a lot.
00:02:42.560 --> 00:02:49.120
And once I did that for about a couple months, I sort of transitioned to doing a little more of the VC kind of roles.
00:02:49.280 --> 00:02:52.400
And that started off with sourcing, and then over time did diligence.
00:02:52.479 --> 00:03:06.719
And as the fund performed and we sort of moved into the new fund, I was starting to get some portfolio responsibilities and then eventually over time started doing my own deals, which was awesome to be able to do at such a young age.
00:03:06.960 --> 00:03:09.439
I actually got a few small exits under my belt.
00:03:09.680 --> 00:03:17.439
I also helped manage a lot of the fund operations and did a ton of LP relations and I actually got to do a little bit of fundraising as well.
00:03:17.599 --> 00:03:19.039
So it was a ton of learning.
00:03:19.199 --> 00:03:27.680
We went from two people managing about$15 million to about 12 people managing close to 200 by the time I left, two geographies.
00:03:27.759 --> 00:03:29.439
We also launched a fund in India.
00:03:29.680 --> 00:03:36.560
So it was a tremendous experience and definitely walked away with a bunch of stuff that were really important learnings.
00:03:36.879 --> 00:03:41.599
Another interesting part of your path back to VC is choosing to go to business school.
00:03:41.759 --> 00:03:49.840
I think nowadays, you know, we're at the point at that age 24, 25, people debating to studying and going back, especially when you're already in VC.
00:03:50.080 --> 00:03:54.639
I think some would argue that going back sometimes makes it harder to get back into VC.
00:03:54.800 --> 00:03:59.759
So I'm curious, when you decided to go to booth, was the plan to come back into VC?
00:04:00.240 --> 00:04:02.319
Yeah, the plan was definitely to come back.
00:04:02.560 --> 00:04:10.240
Business school was primarily because A, I felt that being in Dallas, it's a great ecosystem.
00:04:10.319 --> 00:04:15.360
You have a ton of corporate enterprises and you have a really good research institutions and things like that.
00:04:15.520 --> 00:04:22.319
But at the same time, you sort of don't necessarily have the critical mass that a New York or San Francisco will have.
00:04:22.560 --> 00:04:26.079
And so for me, I really wanted to be in a bigger ecosystem.
00:04:26.160 --> 00:04:40.000
And I felt that going to business school would be sort of a great way to be able to get exposure to those ecosystems, but then also just take a step back from five years of drinking from the fire hose and get a wider perspective in the world of business.
00:04:40.240 --> 00:04:41.680
Booth for me was really interesting.
00:04:41.839 --> 00:04:44.480
It has a very systematic approach to decision making.
00:04:44.639 --> 00:04:48.480
And I think that is really helpful for becoming a disciplined investor.
00:04:48.720 --> 00:04:51.040
But for me, the plan was definitely always to come back.
00:04:51.279 --> 00:04:51.600
Got it.
00:04:51.680 --> 00:04:58.399
And I'm sure the network as well that you've built at business school will only be helpful in VC and whatever you choose to do next.
00:04:58.639 --> 00:04:59.839
Absolutely.
00:05:00.160 --> 00:05:06.079
Would you say that does going to business school make go getting into venture capital easier?
00:05:06.240 --> 00:05:12.240
Let's say whether you come from a non-traditional background and you're looking to switch to that investing role?
00:05:12.639 --> 00:05:19.759
I think it depends on what background you're really coming from and where the world of venture capital you want to target.
00:05:20.000 --> 00:05:23.120
I think for on the whole, it's much easier than coming from undergrad.
00:05:23.680 --> 00:05:28.959
But there are going to be target types that venture funds will primarily look for.
00:05:29.120 --> 00:05:37.920
And that's going to be someone who's either been a prior investor or an industry specialist or an operator or product manager at a fast-growing early stage startup.
00:05:38.160 --> 00:05:43.439
But for people that are not coming from those backgrounds that want to transition into venture, it's also very doable.
00:05:43.600 --> 00:05:47.839
Business school is one of those things where it's a matter of you get out what you put in.
00:05:47.920 --> 00:05:54.319
And so if you're the kind of person that seeks opportunities, seeks internships, it's not that difficult to get a full-time offer.
00:05:54.480 --> 00:06:04.240
Obviously, there are other macro factors like the market, but on the whole, people, at least what I've seen in booths, people that try and really shoot for VC as a role end up getting jobs in VC.
00:06:04.480 --> 00:06:05.439
Yeah, that's a good point.
00:06:05.519 --> 00:06:08.800
And you can also start working too while you're a first or second year.
00:06:08.879 --> 00:06:11.920
So getting that VC experience part-time for a fund.
00:06:12.160 --> 00:06:12.720
Exactly.
00:06:13.040 --> 00:06:16.240
You decide go back into VC, start recruiting.
00:06:16.399 --> 00:06:23.519
What were you looking for in a VC fund, maybe just to broaden your skill set as you were going back into the recruiting landscape?
00:06:23.920 --> 00:06:39.759
One of the things I'd learned at Dallas and some of the other firms I'd interned at was that having a really differentiated strategy is really successful to being a venture fund that goes through multiple funds and is really something that's established.
00:06:39.920 --> 00:06:43.279
And so I looked for, above all, a differentiated strategy.
00:06:43.439 --> 00:06:47.120
I also look for a small team because I feel that I operate best in a small team.
00:06:47.199 --> 00:06:51.920
And that's where I definitely get a lot of FaceTime and learnings from more senior investors.
00:06:52.160 --> 00:06:54.399
And I look for a track record of success.
00:06:54.560 --> 00:07:01.680
So for me, when I was sort of looking at roles, Tusk really stood out because I think it definitely has a differentiated strategy.
00:07:01.759 --> 00:07:05.120
It's a very small and lean team, and the track record is definitely there.
00:07:05.439 --> 00:07:08.240
And quickly touch on the strategy of Tusk.
00:07:08.560 --> 00:07:09.279
Yeah, absolutely.
00:07:09.439 --> 00:07:15.680
So Tusk does seed and series A investing in companies operating in highly regulated markets.
00:07:16.079 --> 00:07:33.839
That basically means that Tusk focuses on companies that they can help in one of three areas, whether that's being able to create a regulatory moat or a competitive advantage, being able to protect against downside risks, and then being able to help manage media and comms in a regulatorily complex environment.
00:07:34.079 --> 00:07:35.519
That's what the fund does as a whole.
00:07:35.680 --> 00:07:44.560
I can talk a lot more in detail about it, but we have a platform team that really helps us in almost in lockstep with the investment team to execute on those strategies.
00:07:44.639 --> 00:07:46.319
And that's where we really focus on.
00:07:46.560 --> 00:07:55.680
I'd love to hear a little bit more about the founding of Tusk and specifically why you focus on verticals that might be exposed to more regulatory risk.
00:07:55.920 --> 00:08:02.639
Seems like you're almost self-selecting towards the regulatory risk, something that some venture capital firms will sometimes avoid, right?
00:08:02.720 --> 00:08:04.079
Because it can be out of their hands.
00:08:04.319 --> 00:08:08.879
I would just love to hear a little bit more about the founding and reasoning for kind of targeting that strategy.
00:08:09.360 --> 00:08:14.240
Tusk definitely tends to lean in where other venture funds might lean away.
00:08:14.480 --> 00:08:23.519
And that does help us generate a ton of alpha because we are able to evaluate what deals we can underwrite in terms of regulatory risks and what we can't.
00:08:23.680 --> 00:08:34.799
So Bradley Tusk is a political strategist and was sort of instrumental in helping Uber get their license in New York and then in other areas on a state-by-state basis.
00:08:34.879 --> 00:08:39.440
And then he also helped Claire get into the um sort of the US airports.
00:08:39.519 --> 00:08:51.279
And so he's seen this background around operating with startups, helping them over overcome regulatory hurdles, and then also saw how there's opportunity to be able to invest and create value there.
00:08:51.440 --> 00:08:53.360
So he partnered up with Jordan Knoff.
00:08:53.519 --> 00:08:55.840
Jordan came from Blackstone's internal VC arm.
00:08:56.159 --> 00:08:58.720
Together they launched Fund One back in 2016.
00:08:58.879 --> 00:09:06.159
The fund did a bunch of investments out of which they were early investors in FanDuel, Lemonade, Coinbase, Bird, and Roe.
00:09:06.399 --> 00:09:12.960
And so it was definitely a very strong fund, and that definitely improved the value add in a lot of specific areas.
00:09:13.120 --> 00:09:20.240
So for example, with the Lemonade, we were able to get help them get some of their initial licenses to operate.
00:09:20.320 --> 00:09:25.600
And that's where we really helped de-risk a lot of the things that in a way that other venture funds can't.
00:09:25.840 --> 00:09:26.320
That's great.
00:09:26.480 --> 00:09:42.399
And have you felt that in pitching to various firms that that core strategy does kind of resonate with firms that are in those targeted, kind of more regulatory at-risk areas, or what's the feedback been from customers from founders that you've spoken to?
00:09:42.799 --> 00:09:54.960
I will say this coming only uh a few months into the role, but especially the founders that I've spoken to, for the ones that it is a good fit, it's a home run pitch and they tend to immediately get the value add.
00:09:55.200 --> 00:10:01.440
The best founders almost already know where you know their regulatory risks or opportunities are.
00:10:01.679 --> 00:10:04.480
And they would have probably thought of those before we even do.
00:10:04.639 --> 00:10:14.000
And for us, it's just helping them sort of think through those guardrails, think through those opportunities, and be a thought partner and helping them execute along those guidelines.
00:10:14.399 --> 00:10:21.840
Would love to touch on your experience and actually specifically a post that you had in the vertical software space.
00:10:22.080 --> 00:10:26.720
I think it was titled Vertical SAS 4.0, the great game theory.
00:10:27.120 --> 00:10:33.600
Something that I've referred to a couple of times and passed along to a couple investors, and it's it's really resonated among the people I've spoken to.
00:10:34.080 --> 00:10:40.960
So I'd love to touch on one, your experience writing the article and the different research that you kind of compiled coming into it.
00:10:41.039 --> 00:10:43.759
And then just your general thoughts on vertical SAS.
00:10:44.080 --> 00:10:48.399
Could you could you just kind of define for the audience what you view as vertical SAS?
00:10:48.799 --> 00:10:49.600
Yeah, absolutely.
00:10:49.679 --> 00:10:51.200
And uh thanks for the kind words.
00:10:51.360 --> 00:10:55.919
I was mostly writing my research thoughts, and it's great to be here that it was useful to other people.
00:10:56.240 --> 00:11:10.480
As far as why I chose to write that and and sort of the history of vertical SAS, you know, I've I've spent quite a bit of time looking at vertical SaaS as investment opportunities, both at B Capital and at Dallas VC, and hopefully continue to do doing that at Tusk.
00:11:10.720 --> 00:11:18.399
Just to give you an overview on what vertical SAS is and sort of the history of it, vertical SAS is simply software that's designed for a specific industry or niche.
00:11:18.639 --> 00:11:22.799
Common household vertical SaaS companies include Service Titan, Procore, Toast.
00:11:23.200 --> 00:11:36.720
And some of the key features I think that defines vertical SaaS is really having industry-specific solutions and capabilities, being able to integrate to other software that the companies in those industries use, and then having very targeted go-to-market strategies.
00:11:36.960 --> 00:11:47.440
I think it's also important to differentiate from horizontal SaaS, which is a lot of the typical SaaS we think about, in the sense that horizontal SaaS tends to cater to multiple industries.
00:11:47.679 --> 00:11:54.559
It also will have a larger market size because those industries are much bigger, and it tends to be a little bit harder to customize.
00:11:54.799 --> 00:12:02.559
But as far as like the overview of history, the way I sort of personally split them up is into four generations.
00:12:03.200 --> 00:12:21.600
The first generation is really the early wave of solutions that came out in the late 90s and early 2000s, and that was almost software with a specific industry focus, maybe a few key specializations, but it was really around having a few core functionalities and reducing IT spend.
00:12:21.840 --> 00:12:27.519
And companies out of that wave were like, for example, Weva, which focused on life sciences.
00:12:27.679 --> 00:12:34.960
I think Gen 2 was really around adding more functionalities, really defining it as a solution for a specific industry.
00:12:35.120 --> 00:12:41.679
At the same time, you had the rise of mobile, so you had mobile integrations, you also had APIs becoming more common.
00:12:41.919 --> 00:12:44.799
So companies like Procore came out of those era.
00:12:45.360 --> 00:12:50.159
And then Gen 3 was really having the foundations of AI in there.
00:12:50.240 --> 00:12:56.480
So some deep learning, some machine learning, having insights in analytics, and also having embedded finance.
00:12:56.799 --> 00:12:58.720
Mindbody is a great example of this.
00:12:58.960 --> 00:13:04.879
I worked on a company at B Capital called Curb Waste that did that is what I would define as a Gen 3 company.
00:13:05.039 --> 00:13:10.320
And then Gen 4 is really being super AI driven and hyper specialized.
00:13:10.399 --> 00:13:16.960
And I think that's what you're really starting to see with foundational models uh really coming into play.
00:13:17.279 --> 00:13:21.200
That's the wave of AI that's coming out into vertical SaaS today.
00:13:21.279 --> 00:13:26.159
And that's why you also see other investors refer to vertical SaaS as vertical AI now.
00:13:26.399 --> 00:13:32.320
I'd love to hear just a little bit more about how AI is specifically disrupting horizontal SaaS.
00:13:32.720 --> 00:13:38.320
There's a series of benefits and cons to having a very big reliance on AI.
00:13:38.480 --> 00:13:43.279
It definitely makes it very easy to build very industry-specific products.
00:13:43.440 --> 00:13:52.240
You can turn out code faster, you can have more integrations and process data a whole lot faster, and you can have improved analytics and outcomes.
00:13:52.399 --> 00:13:56.080
So it makes tools and software a lot more customizable.
00:13:56.399 --> 00:14:00.639
But at the same time, it creates a very heavy data dependency.
00:14:00.879 --> 00:14:14.720
And a result of that is you're gonna see incumbents that have mass sets of data that they're already working with having a first mover advantage because they are already able to train on in models that become much more defensible as they continue to innovate.
00:14:14.879 --> 00:14:23.919
At the same time, the other consideration is that I don't necessarily think AI drill uh delivers on all the hype that people necessarily say it does.
00:14:24.240 --> 00:14:25.919
There's gonna be a lot of limitations, right?
00:14:26.000 --> 00:14:32.080
And you're gonna see a few use cases coming out, you're gonna see like co-pilots, but it's not general purpose AI.
00:14:32.240 --> 00:14:33.600
It's not gonna solve all the problems.
00:14:33.759 --> 00:14:40.720
There's still gonna be a lot of uh manual work and customization that humans need to be involved in as far as software goes.
00:14:41.200 --> 00:14:45.600
It would be great to just walk through one area, like an example of disruption.
00:14:46.240 --> 00:15:03.679
The article you you mentioned, a bunch of some of the ones that have resonated with me personally, or for example, CRM, where historically you have a Salesforce or a HubSpot that's a general purpose CRM, but you're starting to see more vertically built CRMs like an Affinity, who Shubum Gloal is someone that Laura and I have spoken with on the podcast before.
00:15:03.840 --> 00:15:11.440
So I guess we'd love any example or just some way to kind of ground the listeners in just an example of what you're talking about.
00:15:11.679 --> 00:15:22.000
I think I sort of think through this in a broader framework that involves thinking about what is a horizontal solution that doesn't benefit from economies of scale.
00:15:22.320 --> 00:15:41.519
So when I think of economies of scale, I'm thinking about software or just sort of the layers of uh technology and things like having emails, that's not necessarily something that's going to be completely verticalized because there's just such benefit to Microsoft or Google providing a whole platform for that or cloud computing.
00:15:41.759 --> 00:16:04.159
But at the same time, when you think about customer analytics, you think about CRMs, as you mentioned, you think about all these sort of different specific use cases that become part of the enterprise operational stacks that don't necessarily need to have economies to scale, I think those are all the areas that become affected by vertical SaaS, sort of encroaching upon those capabilities.
00:16:04.399 --> 00:16:05.440
Yeah, no, that's very helpful.
00:16:05.519 --> 00:16:09.120
And then I guess it my next question would be where does the line stop?
00:16:09.440 --> 00:16:28.240
If you have general, say it's laundromat software that kind of services laundromats and other washer-dryer, maybe dry cleaning as well, then the next iteration of vertical sass, is that going to be purpose built for laundromats that are in the northeast focused on button-downs?
00:16:28.720 --> 00:16:34.000
At what point does the software actually become to an addressable market to focus on?
00:16:34.159 --> 00:16:37.519
And kind of where does that specialization stop in your mind?
00:16:37.840 --> 00:16:39.360
Yeah, I think that's a great question.
00:16:39.440 --> 00:16:41.679
And I think there's two ways to answer it.
00:16:41.919 --> 00:16:47.360
One is through sort of the VC lens, and the other is really from a customer perspective.
00:16:47.519 --> 00:16:53.279
I would say from a VC perspective, there's a trade-off between market size and differentiation.
00:16:53.759 --> 00:17:12.000
And you can have all the differentiation you want, but as long as your SAM, if your SAM is less than a billion dollars, it becomes very hard from a VC perspective to be able to invest in those companies because those venture fundable returns aren't necessarily there.
00:17:12.480 --> 00:17:14.319
That's sort of the easy answer.
00:17:14.640 --> 00:17:32.640
From a customer perspective, I think the trade-off is really around the customization that vertical SaaS provides, how much of a pain point they particularly need to have, and the integrations that vertical SaaS solutions can provide that are very unique to their businesses.
00:17:32.799 --> 00:17:43.119
And that's in comparison to less vertical or even horizontal softwares, where they do have economies of scale through things like pricing and larger support teams and things like that.
00:17:43.279 --> 00:17:55.119
I think today the limitation is primarily around AI capabilities when it comes to from a customer perspective, whether AI is able to solve a unique value proposition that a more generalist software necessarily isn't able to.
00:17:55.359 --> 00:18:05.920
How do you feel that various AI impacting the cost of developing software, copilot or something like that, reducing the cost of producing software?
00:18:06.000 --> 00:18:12.400
How is that impacting the ability for people to actually produce these vertically specific software solutions?
00:18:12.799 --> 00:18:22.799
Definitely easier, but at the same time, the execution and the capabilities of delivering a software are just going to remain the same, right?
00:18:22.960 --> 00:18:37.759
It comes back to sort of the foundational uh foundations of success that enterprises need, which is high-quality customer service, uh, successful integrations, continued product development, having high quality execution always tends to make or break these businesses.
00:18:38.000 --> 00:18:53.599
So if you're able to lower down the costs significantly and also provide a service that is competitive or much better in terms of being able to return a hard high ROI for these businesses, it definitely is a value added.
00:18:54.160 --> 00:19:02.000
The trade-off that becomes there is between what the use case really is and the size of those companies and sort of their budgets.
00:19:02.160 --> 00:19:26.559
And so if you look at something like a laundromat or like a field service operations, maybe for those businesses, having a lightweight sleeve or a co-pilot that's able to provide some instructional database or a faster way to process some type of analytics is a really great ROI from an IT budget and benefit point of view.
00:19:26.799 --> 00:19:38.160
But let's say tomorrow you're a large megabank and you're looking at a specific use case that's gonna impact you, your customers, and potentially billions or trillions of dollars.
00:19:38.400 --> 00:19:48.799
For them, that probably requires security, it requires customizability, it requires that AI abilities to be able to integrate and iterate itself.
00:19:48.960 --> 00:19:58.880
And so for those, you're gonna have a larger budget and you might have a much more customizable solution that's more deeply integrated into your workflow than just something that's fairly surface level.
00:19:59.200 --> 00:20:12.720
Next, we'd love to compare the benefits of the vertical specific large language models and how maybe they can leverage that industry-specific expertise to potentially outperform some of the general AI models.
00:20:12.960 --> 00:20:17.039
We recently read an article that focuses on this, which was published by Tidemark.
00:20:17.440 --> 00:20:20.240
It's called AI, Extinction or Evolution.
00:20:20.480 --> 00:20:32.079
They're a well-respective vertical SaaS fund for all of our listeners who aren't familiar, going back to kind of the specialized LLMs and they're industry specific compared to the general purpose models.
00:20:32.480 --> 00:20:38.400
On the whole, there's a ton of benefit to having large language models that are trained on vertical specific data sets.
00:20:38.559 --> 00:20:45.119
And they're able to answer contextual answers that general purpose LLMs aren't particularly fine-tuned for.
00:20:45.359 --> 00:20:49.279
And I think it again comes back to sort of those use case requirements.
00:20:49.440 --> 00:20:50.720
And they can vary, right?
00:20:50.799 --> 00:20:54.400
Some of those requirements can be stuff around accuracy rates or errors.
00:20:54.480 --> 00:20:59.680
And in certain fields, there's a much lower margin for error as opposed to other areas.
00:20:59.839 --> 00:21:05.599
Also, explainability and understanding how those large language models are coming to a certain output.
00:21:05.839 --> 00:21:16.480
And then also looking at having the ease of prompt engineering, being able to integrate with workflows, and also having a high level of data security, both in motion in those models themselves.
00:21:16.720 --> 00:21:21.759
There's a lot of variance in the complexity of the capabilities that LLMs can perform.
00:21:21.920 --> 00:21:26.400
And sort of the more specialized it is, the higher complexity it's going to have.
00:21:26.640 --> 00:21:40.720
So, as I was talking about, having a solution that's just sort of a lightweight sleeve is a lot more easier to implement than something that's integrated into a workflow that's able to sort of be fine-tuned and be integrated.
00:21:40.880 --> 00:21:44.559
It really comes down to how much they're able to become customizable.
00:21:44.720 --> 00:21:51.359
And that's sort of a trade off that any business or an IT buyer needs to sort of decide for themselves.
00:21:51.759 --> 00:21:59.599
One consideration that becomes interesting as time goes by is sort of the platform strategies that these foundational models.
00:22:00.160 --> 00:22:01.359
Providers are bringing in.
00:22:01.519 --> 00:22:06.960
So, I mean, pretty recently, OpenAI released the Chat GPT version for enterprise.
00:22:07.200 --> 00:22:21.119
And what you saw was the initial feature or capability that they're providing is a way to protect your data in terms of where who's feeding what data in it and where that data is going to be stored.
00:22:21.279 --> 00:22:26.720
And that already disrupts a whole segment of AI infrastructure companies.
00:22:26.960 --> 00:22:53.440
And so I think it's very interesting to keep an eye on how these platforms continue to evolve their offerings because at some point it may make sense for an enterprise to just use a platform with some lightweight tools that they offer themselves, similar to how AWS might offer a certain set of cloud features versus using a specific cloud management system or some sort of use case in that manner.
00:22:53.680 --> 00:23:02.559
And double-clicking on, I guess, the fields or let's call it like areas or industries within vertical SaaS that would benefit from these models.
00:23:02.720 --> 00:23:09.200
Do you have any top of mind or that kind of you're excited about that might see this change in behavior?
00:23:10.000 --> 00:23:23.839
There are a lot of industries that are that already have some level of analytics, but the more data and contextual knowledge they can have feeding into their operations, the better they'll become.
00:23:24.160 --> 00:23:30.480
These tend to be fairly complex industries in terms of the functions and the workflows.
00:23:30.720 --> 00:23:37.759
So we recently did an investment in a company called the Contract Network, which is basically vertical AI for the legal tech industry.
00:23:37.920 --> 00:23:45.440
And if you look at those industries, you're gonna have companies that have a lot of specialization.
00:23:45.519 --> 00:23:53.599
The law firms will have a ton of specialization in terms of the kind of laws they work on, the kind of ways in which an LOM can provide value for them.
00:23:53.759 --> 00:23:59.519
And so it's a lot more complex than a typical field force might need on a day-to-day basis.
00:23:59.680 --> 00:24:03.839
And so it really depends on sort of that complexity of the business operations.
00:24:04.319 --> 00:24:06.400
Legal Tech is a great example of that.
00:24:06.799 --> 00:24:10.720
So do you think that platforms such as the one you just invested in?
00:24:10.880 --> 00:24:15.759
I guess what was your thesis in terms of maybe competing with some of the legacy players like Clio?
00:24:16.480 --> 00:24:39.599
For us, it was taking a particular use case that's a very big segment within the legal industry, which is contracts, and looking at how did those specifically become more efficient in the way they're agreed upon, in the way they're negotiated, and finding a solution that addresses that with the latest and greatest in technology.
00:24:39.839 --> 00:24:43.920
That's where we found the contract network and it fit our thesis there.
00:24:44.000 --> 00:24:47.599
And it really does solve a ton of pain points there.
00:24:47.920 --> 00:24:57.039
And I'd love to move back to your thoughts on the end state of large language models and their application to vertically specific use cases.
00:24:57.279 --> 00:24:59.519
Right now, I've heard two trains of thought.
00:24:59.680 --> 00:25:15.359
One is leveraging the general purpose foundation models like a GPT-4 and applying it with some training data leveled on top, but generally leveraging that foundation model that's that's not vertically specific and using that.
00:25:15.519 --> 00:25:26.319
Like others think that there could be actual vertically specific foundation models, which obviously have a really high cost to develop, or even enterprises developing their own foundation models.
00:25:26.960 --> 00:25:35.759
So that's kind of a spectrum of of how large language models could be implemented and what the end state could be in a vertically specific application.
00:25:35.839 --> 00:25:42.400
I'm curious to get your thoughts on where you think the application of large language models sits in five or ten years.
00:25:43.039 --> 00:25:48.240
It's a good thought exercise in terms of in which potentially when this shapes out.
00:25:48.480 --> 00:25:52.480
Broadly speaking, there is a combination of the two.
00:25:52.640 --> 00:25:56.079
I don't think this necessarily becomes winner-takes all market.
00:25:56.319 --> 00:25:59.759
And I don't think there's necessarily market domination either.
00:26:00.079 --> 00:26:08.720
For smaller use cases and industries, having an industry agnostic solution uh probably makes a lot of sense in terms of sort of that ROI.
00:26:09.039 --> 00:26:19.039
But with larger organizations, having a particular or their own sort of uh large language model, if that becomes scalable, I think becomes a lot more interesting as well.
00:26:19.279 --> 00:26:21.599
Both of those models long term will work.
00:26:21.680 --> 00:26:23.920
It's gonna be more dependent on the use case.
00:26:24.160 --> 00:26:29.599
One winner on both of these sides is really gonna be a lot of tech-enabled services companies as well.
00:26:29.759 --> 00:26:41.680
There is gonna be a lot of customization and domain-specific knowledge that's gonna be required in terms of implementing best practices and having companies that specialize in this is gonna be really, really helpful.
00:26:42.000 --> 00:26:45.759
There's been a lot of noise in the VC ecosystem regarding platform teams.
00:26:45.920 --> 00:26:52.640
A few months ago, we saw that you know Sequoia laid off a third of their talent team, which is only a division of a platform team.
00:26:52.960 --> 00:26:57.359
There's been debate around their value add and how many people to have on the team.
00:26:57.599 --> 00:27:02.880
And so Tusk actually has a pretty big platform team, almost as big as their investment team.
00:27:03.039 --> 00:27:10.400
So we'd love to hear why that is a compelling strategy for a VC fund and where does your platform team help out?
00:27:10.799 --> 00:27:14.240
Our platform team is core to our model as a fund.
00:27:14.720 --> 00:27:25.519
If you look at Bradley's background with Uber and with Travis, his background is really as an operator and someone who's helped solve really critical problems for companies.
00:27:25.680 --> 00:27:39.839
And so combining that with uh Jordan's previous experience as an investor, obviously everyone on our team is an investor, but we look at how do we really build a platform team that became lockstep with our investment team.
00:27:40.000 --> 00:27:42.799
And so our platform team helps us in a lot of different areas.
00:27:43.119 --> 00:27:54.240
They're able to keep an eye on regulations, they're very critical to helping us sort of think about areas that we should invest in and sort of the thesis and macro trends that matter, and obviously helping our companies after the fact.
00:27:54.480 --> 00:28:00.720
And I think because of this, we're fairly unique as a venture fund because our strategy is very specific and differentiated.
00:28:01.119 --> 00:28:06.640
We tend to focus on being really impactful for the very specific companies that we invest in.
00:28:06.880 --> 00:28:14.400
But because of that, we're often sort of the most important, or maybe the better word is impactful investor on our cap table.
00:28:14.640 --> 00:28:29.440
And because of this, our platform team focuses on three primary activities, which was really identifying and helping startups capture on regulatory tailwinds, monitoring and protecting against sort of regulatory risk, and then managing those media and comps.
00:28:29.599 --> 00:28:32.079
And I think we do that and execute on that really well.
00:28:32.319 --> 00:28:33.119
Yeah, I agree.
00:28:33.200 --> 00:28:39.680
And we can also get into the fees and the valuations and whatnot in another episode, but you're spot on.
00:28:39.920 --> 00:28:43.200
And so to wrap up the episode, we have two questions for you.
00:28:43.359 --> 00:28:46.160
And the first one is what motivates you to work hard?
00:28:46.559 --> 00:28:54.240
I think I'm really lucky in the sense that I see venture for me as a vocation and not just a job.
00:28:54.400 --> 00:29:00.799
I really get energized by just the work that I do and working with amazing founders and helping them be successful.
00:29:01.039 --> 00:29:08.160
And so it's a natural output of that perspective and having that opportunity to do something that I truly love.
00:29:08.400 --> 00:29:12.079
I don't necessarily have to work hard to work hard.
00:29:12.160 --> 00:29:18.160
It's just uh something that fairly comes naturally just to be in a position that I really truly enjoy.
00:29:18.400 --> 00:29:18.799
For sure.
00:29:18.880 --> 00:29:19.920
It's a fun job.
00:29:20.160 --> 00:29:23.359
And what's a piece of advice you would give your younger self?
00:29:23.680 --> 00:29:29.440
I think broadly speaking, I'd say to just focus on my natural strengths and just let them shine.
00:29:29.759 --> 00:29:30.079
Great.
00:29:30.240 --> 00:29:31.759
I think that's an awesome note to end on.
00:29:31.920 --> 00:29:35.039
Manu, thank you so much for taking the time to chat with Laura and I today.
00:29:35.119 --> 00:29:36.559
We've enjoyed the conversation a lot.
00:29:36.799 --> 00:29:37.200
Absolutely.
00:29:37.359 --> 00:29:38.000
It's been a pleasure.
00:29:38.079 --> 00:29:39.839
Thank you so much, Will and Laura.
00:00:00.080 --> 00:00:05.759
I would say from a VC perspective, there's a trade-off between market size and differentiation.
00:00:06.320 --> 00:00:24.559
And you can have all the differentiation you want, but as long as your SAM if your SAM is less than a billion dollars, it becomes very hard from a VC perspective to be able to invest in those companies because those venture fundable returns aren't necessarily there.
00:00:31.920 --> 00:00:38.880
Welcome to Partner Path, a podcast that unpacks the venture capital and growth equity ecosystem from a junior perspective.
00:00:39.039 --> 00:00:45.359
Young entrepreneurs and investors have already had a massive impact on the industry, having started unicorns and launched billion-dollar funds.
00:00:45.759 --> 00:00:52.079
We discuss these success stories and more by sharing perspectives and advice from some of the industry's most prominent role models.
00:00:52.240 --> 00:00:58.640
This podcast expresses our views as of the date published and does not represent the views of and is not endorsed by any company for which we work.
00:01:05.280 --> 00:01:09.760
Manu started his career at Dallas VC before heading to business school in Chicago.
00:01:09.920 --> 00:01:19.519
While in business school, Manu did early stage investing at B Capital, then landed at TVP, an early stage VC fund focused on investing in startups operating in highly regulated industries.
00:01:19.680 --> 00:01:20.560
Welcome, Manu.
00:01:21.040 --> 00:01:21.680
Thank you, Laura.
00:01:21.760 --> 00:01:22.319
Thank you, Will.
00:01:22.400 --> 00:01:23.120
Thanks for having me.
00:01:23.359 --> 00:01:28.319
To start, could you walk us through your background and what led you to come to the US, specifically Texas?
00:01:28.640 --> 00:01:29.680
Yeah, absolutely.
00:01:29.920 --> 00:01:35.120
So I was born in India, but my dad worked for IBM all throughout the 90s.
00:01:35.280 --> 00:01:40.480
And so I lived in a bunch of different areas and had a very interesting childhood growing up.
00:01:40.719 --> 00:01:46.640
I lived in Thailand, Singapore, UK, and the UAE before the age of seven.
00:01:46.799 --> 00:01:50.239
And then ended up moving to Texas when I was eight.
00:01:50.400 --> 00:01:56.560
So it was a lot of hopping, and then all of a sudden in one place for 15, 20 years.
00:01:56.959 --> 00:02:07.599
And you started your VC career in Texas too, specifically working for a solo GP and building the fund essentially from the ground up, which I can imagine was very challenging.
00:02:07.760 --> 00:02:13.759
So it would be great if you could touch on kind of your role in building Dallas VC and what you learned from that experience.
00:02:14.080 --> 00:02:15.120
Yeah, absolutely.
00:02:15.280 --> 00:02:20.960
Uh was very, very lucky to have that opportunity and definitely walked away with a tons of learning.
00:02:21.120 --> 00:02:31.599
So I sort of joined as the first non-partner employee, almost you could think of it at that time as a solo GP uh fund that was sort of nearing the end of its first fund.
00:02:32.000 --> 00:02:35.199
And my first job was to really just help out with the portfolio.
00:02:35.280 --> 00:02:42.400
And I stepped into a bunch of different roles and did anything from PM or BizDev and just learned a lot.
00:02:42.560 --> 00:02:49.120
And once I did that for about a couple months, I sort of transitioned to doing a little more of the VC kind of roles.
00:02:49.280 --> 00:02:52.400
And that started off with sourcing, and then over time did diligence.
00:02:52.479 --> 00:03:06.719
And as the fund performed and we sort of moved into the new fund, I was starting to get some portfolio responsibilities and then eventually over time started doing my own deals, which was awesome to be able to do at such a young age.
00:03:06.960 --> 00:03:09.439
I actually got a few small exits under my belt.
00:03:09.680 --> 00:03:17.439
I also helped manage a lot of the fund operations and did a ton of LP relations and I actually got to do a little bit of fundraising as well.
00:03:17.599 --> 00:03:19.039
So it was a ton of learning.
00:03:19.199 --> 00:03:27.680
We went from two people managing about$15 million to about 12 people managing close to 200 by the time I left, two geographies.
00:03:27.759 --> 00:03:29.439
We also launched a fund in India.
00:03:29.680 --> 00:03:36.560
So it was a tremendous experience and definitely walked away with a bunch of stuff that were really important learnings.
00:03:36.879 --> 00:03:41.599
Another interesting part of your path back to VC is choosing to go to business school.
00:03:41.759 --> 00:03:49.840
I think nowadays, you know, we're at the point at that age 24, 25, people debating to studying and going back, especially when you're already in VC.
00:03:50.080 --> 00:03:54.639
I think some would argue that going back sometimes makes it harder to get back into VC.
00:03:54.800 --> 00:03:59.759
So I'm curious, when you decided to go to booth, was the plan to come back into VC?
00:04:00.240 --> 00:04:02.319
Yeah, the plan was definitely to come back.
00:04:02.560 --> 00:04:10.240
Business school was primarily because A, I felt that being in Dallas, it's a great ecosystem.
00:04:10.319 --> 00:04:15.360
You have a ton of corporate enterprises and you have a really good research institutions and things like that.
00:04:15.520 --> 00:04:22.319
But at the same time, you sort of don't necessarily have the critical mass that a New York or San Francisco will have.
00:04:22.560 --> 00:04:26.079
And so for me, I really wanted to be in a bigger ecosystem.
00:04:26.160 --> 00:04:40.000
And I felt that going to business school would be sort of a great way to be able to get exposure to those ecosystems, but then also just take a step back from five years of drinking from the fire hose and get a wider perspective in the world of business.
00:04:40.240 --> 00:04:41.680
Booth for me was really interesting.
00:04:41.839 --> 00:04:44.480
It has a very systematic approach to decision making.
00:04:44.639 --> 00:04:48.480
And I think that is really helpful for becoming a disciplined investor.
00:04:48.720 --> 00:04:51.040
But for me, the plan was definitely always to come back.
00:04:51.279 --> 00:04:51.600
Got it.
00:04:51.680 --> 00:04:58.399
And I'm sure the network as well that you've built at business school will only be helpful in VC and whatever you choose to do next.
00:04:58.639 --> 00:04:59.839
Absolutely.
00:05:00.160 --> 00:05:06.079
Would you say that does going to business school make go getting into venture capital easier?
00:05:06.240 --> 00:05:12.240
Let's say whether you come from a non-traditional background and you're looking to switch to that investing role?
00:05:12.639 --> 00:05:19.759
I think it depends on what background you're really coming from and where the world of venture capital you want to target.
00:05:20.000 --> 00:05:23.120
I think for on the whole, it's much easier than coming from undergrad.
00:05:23.680 --> 00:05:28.959
But there are going to be target types that venture funds will primarily look for.
00:05:29.120 --> 00:05:37.920
And that's going to be someone who's either been a prior investor or an industry specialist or an operator or product manager at a fast-growing early stage startup.
00:05:38.160 --> 00:05:43.439
But for people that are not coming from those backgrounds that want to transition into venture, it's also very doable.
00:05:43.600 --> 00:05:47.839
Business school is one of those things where it's a matter of you get out what you put in.
00:05:47.920 --> 00:05:54.319
And so if you're the kind of person that seeks opportunities, seeks internships, it's not that difficult to get a full-time offer.
00:05:54.480 --> 00:06:04.240
Obviously, there are other macro factors like the market, but on the whole, people, at least what I've seen in booths, people that try and really shoot for VC as a role end up getting jobs in VC.
00:06:04.480 --> 00:06:05.439
Yeah, that's a good point.
00:06:05.519 --> 00:06:08.800
And you can also start working too while you're a first or second year.
00:06:08.879 --> 00:06:11.920
So getting that VC experience part-time for a fund.
00:06:12.160 --> 00:06:12.720
Exactly.
00:06:13.040 --> 00:06:16.240
You decide go back into VC, start recruiting.
00:06:16.399 --> 00:06:23.519
What were you looking for in a VC fund, maybe just to broaden your skill set as you were going back into the recruiting landscape?
00:06:23.920 --> 00:06:39.759
One of the things I'd learned at Dallas and some of the other firms I'd interned at was that having a really differentiated strategy is really successful to being a venture fund that goes through multiple funds and is really something that's established.
00:06:39.920 --> 00:06:43.279
And so I looked for, above all, a differentiated strategy.
00:06:43.439 --> 00:06:47.120
I also look for a small team because I feel that I operate best in a small team.
00:06:47.199 --> 00:06:51.920
And that's where I definitely get a lot of FaceTime and learnings from more senior investors.
00:06:52.160 --> 00:06:54.399
And I look for a track record of success.
00:06:54.560 --> 00:07:01.680
So for me, when I was sort of looking at roles, Tusk really stood out because I think it definitely has a differentiated strategy.
00:07:01.759 --> 00:07:05.120
It's a very small and lean team, and the track record is definitely there.
00:07:05.439 --> 00:07:08.240
And quickly touch on the strategy of Tusk.
00:07:08.560 --> 00:07:09.279
Yeah, absolutely.
00:07:09.439 --> 00:07:15.680
So Tusk does seed and series A investing in companies operating in highly regulated markets.
00:07:16.079 --> 00:07:33.839
That basically means that Tusk focuses on companies that they can help in one of three areas, whether that's being able to create a regulatory moat or a competitive advantage, being able to protect against downside risks, and then being able to help manage media and comms in a regulatorily complex environment.
00:07:34.079 --> 00:07:35.519
That's what the fund does as a whole.
00:07:35.680 --> 00:07:44.560
I can talk a lot more in detail about it, but we have a platform team that really helps us in almost in lockstep with the investment team to execute on those strategies.
00:07:44.639 --> 00:07:46.319
And that's where we really focus on.
00:07:46.560 --> 00:07:55.680
I'd love to hear a little bit more about the founding of Tusk and specifically why you focus on verticals that might be exposed to more regulatory risk.
00:07:55.920 --> 00:08:02.639
Seems like you're almost self-selecting towards the regulatory risk, something that some venture capital firms will sometimes avoid, right?
00:08:02.720 --> 00:08:04.079
Because it can be out of their hands.
00:08:04.319 --> 00:08:08.879
I would just love to hear a little bit more about the founding and reasoning for kind of targeting that strategy.
00:08:09.360 --> 00:08:14.240
Tusk definitely tends to lean in where other venture funds might lean away.
00:08:14.480 --> 00:08:23.519
And that does help us generate a ton of alpha because we are able to evaluate what deals we can underwrite in terms of regulatory risks and what we can't.
00:08:23.680 --> 00:08:34.799
So Bradley Tusk is a political strategist and was sort of instrumental in helping Uber get their license in New York and then in other areas on a state-by-state basis.
00:08:34.879 --> 00:08:39.440
And then he also helped Claire get into the um sort of the US airports.
00:08:39.519 --> 00:08:51.279
And so he's seen this background around operating with startups, helping them over overcome regulatory hurdles, and then also saw how there's opportunity to be able to invest and create value there.
00:08:51.440 --> 00:08:53.360
So he partnered up with Jordan Knoff.
00:08:53.519 --> 00:08:55.840
Jordan came from Blackstone's internal VC arm.
00:08:56.159 --> 00:08:58.720
Together they launched Fund One back in 2016.
00:08:58.879 --> 00:09:06.159
The fund did a bunch of investments out of which they were early investors in FanDuel, Lemonade, Coinbase, Bird, and Roe.
00:09:06.399 --> 00:09:12.960
And so it was definitely a very strong fund, and that definitely improved the value add in a lot of specific areas.
00:09:13.120 --> 00:09:20.240
So for example, with the Lemonade, we were able to get help them get some of their initial licenses to operate.
00:09:20.320 --> 00:09:25.600
And that's where we really helped de-risk a lot of the things that in a way that other venture funds can't.
00:09:25.840 --> 00:09:26.320
That's great.
00:09:26.480 --> 00:09:42.399
And have you felt that in pitching to various firms that that core strategy does kind of resonate with firms that are in those targeted, kind of more regulatory at-risk areas, or what's the feedback been from customers from founders that you've spoken to?
00:09:42.799 --> 00:09:54.960
I will say this coming only uh a few months into the role, but especially the founders that I've spoken to, for the ones that it is a good fit, it's a home run pitch and they tend to immediately get the value add.
00:09:55.200 --> 00:10:01.440
The best founders almost already know where you know their regulatory risks or opportunities are.
00:10:01.679 --> 00:10:04.480
And they would have probably thought of those before we even do.
00:10:04.639 --> 00:10:14.000
And for us, it's just helping them sort of think through those guardrails, think through those opportunities, and be a thought partner and helping them execute along those guidelines.
00:10:14.399 --> 00:10:21.840
Would love to touch on your experience and actually specifically a post that you had in the vertical software space.
00:10:22.080 --> 00:10:26.720
I think it was titled Vertical SAS 4.0, the great game theory.
00:10:27.120 --> 00:10:33.600
Something that I've referred to a couple of times and passed along to a couple investors, and it's it's really resonated among the people I've spoken to.
00:10:34.080 --> 00:10:40.960
So I'd love to touch on one, your experience writing the article and the different research that you kind of compiled coming into it.
00:10:41.039 --> 00:10:43.759
And then just your general thoughts on vertical SAS.
00:10:44.080 --> 00:10:48.399
Could you could you just kind of define for the audience what you view as vertical SAS?
00:10:48.799 --> 00:10:49.600
Yeah, absolutely.
00:10:49.679 --> 00:10:51.200
And uh thanks for the kind words.
00:10:51.360 --> 00:10:55.919
I was mostly writing my research thoughts, and it's great to be here that it was useful to other people.
00:10:56.240 --> 00:11:10.480
As far as why I chose to write that and and sort of the history of vertical SAS, you know, I've I've spent quite a bit of time looking at vertical SaaS as investment opportunities, both at B Capital and at Dallas VC, and hopefully continue to do doing that at Tusk.
00:11:10.720 --> 00:11:18.399
Just to give you an overview on what vertical SAS is and sort of the history of it, vertical SAS is simply software that's designed for a specific industry or niche.
00:11:18.639 --> 00:11:22.799
Common household vertical SaaS companies include Service Titan, Procore, Toast.
00:11:23.200 --> 00:11:36.720
And some of the key features I think that defines vertical SaaS is really having industry-specific solutions and capabilities, being able to integrate to other software that the companies in those industries use, and then having very targeted go-to-market strategies.
00:11:36.960 --> 00:11:47.440
I think it's also important to differentiate from horizontal SaaS, which is a lot of the typical SaaS we think about, in the sense that horizontal SaaS tends to cater to multiple industries.
00:11:47.679 --> 00:11:54.559
It also will have a larger market size because those industries are much bigger, and it tends to be a little bit harder to customize.
00:11:54.799 --> 00:12:02.559
But as far as like the overview of history, the way I sort of personally split them up is into four generations.
00:12:03.200 --> 00:12:21.600
The first generation is really the early wave of solutions that came out in the late 90s and early 2000s, and that was almost software with a specific industry focus, maybe a few key specializations, but it was really around having a few core functionalities and reducing IT spend.
00:12:21.840 --> 00:12:27.519
And companies out of that wave were like, for example, Weva, which focused on life sciences.
00:12:27.679 --> 00:12:34.960
I think Gen 2 was really around adding more functionalities, really defining it as a solution for a specific industry.
00:12:35.120 --> 00:12:41.679
At the same time, you had the rise of mobile, so you had mobile integrations, you also had APIs becoming more common.
00:12:41.919 --> 00:12:44.799
So companies like Procore came out of those era.
00:12:45.360 --> 00:12:50.159
And then Gen 3 was really having the foundations of AI in there.
00:12:50.240 --> 00:12:56.480
So some deep learning, some machine learning, having insights in analytics, and also having embedded finance.
00:12:56.799 --> 00:12:58.720
Mindbody is a great example of this.
00:12:58.960 --> 00:13:04.879
I worked on a company at B Capital called Curb Waste that did that is what I would define as a Gen 3 company.
00:13:05.039 --> 00:13:10.320
And then Gen 4 is really being super AI driven and hyper specialized.
00:13:10.399 --> 00:13:16.960
And I think that's what you're really starting to see with foundational models uh really coming into play.
00:13:17.279 --> 00:13:21.200
That's the wave of AI that's coming out into vertical SaaS today.
00:13:21.279 --> 00:13:26.159
And that's why you also see other investors refer to vertical SaaS as vertical AI now.
00:13:26.399 --> 00:13:32.320
I'd love to hear just a little bit more about how AI is specifically disrupting horizontal SaaS.
00:13:32.720 --> 00:13:38.320
There's a series of benefits and cons to having a very big reliance on AI.
00:13:38.480 --> 00:13:43.279
It definitely makes it very easy to build very industry-specific products.
00:13:43.440 --> 00:13:52.240
You can turn out code faster, you can have more integrations and process data a whole lot faster, and you can have improved analytics and outcomes.
00:13:52.399 --> 00:13:56.080
So it makes tools and software a lot more customizable.
00:13:56.399 --> 00:14:00.639
But at the same time, it creates a very heavy data dependency.
00:14:00.879 --> 00:14:14.720
And a result of that is you're gonna see incumbents that have mass sets of data that they're already working with having a first mover advantage because they are already able to train on in models that become much more defensible as they continue to innovate.
00:14:14.879 --> 00:14:23.919
At the same time, the other consideration is that I don't necessarily think AI drill uh delivers on all the hype that people necessarily say it does.
00:14:24.240 --> 00:14:25.919
There's gonna be a lot of limitations, right?
00:14:26.000 --> 00:14:32.080
And you're gonna see a few use cases coming out, you're gonna see like co-pilots, but it's not general purpose AI.
00:14:32.240 --> 00:14:33.600
It's not gonna solve all the problems.
00:14:33.759 --> 00:14:40.720
There's still gonna be a lot of uh manual work and customization that humans need to be involved in as far as software goes.
00:14:41.200 --> 00:14:45.600
It would be great to just walk through one area, like an example of disruption.
00:14:46.240 --> 00:15:03.679
The article you you mentioned, a bunch of some of the ones that have resonated with me personally, or for example, CRM, where historically you have a Salesforce or a HubSpot that's a general purpose CRM, but you're starting to see more vertically built CRMs like an Affinity, who Shubum Gloal is someone that Laura and I have spoken with on the podcast before.
00:15:03.840 --> 00:15:11.440
So I guess we'd love any example or just some way to kind of ground the listeners in just an example of what you're talking about.
00:15:11.679 --> 00:15:22.000
I think I sort of think through this in a broader framework that involves thinking about what is a horizontal solution that doesn't benefit from economies of scale.
00:15:22.320 --> 00:15:41.519
So when I think of economies of scale, I'm thinking about software or just sort of the layers of uh technology and things like having emails, that's not necessarily something that's going to be completely verticalized because there's just such benefit to Microsoft or Google providing a whole platform for that or cloud computing.
00:15:41.759 --> 00:16:04.159
But at the same time, when you think about customer analytics, you think about CRMs, as you mentioned, you think about all these sort of different specific use cases that become part of the enterprise operational stacks that don't necessarily need to have economies to scale, I think those are all the areas that become affected by vertical SaaS, sort of encroaching upon those capabilities.
00:16:04.399 --> 00:16:05.440
Yeah, no, that's very helpful.
00:16:05.519 --> 00:16:09.120
And then I guess it my next question would be where does the line stop?
00:16:09.440 --> 00:16:28.240
If you have general, say it's laundromat software that kind of services laundromats and other washer-dryer, maybe dry cleaning as well, then the next iteration of vertical sass, is that going to be purpose built for laundromats that are in the northeast focused on button-downs?
00:16:28.720 --> 00:16:34.000
At what point does the software actually become to an addressable market to focus on?
00:16:34.159 --> 00:16:37.519
And kind of where does that specialization stop in your mind?
00:16:37.840 --> 00:16:39.360
Yeah, I think that's a great question.
00:16:39.440 --> 00:16:41.679
And I think there's two ways to answer it.
00:16:41.919 --> 00:16:47.360
One is through sort of the VC lens, and the other is really from a customer perspective.
00:16:47.519 --> 00:16:53.279
I would say from a VC perspective, there's a trade-off between market size and differentiation.
00:16:53.759 --> 00:17:12.000
And you can have all the differentiation you want, but as long as your SAM, if your SAM is less than a billion dollars, it becomes very hard from a VC perspective to be able to invest in those companies because those venture fundable returns aren't necessarily there.
00:17:12.480 --> 00:17:14.319
That's sort of the easy answer.
00:17:14.640 --> 00:17:32.640
From a customer perspective, I think the trade-off is really around the customization that vertical SaaS provides, how much of a pain point they particularly need to have, and the integrations that vertical SaaS solutions can provide that are very unique to their businesses.
00:17:32.799 --> 00:17:43.119
And that's in comparison to less vertical or even horizontal softwares, where they do have economies of scale through things like pricing and larger support teams and things like that.
00:17:43.279 --> 00:17:55.119
I think today the limitation is primarily around AI capabilities when it comes to from a customer perspective, whether AI is able to solve a unique value proposition that a more generalist software necessarily isn't able to.
00:17:55.359 --> 00:18:05.920
How do you feel that various AI impacting the cost of developing software, copilot or something like that, reducing the cost of producing software?
00:18:06.000 --> 00:18:12.400
How is that impacting the ability for people to actually produce these vertically specific software solutions?
00:18:12.799 --> 00:18:22.799
Definitely easier, but at the same time, the execution and the capabilities of delivering a software are just going to remain the same, right?
00:18:22.960 --> 00:18:37.759
It comes back to sort of the foundational uh foundations of success that enterprises need, which is high-quality customer service, uh, successful integrations, continued product development, having high quality execution always tends to make or break these businesses.
00:18:38.000 --> 00:18:53.599
So if you're able to lower down the costs significantly and also provide a service that is competitive or much better in terms of being able to return a hard high ROI for these businesses, it definitely is a value added.
00:18:54.160 --> 00:19:02.000
The trade-off that becomes there is between what the use case really is and the size of those companies and sort of their budgets.
00:19:02.160 --> 00:19:26.559
And so if you look at something like a laundromat or like a field service operations, maybe for those businesses, having a lightweight sleeve or a co-pilot that's able to provide some instructional database or a faster way to process some type of analytics is a really great ROI from an IT budget and benefit point of view.
00:19:26.799 --> 00:19:38.160
But let's say tomorrow you're a large megabank and you're looking at a specific use case that's gonna impact you, your customers, and potentially billions or trillions of dollars.
00:19:38.400 --> 00:19:48.799
For them, that probably requires security, it requires customizability, it requires that AI abilities to be able to integrate and iterate itself.
00:19:48.960 --> 00:19:58.880
And so for those, you're gonna have a larger budget and you might have a much more customizable solution that's more deeply integrated into your workflow than just something that's fairly surface level.
00:19:59.200 --> 00:20:12.720
Next, we'd love to compare the benefits of the vertical specific large language models and how maybe they can leverage that industry-specific expertise to potentially outperform some of the general AI models.
00:20:12.960 --> 00:20:17.039
We recently read an article that focuses on this, which was published by Tidemark.
00:20:17.440 --> 00:20:20.240
It's called AI, Extinction or Evolution.
00:20:20.480 --> 00:20:32.079
They're a well-respective vertical SaaS fund for all of our listeners who aren't familiar, going back to kind of the specialized LLMs and they're industry specific compared to the general purpose models.
00:20:32.480 --> 00:20:38.400
On the whole, there's a ton of benefit to having large language models that are trained on vertical specific data sets.
00:20:38.559 --> 00:20:45.119
And they're able to answer contextual answers that general purpose LLMs aren't particularly fine-tuned for.
00:20:45.359 --> 00:20:49.279
And I think it again comes back to sort of those use case requirements.
00:20:49.440 --> 00:20:50.720
And they can vary, right?
00:20:50.799 --> 00:20:54.400
Some of those requirements can be stuff around accuracy rates or errors.
00:20:54.480 --> 00:20:59.680
And in certain fields, there's a much lower margin for error as opposed to other areas.
00:20:59.839 --> 00:21:05.599
Also, explainability and understanding how those large language models are coming to a certain output.
00:21:05.839 --> 00:21:16.480
And then also looking at having the ease of prompt engineering, being able to integrate with workflows, and also having a high level of data security, both in motion in those models themselves.
00:21:16.720 --> 00:21:21.759
There's a lot of variance in the complexity of the capabilities that LLMs can perform.
00:21:21.920 --> 00:21:26.400
And sort of the more specialized it is, the higher complexity it's going to have.
00:21:26.640 --> 00:21:40.720
So, as I was talking about, having a solution that's just sort of a lightweight sleeve is a lot more easier to implement than something that's integrated into a workflow that's able to sort of be fine-tuned and be integrated.
00:21:40.880 --> 00:21:44.559
It really comes down to how much they're able to become customizable.
00:21:44.720 --> 00:21:51.359
And that's sort of a trade off that any business or an IT buyer needs to sort of decide for themselves.
00:21:51.759 --> 00:21:59.599
One consideration that becomes interesting as time goes by is sort of the platform strategies that these foundational models.
00:22:00.160 --> 00:22:01.359
Providers are bringing in.
00:22:01.519 --> 00:22:06.960
So, I mean, pretty recently, OpenAI released the Chat GPT version for enterprise.
00:22:07.200 --> 00:22:21.119
And what you saw was the initial feature or capability that they're providing is a way to protect your data in terms of where who's feeding what data in it and where that data is going to be stored.
00:22:21.279 --> 00:22:26.720
And that already disrupts a whole segment of AI infrastructure companies.
00:22:26.960 --> 00:22:53.440
And so I think it's very interesting to keep an eye on how these platforms continue to evolve their offerings because at some point it may make sense for an enterprise to just use a platform with some lightweight tools that they offer themselves, similar to how AWS might offer a certain set of cloud features versus using a specific cloud management system or some sort of use case in that manner.
00:22:53.680 --> 00:23:02.559
And double-clicking on, I guess, the fields or let's call it like areas or industries within vertical SaaS that would benefit from these models.
00:23:02.720 --> 00:23:09.200
Do you have any top of mind or that kind of you're excited about that might see this change in behavior?
00:23:10.000 --> 00:23:23.839
There are a lot of industries that are that already have some level of analytics, but the more data and contextual knowledge they can have feeding into their operations, the better they'll become.
00:23:24.160 --> 00:23:30.480
These tend to be fairly complex industries in terms of the functions and the workflows.
00:23:30.720 --> 00:23:37.759
So we recently did an investment in a company called the Contract Network, which is basically vertical AI for the legal tech industry.
00:23:37.920 --> 00:23:45.440
And if you look at those industries, you're gonna have companies that have a lot of specialization.
00:23:45.519 --> 00:23:53.599
The law firms will have a ton of specialization in terms of the kind of laws they work on, the kind of ways in which an LOM can provide value for them.
00:23:53.759 --> 00:23:59.519
And so it's a lot more complex than a typical field force might need on a day-to-day basis.
00:23:59.680 --> 00:24:03.839
And so it really depends on sort of that complexity of the business operations.
00:24:04.319 --> 00:24:06.400
Legal Tech is a great example of that.
00:24:06.799 --> 00:24:10.720
So do you think that platforms such as the one you just invested in?
00:24:10.880 --> 00:24:15.759
I guess what was your thesis in terms of maybe competing with some of the legacy players like Clio?
00:24:16.480 --> 00:24:39.599
For us, it was taking a particular use case that's a very big segment within the legal industry, which is contracts, and looking at how did those specifically become more efficient in the way they're agreed upon, in the way they're negotiated, and finding a solution that addresses that with the latest and greatest in technology.
00:24:39.839 --> 00:24:43.920
That's where we found the contract network and it fit our thesis there.
00:24:44.000 --> 00:24:47.599
And it really does solve a ton of pain points there.
00:24:47.920 --> 00:24:57.039
And I'd love to move back to your thoughts on the end state of large language models and their application to vertically specific use cases.
00:24:57.279 --> 00:24:59.519
Right now, I've heard two trains of thought.
00:24:59.680 --> 00:25:15.359
One is leveraging the general purpose foundation models like a GPT-4 and applying it with some training data leveled on top, but generally leveraging that foundation model that's that's not vertically specific and using that.
00:25:15.519 --> 00:25:26.319
Like others think that there could be actual vertically specific foundation models, which obviously have a really high cost to develop, or even enterprises developing their own foundation models.
00:25:26.960 --> 00:25:35.759
So that's kind of a spectrum of of how large language models could be implemented and what the end state could be in a vertically specific application.
00:25:35.839 --> 00:25:42.400
I'm curious to get your thoughts on where you think the application of large language models sits in five or ten years.
00:25:43.039 --> 00:25:48.240
It's a good thought exercise in terms of in which potentially when this shapes out.
00:25:48.480 --> 00:25:52.480
Broadly speaking, there is a combination of the two.
00:25:52.640 --> 00:25:56.079
I don't think this necessarily becomes winner-takes all market.
00:25:56.319 --> 00:25:59.759
And I don't think there's necessarily market domination either.
00:26:00.079 --> 00:26:08.720
For smaller use cases and industries, having an industry agnostic solution uh probably makes a lot of sense in terms of sort of that ROI.
00:26:09.039 --> 00:26:19.039
But with larger organizations, having a particular or their own sort of uh large language model, if that becomes scalable, I think becomes a lot more interesting as well.
00:26:19.279 --> 00:26:21.599
Both of those models long term will work.
00:26:21.680 --> 00:26:23.920
It's gonna be more dependent on the use case.
00:26:24.160 --> 00:26:29.599
One winner on both of these sides is really gonna be a lot of tech-enabled services companies as well.
00:26:29.759 --> 00:26:41.680
There is gonna be a lot of customization and domain-specific knowledge that's gonna be required in terms of implementing best practices and having companies that specialize in this is gonna be really, really helpful.
00:26:42.000 --> 00:26:45.759
There's been a lot of noise in the VC ecosystem regarding platform teams.
00:26:45.920 --> 00:26:52.640
A few months ago, we saw that you know Sequoia laid off a third of their talent team, which is only a division of a platform team.
00:26:52.960 --> 00:26:57.359
There's been debate around their value add and how many people to have on the team.
00:26:57.599 --> 00:27:02.880
And so Tusk actually has a pretty big platform team, almost as big as their investment team.
00:27:03.039 --> 00:27:10.400
So we'd love to hear why that is a compelling strategy for a VC fund and where does your platform team help out?
00:27:10.799 --> 00:27:14.240
Our platform team is core to our model as a fund.
00:27:14.720 --> 00:27:25.519
If you look at Bradley's background with Uber and with Travis, his background is really as an operator and someone who's helped solve really critical problems for companies.
00:27:25.680 --> 00:27:39.839
And so combining that with uh Jordan's previous experience as an investor, obviously everyone on our team is an investor, but we look at how do we really build a platform team that became lockstep with our investment team.
00:27:40.000 --> 00:27:42.799
And so our platform team helps us in a lot of different areas.
00:27:43.119 --> 00:27:54.240
They're able to keep an eye on regulations, they're very critical to helping us sort of think about areas that we should invest in and sort of the thesis and macro trends that matter, and obviously helping our companies after the fact.
00:27:54.480 --> 00:28:00.720
And I think because of this, we're fairly unique as a venture fund because our strategy is very specific and differentiated.
00:28:01.119 --> 00:28:06.640
We tend to focus on being really impactful for the very specific companies that we invest in.
00:28:06.880 --> 00:28:14.400
But because of that, we're often sort of the most important, or maybe the better word is impactful investor on our cap table.
00:28:14.640 --> 00:28:29.440
And because of this, our platform team focuses on three primary activities, which was really identifying and helping startups capture on regulatory tailwinds, monitoring and protecting against sort of regulatory risk, and then managing those media and comps.
00:28:29.599 --> 00:28:32.079
And I think we do that and execute on that really well.
00:28:32.319 --> 00:28:33.119
Yeah, I agree.
00:28:33.200 --> 00:28:39.680
And we can also get into the fees and the valuations and whatnot in another episode, but you're spot on.
00:28:39.920 --> 00:28:43.200
And so to wrap up the episode, we have two questions for you.
00:28:43.359 --> 00:28:46.160
And the first one is what motivates you to work hard?
00:28:46.559 --> 00:28:54.240
I think I'm really lucky in the sense that I see venture for me as a vocation and not just a job.
00:28:54.400 --> 00:29:00.799
I really get energized by just the work that I do and working with amazing founders and helping them be successful.
00:29:01.039 --> 00:29:08.160
And so it's a natural output of that perspective and having that opportunity to do something that I truly love.
00:29:08.400 --> 00:29:12.079
I don't necessarily have to work hard to work hard.
00:29:12.160 --> 00:29:18.160
It's just uh something that fairly comes naturally just to be in a position that I really truly enjoy.
00:29:18.400 --> 00:29:18.799
For sure.
00:29:18.880 --> 00:29:19.920
It's a fun job.
00:29:20.160 --> 00:29:23.359
And what's a piece of advice you would give your younger self?
00:29:23.680 --> 00:29:29.440
I think broadly speaking, I'd say to just focus on my natural strengths and just let them shine.
00:29:29.759 --> 00:29:30.079
Great.
00:29:30.240 --> 00:29:31.759
I think that's an awesome note to end on.
00:29:31.920 --> 00:29:35.039
Manu, thank you so much for taking the time to chat with Laura and I today.
00:29:35.119 --> 00:29:36.559
We've enjoyed the conversation a lot.
00:29:36.799 --> 00:29:37.200
Absolutely.
00:29:37.359 --> 00:29:38.000
It's been a pleasure.
00:29:38.079 --> 00:29:39.839
Thank you so much, Will and Laura.
