Vertical AI, Hardware + Software and Moats

Today I have Jason Shuman on our podcast. Jason is a general partner at Primary Venture Partners, one of the NYC's top seed stage firms with a billion dollars in AUM. We most recently just announced a $625 million set of funds, making us the largest seed fund in New York. I saw your tweet, you know, on x.com and you said that basically Stripe's real genius was abstracting compliance, fraud, and complexity and not just processing payments. And that's the bar for vertical AI. Now, do you see something like this happening in some segments of vertical AI?

Vertical software required so much interaction by the human being to input things into the system to get outputs. And so, we are seeing things that are jaw-dropping customer experiences. They are things that are doing a lot of the heavy lifting and the work for people and extracting all of the complexity. This is what's happening in vertical AI. As an investor, how do you look at it? Like, what are the things which are more important than the others?

Is it the data modes? Is it workflow ownership? Domain expertise? Is it distribution? How do you think about picking companies? I actually think there's like two different clear entry points the way that we see it today.

You can either build a very high-velocity wedge product, and if the customer sees it, they're like, "Oh my god, I want that." The other way that people tend to approach it is trying to build an AI-native system of record. If you are building an AI-native system of record, oftentimes you have a much more challenging go-to-market because it's a rip-and-replace motion for a very, very bulky thing. And I think I'd rather be betting on companies that can have the higher-velocity wedges right now just because it seems easier to to go in that Have you started looking at hardware plus plus software as a mode? I absolutely love hardware. Hardware can capture things that our eyes can capture, and hardware can also capture things that our eyes can't see. My belief is that, you know, so much human error and so much time and effort goes into capturing things with our eyes and manually inputting things into a system.

And then once those things are into a system, there are many downstream workflows that come out of that. You had tweeted about Anthropic's Series A round which was passed by every tier-one VC. That seems like such a glaring example. One of the most important companies of the generation was actually passed by most of the tier-one VCs. Do you think that still there is some pattern matching, something which can still work in terms of like, "Okay, these three things have to be there and then this will be like a generational company"? Yes, I do think there's still pattern matching, especially when it comes to founders.

And I do think that oftentimes a lot of the best companies of any generation appear non-consensus at the very early stages, although the the risk is changing because we are writing checks at much, much higher valuations than we did before, but they step they tend to become consensus over time. Hottest topic of right now, you know, there's some people call it SaaS-pocalypse. I think you said that software-only modes are dead. And I want to touch upon that. How should investors who have a lot of investments in you know, software companies, your pure-play software companies, how should they think about all the investments that they have done in the last 5 or 10 years? Welcome everyone to yet another episode of Build AI podcast.

Uh today I have I have Jason Shuman, um you know, on our podcast. It it took a while to get him on the pod, but here we are. Thanks Thanks, Jason, for joining us. I appreciate it, man. Thanks for having me. It's uh the the AI boom has everybody working around the clock, so I got to try to carve out more time for stuff like this.

Absolutely. So, um I have been following you on Twitter and also we chatted a couple of times before, so there's a bunch of contexts. And so, you know, I'm looking forward to to this discussion. There are a few different angles and questions that I want to ask you. Uh but before that, for everyone, uh Jason is a general partner at Primary Venture Partners, one of the NYC's top seed stage firms with a billion dollars in AUM. Uh Jason actually is a founder, ex-founder who turned into VC, and he went from building a D2C footwear brand in college uh to becoming an investor.

His current investment thesis is squarely focused on vertical AI and hardware and autonomy and and a few other things. We'll we'll touch upon that. In terms of the portfolio, Primary VC has invested in Dandy, Edge, Walmar, and so on. And then specifically, Jason has invested in companies like Bobbie Yard, Light Level, Fuse. Anything else, Jason, to add to this this list? No, I mean, I think that's uh a good summary there.

Uh we most recently just announced a $625 million set of funds, making us the largest seed fund in New York. And I'd say like our model is quite a bit different from most other seed funds in the sense that one, we're incredibly concentrated. We only do three or four deals a year per partner. The second is we all specialize. So, we've gone out and we've built, you know, multi-thousand-person expert and customer networks. The third is we have a forward-deployed impact teams.

We actually have staffed up from now 45, soon to be 60 people that embed in our companies whenever they want them around go-to-market. So, we have a go-to-market engineering product. We do some outbound for companies, recruiting, finance, corp dev, anything you really need in the early days to kind of help you get off the ground. And then the last thing, and and happy to chat a little bit about as well, I've been launching a number of companies on the incubation side with founders. Uh and and oftentimes, you know, building my own It's been a lot of fun over the last couple of years as AI has certainly made a lot easier for a non-technical person like me. Right.

And we'll definitely talk about that because that is something that we truly believe in at GI Ventures. So, it will be an interesting topic. Also, I I believe at Primary VC you guys have seen a few exits. Latch went public and, you know, I think one of your partners was invested in Coupang which got acquired by Amazon and then Walmar. So, there has been a decent amount of success as well. Yeah.

No, I mean, I think, you know, the first couple funds we've been lucky enough where about 90% of the companies have raised Series A and about 20% have become unicorns. And so, hoping that we can keep up with the new wave and get the opportunity to work with the best founders here. Right. Now, I want to start by talking about like uh you know, your transition. You started Category 5, I think your company was called Category 5. And from there to working, you know, for Corigin Ventures and then uh you know, working for Mark Gerson, GLG Group co-founder.

How Like, how has been the journey all the way to joining Primary Venture Capital? How have the roles changed? How did you prepare yourself for it? Man, you know, I was obsessed with startups when I was a kid, and my family are a bunch of small business owners and entrepreneurs. And honestly, I was very ADD growing up and I didn't really like school. So, I was always hustling around doing things.

And as you mentioned, I launched a direct-to-consumer footwear company when I was an undergrad. And it actually reminds me a little bit of the times right now in a sense because for me, it wasn't that I was like super passionate about footwear or super passionate about commerce. What it was was really an arbitrage at the time where Facebook was very cheap and, you know, excuse my language, Toby, but like Shopify was, you know, not good at all. It was very, very crappy. And so, there was an opportunity to go online and start to acquire customers directly. And candidly, I didn't know what I was doing at all.

I mean, I had not really operated and scaled a direct-to-consumer company, but I learned some incredible learnings from that time. Bootstrapped it, you know, we got close to about a million in revenue. Um and along the way, I would just say like you learn uh the more people you meet and like what great looks like and what the real like tactical hardships are. Uh and overcoming them is not so easy. And so, I have a lot of empathy for founders along the way. But about a year out of college, I realized it's not what I wanted to do for the next 10 years.

And so, I ended up winding down the business. And I drove for Uber actually for a little while and was sourcing deals in Boston, sending them to VCs in New York. And got an offer to move to New York to become a venture capitalist. And at the time, I would say VC was not super popular. It was not like the sexy job to have. But my way of doing it was just becoming completely obsessed with the craft and, you know, reading everything I could read, listening to podcasts or YouTube videos, and really admiring some of the great seed investors of the time, folks like Chris Sacca, you know, who started uh Lowercase Capital, now Lowercarbon, uh David Frankel from Founder Collective, you know, a number of these folks.

And I just tried to understand them. And then I really tried to understand what the great businesses look like. And so, studying those and coming up with investment theses. And I don't think anything fully, fully prepares you for for getting a job in venture capital because it's uh it's a job that requires you to source companies, requires you to understand the psychology of what makes a great founder, and then it's a sales job because you need to convince people to take your money. Uh but you need to be picking the right founders in the right markets at the end of the day. So, I spent a couple years there, and I got lucky enough to have backed Latch and Luke Schoenfelder there in the seed.

My co-investors were the guys at Primary, so that's how we got to know each other, you know, throughout that journey what as they went public. But after a couple years at Corigin, I got the opportunity to launch a family office for Mark. And, you know, Mark is an unbelievable entrepreneur, and he really believes in trusting [clears throat] and giving a lot of rope to young people. And so, I think I was doing deals across like real estate, credit, litigation finance. We were doing a lot of nonprofit work. But on the venture side, uh what I kind of find interesting and funny was I I was doing deals that I called the future of automation back then.

They really are AI companies, but I think of the nine investments I made, you know, two of them are are unicorns or soon to be unicorns, and the rest are dead. And the reason why the rest are dead was probably because AI just was not ready in for prime time, and I was too early. But a lot of really good, interesting learnings there because I didn't have a lot of time to do VC back then. So, I had to make commitments in meetings. I had to understand market structures pretty quickly. Um which was very, very helpful for for what I do now, which is much more concentrated investing, but simultaneously Mark had me doing some operating things and scaling a B2B software company.

So, there was a lot of high-pressure situations, there was a lot of prioritization that was required, and and a lot of self-directed learning. And I think if there's one takeaway from my time working with Mark for everybody listening to this, it's that you don't need to be told by anybody what you are capable of. go out there and learn and figure things out on your own. Information was completely democratized back then. If anything, it's now even more democratized with AI. And so, uh I think people that have drive and are just curious are going to thrive right now.

Well said. Thank you so much for sharing that. Okay, let let's uh let's jump on one of the topics that's uh very dear and near to you and for us as well, which is vertical AI. I really like that uh one I saw your tweet uh um you know, x.com now. I I still can't get used to it. Uh and you said that uh basically Stripe's real genius was abstracting compliance, fraud, and complexity and not just processing payment payments, and that's the bar for vertical AI.

Now, do you see something like this happening in some segments of vertical AI? Like are your favorites and not so favorites today in vertical AI? Yeah, I mean, when you think about vertical AI today versus what vertical software was uh not even a handful, but really only a few years ago, vertical software required so much interaction by the human being to input things into the system to get outputs. And so, we are seeing things that are jaw-dropping customer experiences. They are things that are doing a lot of the heavy lifting and the work for people and extracting all of the complexity and just providing a very strong outcome without having to go through the brain drain of like the numerous steps that certain jobs might require. So, an example that I could walk you through is a company like Bob Yard.

So, Bob Yard raised a $35 million Series A from 8VC. What they do is AI takeoffs for the trades. Previously, OCR technology existed, and you know, folks at a landscaping company might get a drawing set from, you know, an architect, and they might sit there and literally have to click on the drawing and count every single bush, you know, from one type of thing. And then every single tree. And like there might be a thousand items in there. So, that was somebody's full-time job was just going in there and trying to calculate how much of a certain quantities were there for different items.

And now with Bob Yard's AI, you know, algorithms, they can go and just do that job in minutes. And so, when you think about that industry, the number one bottleneck to revenue is being able to do takeoffs and being able to do quotes for people so they can respond to the RFQs. And previously, people either had those teams in-house or they outsourced it, and oftentimes there was like a one-week, two-week, three-week lead time, or they wouldn't even bid on certain projects because it just took too much time. Now, they can do it all in in a fraction of the time. And so, I think it's delivering outcomes that are ultimately driving revenue that get people really excited. And so, if you can take the AI products and back into what is a very obviously clear ROI and pain point or bottleneck for the business, I think you have a really good opportunity ahead.

And because there's so much happening in vertical AI, like literally any segment, subsegment you look at today, there are like 10, 20, or even 50 companies in in certain segments. How do you like as a as an investor, how do you look at it? Like what are the things which are more important than the others? Is it the data modes? Is it workflow ownership? Domain expertise?

Is it distribution? How do you think about picking companies? Yeah, we've had this debate a lot internally because I actually think there's like two different clear entry points the way that we see it today. You can either build a very high-velocity wedge product, and that product likely has a single-player mode use case to it that is that jaw-dropping customer experience, and ideally can be demoed or shown on X like you were saying and be understood in like a 30-second clip. And if the customer sees it, they're like, "Oh my god, I want that." And as a result, the conversion rate when you're on sales pitches tends to be north of 50%, which, you know, you and I've been doing this awhile, like that's really high. So, those types of plays though, when you start in a single-player wedge, might not be as defensible when it is in the wedge.

With that being said, the other way that people tend to approach it is trying to build an AI-native system of record. If you are building an AI-native system of record, oftentimes you have a much more challenging go-to-market because it's a rip and replace motion for a very very heavy, bulky thing. And while the defensibility is incredibly clear to me at least with the first step in that process, they tend not to grow as quickly, which then means they won't attract as much capital as the folks that have the high-velocity wedges. And so, the bet with the high-velocity wedge companies is that then they can go deeper into the system of record as migration times start to drop with the advancements in AI coding, and they can have a higher velocity product roadmap because AI coding is also unlocking the ability to build different features. I can unpack that a little bit more and how do I think about specific modes, but those are at least the two shapes of companies that I think about, and I think I'd rather be betting on companies that can have the higher velocity wedges right now just because it seems easier to to go into the system of record than vice versa. Right.

And and would love to go into more details. Uh you know, I I I think we'll touch upon it in the subsequent discussion. Something very related to this, like you you mentioned, you know, and that's the hottest topic right now, you know, this some people call it cesspool calves. I I think you said that software only modes are dead. Um and I want to touch upon that that how how should how should investors who have a lot of investments in uh you know, software companies, your pure-play software companies, how how should they think about all the investments that they have done in the last 5 or 10 years? Yeah, I mean, what's really interesting is none of us I really think know what are the modes going to be longer term right now.

It is a very very ambiguous and uncertain time. With that said, I do believe that certain software modes will continue to exist. And switching costs might be able to become higher if you have a network effect and collaboration software where you're working both internally with the AI tool and then pulling other stakeholders externally into the AI tool. And what is fascinating about today is that certain AI products are able to capture these ACVs that are as large if not larger than the core system of records with just one wedge product and then start to expand out into that multiplayer collaboration. So, an example of that would be a portfolio company of ours called Light Table where they've done AI class detection, AI peer reviews for um developers, large-scale developers. Historically, that has looked like a human being reviewing these other drawing sets and looking for where a structural beam might hit an HVAC like unit.

And a lot of developers didn't even do it before. They didn't even look over the drawings. And so, this company created like a Grammarly for construction drawings, and instead of having to wait 20 weeks and it being 50% accurate, these guys are turning it around in the same day. And so, that turnaround time is unlocking a completely different market, and simultaneously it's increasing the returns for investors by called two points because it's reducing change orders by 70%. Why is that important and why do I like that as a software mode? Because you're going in fast, but then developers work with general contractors, architects, subcontractors like electricians.

And now all of them need to come in on top of the core unit of work, which is the drawing set, and that is exactly where Light Table owns it. And so, I do think like that mode will continue to exist, but the challenge for incumbents right now is that they're going to get eaten from the outside and from so many other angles, and then recreating themselves internally to have a higher velocity product roadmap, go-to-market, and even pricing structure, which I think we could have a longer conversation about, is certainly challenging for the incumbents. Right. And do you see like on a related but slightly different topic, do you see role of hardware becoming more important? Like are you are you have you started looking at hardware plus plus software as a mode? I love hardware.

I absolutely love hardware. Um look, when you think about the physical world today, like hardware can capture things that our eyes can capture, and hardware can also capture things that our eyes can't see, things that live inside of pipes, for instance. And so, my belief is that, you know, so much human error and so much time and effort goes into capturing things with our eyes and manually inputting things into a system. And then once those things are into a system, there are many downstream workflows that come out of that, things like compliance documentation, billing, procurement, dispatching folks to come on site to fix the machine that I see is broken. And if all of a sudden you have hardware products that are dropping in cost and they're easy to deploy, whether it be traditional cameras, hyperspectral imaging cameras, or other sorts of sensors, and now it can catch it in real time, maybe even before my eyes saw it, and then it can get the person on site to fix the thing, or it can then do the billing or procurement. I just think that it's going to be really hard to disrupt that because once you install hardware, and I learned this very early on in my career, typically it's very hard to rip out because there's a lot of physical work that then has to go on.

That's a very interesting point. Like sensors and cameras and data that gets captured through these can supersede human what human eye can capture. For sure. I like to call them HANs, and people are like, "What the hell is a HAN?" It's like a hardware activated agent network because hardware it is capturing that data to then activate these sequencing of different agents who can then do the work that humans used to have to do. Right. Okay, I'm going to change gears now and talk about one like I tweeted actually based on something that you had shared.

Um and and this is like one of my favorite topics as I started learning about venture and investing. You know, I have invested in about 30 to 33 startups between my own personal investments as well as of the company. And you know, of course you have been doing it for much longer and at a much bigger scale. I always often ask investors that is there any pattern matching that still works in VC? And I remember you had tweeted about Anthropic Series A round which was passed by every tier one VC. And it was ultimately done by I think the co-founder of Stripe.

And that seems like such a glaring example. Like the the most one of the most important companies of the generation was actually passed by most of the tier one VCs. Is there like especially in the AI world, do you think that there is still there is some pattern matching or some something which can still work in terms of like, "Okay, these three things have to be there, and then this will be like a generational company." Yes, I do think there's still pattern matching, especially when it comes to founders. I do think that people's mental models have to are shifting a bit, but oftentimes we still are relying on things like the seven powers from a moat's perspective, and we're still trying to ask ourselves why now. And lastly, I think that a lot of investors are still looking for large gross profit pools. Now, this is where the mental model has to shift because historically, when you were investing in a company for instance like a ramp, what were they going after?

They were going after the profit pools of the credit card companies, you know, that were that were working with businesses, and they might have been going after like the accounting and bookkeeping software or spend management software. Today, I think the new thing that exists is like we are going after the profit pools that are labor and services companies. And like that is a completely different mental model, but why did people pass on Anthropic might be a question underneath the question that you're asking. Look, I can't speak for any of the people who looked at it, but back then, it was probably a hell of a lot easier to invest in Sam Altman, you know, coming out of YC and based on his network and to write a big check there than it was Dario, but I think the important piece that Dario, you know, had thought about really early on was just a different way of building the product from the ground up and having an enterprise focus initially. So, writing a check a $100 million check or whatever, I think it was like $400 million back then in Dario probably felt like a very risky bet for for VCs, and it was not consensus, but now it's become consensus. And I do think that oftentimes a lot of the best companies of any generation appear non-consensus at the very early stages, although the the risk is changing because we are writing checks at much, much higher valuations than we did before, but they they tend to become consensus over time.

Right. And it has become such an important company, right? That they almost have crashed multiple like verticals and industries in the last few months as we saw. And there's a school of thought that a lot of the SaaS-pocalypse is actually overstated, and it may be very difficult especially in certain industries to replace, completely wipe out traditional software. Do you think SaaS-pocalypse is like overstated? TBD.

I mean, I do think that if you just look at SaaS-pocalypse and like say it's everything, I do think it's overstated, but I do think there are some companies that are very vulnerable and others that are not. What I will say is that I believe in order for for Anthropic to hit the $100 billion revenue that they're projecting in a few years, they will still need to partner with companies that are deploying their technology. There's too much demand right now for Anthropic. They can't take deals that are under $10 million revenue. And so there is a massive pool of customers to go after. Like Anthropic's not going to hire a team to go focus on SMBs, but you know who does focus on SMBs is MSPs, like managed service providers.

Those are great distribution partners for something like Anthropic, and if you can build software for MSPs to deploy this into small businesses, I think that that's really compelling. I do think that that dynamic exists in many other markets, and so even though, you know, we're sitting here and we're calling vertical AI companies, you know, vertical software businesses, I think they're going to end up taking the shape more of like a Palantir or systems integrator or an MSP or consulting firm than they are the old way of thinking about software. That's actually a good thing. We can do like you know, a lot of listeners to this podcast are actually early stage companies or like, you know, first time founders, early stage founders. And I I think Sequoia recently released this you know, study or or an article about service as a software. And I think a lot of people have talked about it.

You have talked about it. I I've seen you talking about it on X. It looks like for last 10, 15 years, a lot of these founders were taught that you have to build things which are scalable, and because software is scalable, you write it once and you can sell it again and again at 80% profit margins. And then suddenly AI services seem so counterintuitive to all that. Can you explain like the rationale behind why AI services a great opportunity right now? So, let's just take some of the businesses that people have heard about before like in Accenture.

Um the Accenture is making billions of dollars of revenue off of AI deployments right now. And then there are other companies outside of consulting like law firms. And law firms are making a ton of money and, you know, have their old structure of doing things. So, there's two different examples, but what I'll say is like in Accenture has many, many business lines internally. And those business lines go after many different industries. And if you are a founder today and you are working on a vertical AI company to go out and try to attack one revenue line of Accenture, chances are it's very, very big.

And if you brand yourself specifically for that industry, you can go out and you can build a product like a Logora, you can deploy it into a company and start to build what feels like a personalized AI system internally for that business. And sure, the margins might not be as high in the early days, which by the way, all the numbers that I'm hearing on margins for forward deployed engineers are basically software margins. So, I don't know if that's valid argument, but let's say it was lower. If you're going in and you're building on a strong AI system of record for a business to operate on top of, a company likely is not going to turn off that thing. And so if you have a 10-year LTV, you should be willing to spend against that to acquire customers and to build the best product possible. So, that's one version of that.

With that said, the like other way of thinking about disrupting the services market is not deploying software to these companies and like building it out internally, it's just delivering outcomes. And so if I used to hire a marketing agency, and that marketing agency used to cost me two grand a month, and only, you know, 20 or 30 or 40% of that went to my ad spend, all of a sudden, you might be able to drop the cost by 50%, make it so 90% of my ad spend or my spend with the agency is going towards ad spend. So, the economics of every market are starting to flip, and the easiest way if I were a founder today to think about it is I would just go study what counter-positioning is. I would study what that is from the seven powers from Hamilton Helmer, and I would wrap your head around it. In any market, I would go look at, I would try to take a lens, and and literally you can ask Claude this cuz I do this all the time. How would you try to counter-position against this type of a market?

Very interesting point. And similarly, like a lot of startups are thinking about should they especially on the product side that should they be building and focusing on building one small thing, trying to solve one problem for a customer, or should they build an array of products or, you know, multiple things? Which also it never used to be a question, but today because building is has become so much easier, you know, for a particular customer segment, there are companies that are thinking of building multiple products, and doesn't seem like a very distant idea today. So, I don't know what your thoughts on that. I'll say this. If you look and you study Rippling, what was Rippling really good at?

It was building a compound startup with many products and features. And why is that a powerful model? It's a powerful model because one, you control the full stack, so the experience can be dramatically better for the end customer. But two, it provides you with probably the best business strategy of all time, which is bundling. And today, it is easier to bundle products than ever has been in the history of the world. You can build a compound startup in a matter of months, not a matter of years, and that unlocks many different pricing strategies and product opportunities than people had ever thought about before, and that then, you know, provides value propositions to end customers that I believe will get them extremely excited.

Great. Very interesting. This something really related to this you had tweeted about. I think it was Mecha base piece on how SAS can be saved and and you said like it's a it's a essential reading for anyone. What is it about you know the SAS survival question that you really liked from that Mecha base piece if you remember? I don't I don't remember exactly which which which piece that was.

I feel like I'm reading so much right now but when it comes to pure AI SAS right now, I think that and I think actually what that piece was really about was the shifting from old SAS to new SAS and when you're in board meetings today, I think the conversations with founders need to be completely different than they were 1 year ago. And the reason why is that if you start from scratch today, your business will be a completely different shape. You will have a completely different OPEX structure than you had before. You will have a different way of managing your engineering team and your product roadmap and your sprints and your sales team and marketing team. And so if I like I'm sitting down now with founders and they're not having conversations about how every department is using AI, I'm going to push on that. And I'll give you one quick example without using the name of the company but we had a company this week that did something very similar to Block and they laid off you know 40% of their staff.

Why did they do that? It was because a few weeks ago when they started to really lean into to Cloud Code and Codex, they realized that they don't need four people for implementations on a per customer basis anymore. They need one. And that they could actually not just try to customize their product in the current architecture the way that it was for each of the customers based on the asks and demands but they were actually just able to build a net new product every single time on top of their architecture and and you know foundation and as a result with one person only doing that, they can build a 10x better customer experience and they can move much faster for their end customers. And so I think that I'd say like maybe at least 70% of the companies that I work with are kind of on the back end of that shift but like there's another 30% still have a lot of catching up to do. And if you're not a founder looking yourself in the mirror every morning and figuring out how the hell do I catch up, I think that you probably need to be changing your morning routine.

Absolutely. You have to be a learning machine. Everyone actually not just founders. There's a high bar to you know to catching up to things every day. One last question Jason from my side that what was the last investment you made and why did you make it and how is it going? Yeah.

It's not announced yet but we're we're wiring the money here shortly for a company that's building an AI utility management business. And I had met the founder many years ago. He was an investor and I always thought he was like the smartest investor out there when it came to the market that I was looking at. And that is a that was a tough market by the way. It was proptech. And so one, he was incredibly intelligent.

Two, it was his second company. And then three, he experienced the pain point himself. So he went into the real estate market. He was managing a large portfolio of single family rentals and he realized the pain point when he was actually managing that portfolio for the family office he was working for. But so from a founder perspective, I loved it. And then from a pitch perspective, what did he share with me?

He shared with me that there was a company that was worth $5 billion and it had 70% market share. And what was it? It was a BPO. And what was that BPO's number one cost? Call centers. And if your call center cost is 70% of your COGS, that's really interesting because that means your pricing model was derived from your old cost structure which means that another company could come in and do what?

Counter position against you. And so if he's building an AI native version of that company that's worth $5 billion and he can now provide that product or service to the end customers for a fraction of the price, even if that other old company wants to become AI native, they're not going to drop their price by 50%. And so I love that type of a business and so if anyone's working on things to disrupt you know BPOs or companies with very bloated cost structures that could be disrupted with an AI native native way of operating system, I'd love to chat. That's a great example and great investment. Thank you so much Jason for sharing all these insights with us. Thanks for your time.

It was it was fun. Thank you. Thank you man. Have a good one.