WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=J7n3WJnFTto

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/J7n3WJnFTto.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:13.080 --> 00:00:13.520
Hey everyone.

s3
00:00:13.520 --> 00:00:15.240
My name is Pratik Naidu.

s4
00:00:15.240 --> 00:00:26.480
I'm a founding partner at Long Lake and today I'm excited to talk a little bit more about why acquiring businesses is a unique strategy to deploy AI into the real world.

s5
00:00:26.480 --> 00:00:30.880
So to start out very familiar experience to many of you.

s6
00:00:30.880 --> 00:00:35.040
Let's say you're building an AI product and now you're trying to sell it into the enterprise.

s7
00:00:35.040 --> 00:00:36.600
Well, where do you start?

s8
00:00:36.600 --> 00:00:38.040
You found a pain.

s9
00:00:38.040 --> 00:00:42.880
You built an agent for it and it works, but for one very narrow use case.

s10
00:00:42.880 --> 00:00:44.560
It's a great starting point.

s11
00:00:44.560 --> 00:00:47.720
But now you have to actually get it into production.

s12
00:00:47.720 --> 00:00:54.960
And that might involve dealing with security review, internal coordination, integrations with 20-year-old systems.

s13
00:00:54.960 --> 00:01:00.880
It's a lot of hard work and this is typically the first wall in order to deploy AI into an enterprise.

s14
00:01:00.880 --> 00:01:05.400
But once you get past this, now you have to get the team to use it.

s15
00:01:05.400 --> 00:01:11.120
So you typically will iterate on the product interface, go through trainings, get feedback from users.

s16
00:01:11.120 --> 00:01:17.360
And there's a typical question that comes up, which is whose job is it to actually drive up that adoption?

s17
00:01:17.360 --> 00:01:22.800
Is it yours as the AI vendor or is it the client that you're working with?

s18
00:01:22.840 --> 00:01:30.880
And then let's say you actually go through all this and now you're trying to figure out how do I actually get revenue for the agent that I just deployed.

s19
00:01:30.880 --> 00:01:33.120
You'll typically work through a few different questions.

s20
00:01:33.120 --> 00:01:36.000
Do you sell tokens or seats, maybe a bit of both?

s21
00:01:36.000 --> 00:01:39.000
How do you deal with price sensitivity from your client?

s22
00:01:39.000 --> 00:01:42.000
And how do you actually then measure ROI?

s23
00:01:42.000 --> 00:01:51.800
And the bigger picture question here is as an AI startup deploying into the enterprise, how do you capture the value that you create from your deployment?

s24
00:01:51.920 --> 00:01:53.840
This all might sound very familiar to many of you.

s25
00:01:53.840 --> 00:02:04.600
It certainly is to me and these walls of deployment reflect a pretty common challenge, which is you're building on the outside and that makes deploying AI really hard.

s26
00:02:04.600 --> 00:02:06.280
And importantly, it's not the model.

s27
00:02:06.280 --> 00:02:08.679
The models, as you know, are getting incredibly good.

s28
00:02:08.679 --> 00:02:11.920
Intelligence is exponentially increasing.

s29
00:02:11.920 --> 00:02:14.800
So, what makes this really hard?

s30
00:02:14.800 --> 00:02:22.080
And I want to talk about a few different observations of why deploying AI into businesses is challenging.

s31
00:02:22.120 --> 00:02:26.160
The first is that organizations are incredibly complex.

s32
00:02:26.160 --> 00:02:29.160
On paper, your org chart might look something on the left.

s33
00:02:29.160 --> 00:02:34.600
You know, you have your owner, you have teams, managers, and this is a standard team structure.

s34
00:02:34.600 --> 00:02:38.760
But, the reality is organizations are far more nuanced.

s35
00:02:38.760 --> 00:02:45.240
There are decisions that are entrenched in a web of team members, often times folks with decades of experience.

s36
00:02:45.240 --> 00:02:48.240
They're interacting with the sales team, the operations team.

s37
00:02:48.240 --> 00:02:50.920
The owner often times leans on them.

s38
00:02:50.920 --> 00:02:58.320
And so, building an agent as a third-party vendor is difficult because you actually need to embed into the organization.

s39
00:02:59.239 --> 00:03:09.440
The second reason I want to touch on is you can't deliver the full value of your deployment if you don't understand the nuances of how a business actually runs.

s40
00:03:09.440 --> 00:03:19.760
So, a vendor might integrate with your systems of record like a CRM or a ticketing system, or even, you know, document repositories where your SOPs, contracts, or, you know, old email threads live.

s41
00:03:19.760 --> 00:03:23.880
And this is really valuable source of information, very useful.

s42
00:03:23.880 --> 00:03:29.720
But, the reality is businesses, especially in complex domains, run on human judgment.

s43
00:03:29.720 --> 00:03:37.440
It's the exceptions, the edge cases, the tacit knowledge, the, you know, quick questions that get asked to the expert that really matters.

s44
00:03:37.440 --> 00:03:40.920
And that's really what governs a lot of how a company runs.

s45
00:03:40.920 --> 00:03:42.440
And that's the operating reality.

s46
00:03:42.440 --> 00:03:49.440
You need to work with the people doing the work to understand what great work actually looks like.

s47
00:03:50.920 --> 00:03:59.200
The third reason that I want to highlight is that even if you are able to deliver end-to-end value, it's pretty hard to capture value as an AI vendor.

s48
00:03:59.200 --> 00:04:04.280
So, let's say you land a seven-figure million-dollar contract, which is pretty impressive.

s49
00:04:04.280 --> 00:04:08.600
But, how do you think about the enduring business value of that contract?

s50
00:04:08.600 --> 00:04:10.000
A lot of big questions to answer.

s51
00:04:10.000 --> 00:04:11.360
Will customers keep renewing?

s52
00:04:11.360 --> 00:04:16.760
Can you maintain pricing as models improve and as alternatives get cheaper?

s53
00:04:16.760 --> 00:04:22.919
But, let's say as part of that million-dollar-a-year revenue engagement, your client actually gets a lot of value from it.

s54
00:04:22.919 --> 00:04:26.520
In fact, maybe $5 million a year in EBITDA impact.

s55
00:04:26.520 --> 00:04:37.880
And so, what's interesting about this is that with an established business, you have years of operating history, enduring client relationships, and a proven business model with real economics.

s56
00:04:37.880 --> 00:04:43.360
And so, you're adding earnings to a business with value that you can already reason about.

s57
00:04:43.360 --> 00:04:53.400
And so, you know, using a, you know, roughly conservative 10x multiple, you can actually generate for your client $50 million in enterprise value, which is really interesting.

s58
00:04:53.400 --> 00:04:56.960
And so, there's this question here, which is, you know, as a vendor, you did all the work.

s59
00:04:56.960 --> 00:05:03.040
You've deployed your agents, you went through all the walls of integrating with the systems, working with the team.

s60
00:05:03.040 --> 00:05:09.520
But, the largest impact actually shows up in the P&L of the business that you actually worked with.

s61
00:05:10.040 --> 00:05:14.640
And so, these are the typical walls and challenges of deploying AI outside in.

s62
00:05:14.640 --> 00:05:21.560
So, why not just go build that firm and build that business yourself with AI at its core from day one?

s63
00:05:22.040 --> 00:05:23.720
This is a very valid strategy, actually.

s64
00:05:23.720 --> 00:05:27.880
You can improve the whole workflow, not just a narrow slice.

s65
00:05:27.880 --> 00:05:31.480
You sell the work itself, not just one specific tool.

s66
00:05:31.480 --> 00:05:32.960
And importantly, you own the outcomes.

s67
00:05:32.960 --> 00:05:36.360
So, the P&L and the value that you're generating is yours.

s68
00:05:36.360 --> 00:05:38.520
And many of these firms are going to be amazing.

s69
00:05:38.520 --> 00:05:40.919
We think this is a great strategy.

s70
00:05:40.919 --> 00:05:45.960
But, I also want to call out some of the challenges that come up with this approach.

s71
00:05:47.120 --> 00:05:49.880
The first is you don't start with any customers.

s72
00:05:49.880 --> 00:05:51.960
It's hard to get distribution.

s73
00:05:51.960 --> 00:05:57.400
You don't have any minimal you don't have any a brand reputation, so that makes it challenging to to grow.

s74
00:05:57.400 --> 00:06:07.480
And it's hard to hire expertise especially in you know niche niche industries and important industries where clients might need multiple specialty offerings.

s75
00:06:07.480 --> 00:06:08.520
So take an example here.

s76
00:06:08.520 --> 00:06:19.560
If you are the leader of a health system and you're planning a new surgery building, do you hand that project to an AI architecture startup or the firm that you've been working with for 20 years who knows

s77
00:06:19.560 --> 00:06:27.160
the standards, the nuances of how you actually design your buildings, and you can trust them to deliver the full scope of your project.

s78
00:06:27.160 --> 00:06:30.280
It's a very different kind of execution model.

s79
00:06:30.280 --> 00:06:44.120
And so the important insight here is that the best firms have spent decades earning their customers, building brand and trust, and building an incredible team to deliver great service to their clients.

s80
00:06:45.120 --> 00:06:58.440
So instead of being outside in and selling an AI agent to a company or trying to build from scratch, what if you could partner far more closely with incredible businesses that already exist

s81
00:06:58.440 --> 00:07:00.320
through acquisition?

s82
00:07:00.320 --> 00:07:06.120
And I want to talk a little bit more about why ownership is such an interesting strategy here.

s83
00:07:06.120 --> 00:07:08.919
Starting with on the engineering side, which is access.

s84
00:07:08.919 --> 00:07:20.680
The ability to have your team sit down with the people who do work, understand what they do, build those relationships, understand the key people, the veterans, the folks that have the 20 years of experience is super important

s85
00:07:20.680 --> 00:07:32.040
to really designing agents that not just solve one workflow or one use case, but really tackle the full automation and service delivery opportunity for the entire business.

s86
00:07:33.680 --> 00:07:35.520
The second is data.

s87
00:07:35.520 --> 00:07:44.960
Every business you can think of as really a live environment with real tasks, real edge cases, and real answers on whether or not an agent actually got something right versus not.

s88
00:07:44.960 --> 00:07:52.120
And this is a huge opportunity to use that information, use that signal to actually improve your agents and models.

s89
00:07:52.520 --> 00:07:58.240
But importantly, on the operational side, incentive alignment is huge.

s90
00:07:58.240 --> 00:07:59.840
Ownership solves for this.

s91
00:07:59.840 --> 00:08:02.640
You only win when your team and your business wins.

s92
00:08:02.640 --> 00:08:06.440
And there's no separate revenue goals, no distractions from other clients.

s93
00:08:06.440 --> 00:08:17.000
It creates an incredible amount of alignment that enables you to uncover and solve a lot of the blockers to ultimately get to faster deployment and faster shared goals.

s94
00:08:17.000 --> 00:08:19.760
Outcomes really matter here.

s95
00:08:21.880 --> 00:08:23.760
The other piece is trust, which we talked about.

s96
00:08:23.760 --> 00:08:31.360
And, you know, importantly, rather than coming outside-in or building from scratch, ownership has that trust element baked in from the start.

s97
00:08:31.360 --> 00:08:35.120
You know, you have the brand that took decades to build with all of the customer relationships.

s98
00:08:35.120 --> 00:08:38.000
And and naturally plays into this idea of scale.

s99
00:08:38.000 --> 00:08:49.760
Starting with a large base where you have those customers, you have an expert team that can deliver an incredible service, and the cash flow to continue to invest and compound in growth and new products is very powerful.

s100
00:08:49.760 --> 00:08:56.240
And the important piece here is these five characteristics aren't just independent benefits.

s101
00:08:56.240 --> 00:09:06.720
They actually create something far more powerful, which is a compounding advantage where a combination of your agents and your people continue to act in live operations.

s102
00:09:06.720 --> 00:09:14.160
You can understand how work is done, create real-world data that leads to real outcomes, and that learning all compounds together.

s103
00:09:14.160 --> 00:09:23.040
And this flywheel is really interesting, and it comes with all the elements of ownership to uh you know, power this flywheel from from day one.

s104
00:09:23.040 --> 00:09:28.400
And so, what's interesting here is that ownership uh can compound rapidly over time.

s105
00:09:28.720 --> 00:09:30.880
So, we talked about why ownership is is really interesting.

s106
00:09:30.880 --> 00:09:37.640
But what markets are ripe for acquisition, especially in the AI era?

s107
00:09:37.760 --> 00:09:39.840
I want to start with the software economy.

s108
00:09:39.840 --> 00:09:41.000
This is a global software market.

s109
00:09:41.000 --> 00:09:44.440
Most of us have spent a lot of our time and careers building in this part.

s110
00:09:44.440 --> 00:09:46.400
And you know, it's a pretty big market.

s111
00:09:46.400 --> 00:09:50.680
But what's interesting is that software is always sold tools to people doing the work.

s112
00:09:50.680 --> 00:09:53.440
If you zoom out a little bit, this is a services economy.

s113
00:09:53.440 --> 00:09:56.080
It's a 7 trillion 70 trillion global market.

s114
00:09:56.080 --> 00:10:03.400
And the important insight here is that owning the outcomes of a service end-to-end is far more valuable.

s115
00:10:03.400 --> 00:10:04.960
And that's what's exciting to us.

s116
00:10:04.960 --> 00:10:07.240
It's a huge market opportunity.

s117
00:10:07.240 --> 00:10:14.880
But there's an important observation here, which is under the hood, there's a challenge that services businesses typically face.

s118
00:10:15.600 --> 00:10:18.280
Labor costs typically grow with revenue.

s119
00:10:18.280 --> 00:10:25.200
So, uh for example, a community manager at an HOA management company can only run so many associations.

s120
00:10:25.200 --> 00:10:28.520
An architect can only design so many hospitals or schools.

s121
00:10:28.520 --> 00:10:31.720
So, if you want more clients, you have to hire more people.

s122
00:10:31.720 --> 00:10:34.040
And this is the typical challenge of uh a business.

s123
00:10:34.040 --> 00:10:41.760
And you know, if you compare it to a software company, if you add $100 of revenue in a typical software business, you keep about 80.

s124
00:10:41.760 --> 00:10:49.240
So, there's really low marginal cost of growth, which is why these businesses are incredible and you can scale very rapidly.

s125
00:10:49.960 --> 00:10:53.320
On the contrary, a service business operates very differently.

s126
00:10:53.320 --> 00:10:59.040
For every $100 of incremental revenue, you typically have around $80 or so of delivery and overhead.

s127
00:10:59.040 --> 00:11:01.920
You have to hire, train, manage your team.

s128
00:11:01.920 --> 00:11:10.680
And so many founders who have built incredible services businesses have really stopped chasing growth, not because they lack ambition and not because there's no market opportunity.

s129
00:11:10.680 --> 00:11:12.160
These are massive markets.

s130
00:11:12.160 --> 00:11:19.560
But because the challenges of scaling, when you have to do a lot of the work for every $100, you're only keeping about 20.

s131
00:11:19.560 --> 00:11:30.880
And so you can think of this as a pretty huge, you know, 80% marginal tax on growth, which makes it very challenging to both come outside in, start these businesses, and even scale them.

s132
00:11:31.160 --> 00:11:37.760
What's very interesting here is that AI actually reduces the cost of service while also improving the quality of your service.

s133
00:11:37.760 --> 00:11:47.240
So, look back at our chart here where revenue scales proportionally to labor costs, but when you deploy AI you can see here that you get actually operating leverage.

s134
00:11:47.240 --> 00:11:50.280
So, your labor cost no longer directly scales with AI.

s135
00:11:50.280 --> 00:12:01.880
You're still going to need to hire an incredible team of experts to deliver your service to your clients, but the service economics for these businesses start to look a lot more like software.

s136
00:12:01.880 --> 00:12:14.800
And so, the thesis of acquisition, deep partnership with incredible companies, and leveraging AI to scale these businesses and improve the quality of service is exactly what we're excited about here at Long Lake.

s137
00:12:14.800 --> 00:12:16.320
So, just quickly a little bit more about us.

s138
00:12:16.320 --> 00:12:19.400
Long Lake is an applied AI holding company.

s139
00:12:19.400 --> 00:12:27.960
We partner with incredible businesses, have raised 7 and 1/2 billion in capital from Elad Gil, General Catalyst, Alpha Wave, D1, and many others.

s140
00:12:27.960 --> 00:12:29.760
We were started about 3 years ago.

s141
00:12:29.760 --> 00:12:42.800
Our Our whole thesis is to acquire best-in-class services businesses, partner with them to transform their operations, help them grow faster, and ultimately deliver a better client experience.

s142
00:12:42.800 --> 00:12:50.360
We've today partnered with 40 businesses across a number of markets, residential services, architecture, HR, and many more.

s143
00:12:50.360 --> 00:12:55.360
We're an amazing team of 55 team members across technology, finance, and operations.

s144
00:12:55.360 --> 00:12:56.400
We're growing.

s145
00:12:56.400 --> 00:13:02.960
We have an eclectic mix of former founders, folks from the military, Palantir, a lot of familiar names here.

s146
00:13:02.960 --> 00:13:06.000
And we recently just announced our partnership with Amex GBT.

s147
00:13:06.000 --> 00:13:12.080
It's a 6.3 billion take private, bringing us into the corporate travel space, which we're incredibly excited about.

s148
00:13:12.080 --> 00:13:21.520
You know, we're modeled after a lot of the holding companies that have come before us, but with a really interesting observation that AI raises the ceiling for business transformation.

s149
00:13:21.520 --> 00:13:26.000
And this is the lever of excellence that we're excited to go and pursue.

s150
00:13:27.000 --> 00:13:31.080
So, we talked about why We talked about services in particular.

s151
00:13:31.080 --> 00:13:34.840
What services businesses though are actually attractive to own?

s152
00:13:34.840 --> 00:13:43.360
And I want to start with an important observation, which is as models continue to get better, the price of intelligence is rapidly declining.

s153
00:13:43.360 --> 00:13:44.800
And it's pretty incredible to see.

s154
00:13:44.800 --> 00:13:46.640
I think this is only going to continue.

s155
00:13:46.640 --> 00:13:51.480
So, if intelligence is getting cheap, what remains expensive?

s156
00:13:51.480 --> 00:13:58.720
We really focus on mission-critical businesses, and it starts with a few key questions that we typically evaluate these companies on.

s157
00:13:58.720 --> 00:14:04.720
The first is is there a customer base that trust these businesses with real end-to-end outcomes?

s158
00:14:04.720 --> 00:14:10.000
The second is is there human judgment in the service that's actually being provided?

s159
00:14:10.000 --> 00:14:18.000
And the third is that are there, you know, regulatory reasons that govern accountability when making these acquisition decisions?

s160
00:14:18.000 --> 00:14:25.080
And I think the important piece here is you have to own the outcome, and that's the typical characteristic of a lot of these services businesses.

s161
00:14:25.080 --> 00:14:35.400
So, to color this a bit more, it's not just about drafting a newsletter to a homeowner, but actually handling the end-to-end outcome of an urgent leak.

s162
00:14:35.400 --> 00:14:41.920
It's not just about extracting data from documents, but actually taking accountability for a complicated tax decision.

s163
00:14:41.920 --> 00:14:49.600
And, you know, it's not just about helping you book a flight, but handling the last-minute disruptions, which I'm sure many of you have faced before.

s164
00:14:50.880 --> 00:14:52.200
Businesses like these are rare.

s165
00:14:52.200 --> 00:14:59.400
You have to fit all of these questions, and many more, and also have a team that's excited about that partnership.

s166
00:14:59.400 --> 00:15:02.880
And to do this, you know, you have to see a lot of businesses.

s167
00:15:02.880 --> 00:15:05.560
Great investment decisions come with choice.

s168
00:15:05.560 --> 00:15:09.560
So, you have to keep a big funnel, and then an incredibly tight filter.

s169
00:15:09.560 --> 00:15:21.760
Our focus is on businesses that have incredible fundamentals from a business and market standpoint, but also have an awesome AI opportunity where we can actually go in and partner very closely with these businesses.

s170
00:15:21.760 --> 00:15:30.680
We see about 5,000, you know, deals a year, but again, are very focused and diligent on what we actually focus on and who we partner with.

s171
00:15:32.240 --> 00:15:32.560
Awesome.

s172
00:15:32.560 --> 00:15:37.800
So, once you've found your business and you've partnered with a company, what does P&L transformation actually look like?

s173
00:15:37.800 --> 00:15:43.760
And I think a very simple way to explain this is actually through the visual of a J-curve.

s174
00:15:43.760 --> 00:15:46.840
So, you start with your investment in AI deployment.

s175
00:15:46.840 --> 00:15:52.840
You have your platform, you build your tech, capture the workflows, commit a lot of team time to making this work.

s176
00:15:52.840 --> 00:15:54.600
And it's a lot of hard work.

s177
00:15:54.600 --> 00:15:57.040
You have to make it work within the business.

s178
00:15:57.040 --> 00:16:11.600
Thinking back to, you know, some of the challenges of building outside-in, you have to partner very closely with their team, iterate, validate your product, potentially train new models, look at all the different unique workflows and data that exist within that business.

s179
00:16:11.600 --> 00:16:19.560
But once you make it up this hill, you'll start to see folks actually using your product, which is incredibly exciting, and you start to actually unlock a lot of capacity.

s180
00:16:19.560 --> 00:16:22.640
Time savings, usage is going through the roof.

s181
00:16:22.640 --> 00:16:25.560
They're producing the same output, fewer hours.

s182
00:16:25.560 --> 00:16:28.400
But this isn't the the the end of the journey.

s183
00:16:28.400 --> 00:16:34.160
Importantly, we want to actually capture value and we want to generate real business impact with our partner companies.

s184
00:16:34.160 --> 00:16:36.680
And so, you actually have to put that capacity to work.

s185
00:16:36.680 --> 00:16:45.720
Think about incentive alignment, working very closely with our management teams and trusted partners to drive business outcomes, help with growth, improve customer and employee NPS.

s186
00:16:45.720 --> 00:16:50.080
These are all part of the equation to actually driving impact within these businesses.

s187
00:16:50.080 --> 00:16:53.839
And once you do all the those things, you get to real P&L impact.

s188
00:16:53.839 --> 00:16:59.200
Higher revenue per employee, but importantly, a far better service quality for our clients.

s189
00:16:59.200 --> 00:17:05.640
And I think what's cool about this is that this isn't just a bespoke playbook that you run one-off for every single business.

s190
00:17:05.640 --> 00:17:14.000
If you have the platform and the playbook to do both the AI deployment and the business transformation, this can rapidly compound.

s191
00:17:14.000 --> 00:17:21.480
You You get to see faster time to deploy agents and a faster time to P&L impact for all the new businesses that you partner with.

s192
00:17:21.480 --> 00:17:26.640
And we see this at Long Lake across the 40 plus businesses that we've already partnered with today.

s193
00:17:29.040 --> 00:17:31.640
Of course, it's not just about AI and business transformation.

s194
00:17:31.640 --> 00:17:36.560
Under the hood, it takes a pretty unique team to actually own, operate, and and transform.

s195
00:17:36.560 --> 00:17:47.160
So, you have to actually go and and buy these businesses, have an M&A team that can partner with world-class services businesses, work with them very closely to build that relationship, should build that excitement for the future.

s196
00:17:47.160 --> 00:17:49.600
Then you actually have to go to do the work of building, and this is a lot of hard work.

s197
00:17:49.600 --> 00:17:58.520
You have to fly out to Minneapolis, fly out to Portland, uh go to London, spend time with the team, and actually build and push the frontier of how agents get deployed.

s198
00:17:58.520 --> 00:18:03.400
Again, not just for one narrow use case, but really for the entire operation.

s199
00:18:03.400 --> 00:18:08.880
And then to actually do that deployment, you know, you need an amazing deployment and ops team to actually drive adoption, right?

s200
00:18:08.880 --> 00:18:10.480
Think about that at that J-curve, right?

s201
00:18:10.480 --> 00:18:13.200
Don't just stop at capacity and time savings.

s202
00:18:13.200 --> 00:18:24.800
How do you actually ensure that the work you're doing is driving real business impact for the companies that you work with, and delivering a far better client experience uh for the service that you're providing?

s203
00:18:25.720 --> 00:18:32.040
So, if you're excited about owning outcomes and partnering with uh incredible services businesses, feel free to reach out.

s204
00:18:32.040 --> 00:18:33.400
Uh we're excited about this thesis.

s205
00:18:33.400 --> 00:18:42.360
We think there's a lot of opportunity to go continue to build in the services economy, and uh feel free to chat more if there's any questions.

s206
00:18:42.360 --> 00:18:44.520
Thank you.