Compiled by: Jiang Ye, AIGC News
Text: Summarized using Ali Tongyi Efficiency, thanks for the errata.
In YC's latest interview " Vertical AI Agents Could Be 10X Bigger Than SaaS ", four senior YC investors Gary, Jared, Harj and Diana started from the development history of the SaaS industry and combined with a large number of examples, deeply analyzed why vertical AI agents will become the next entrepreneurial trend.
As AI models continue to rapidly improve and compete with each other, a new business model is emerging: vertical AI brokers. In this episode of Lightcone, the hosts consider what impact vertical AI brokers will have on existing SaaS companies, which use cases make the most sense, and how there are $300 billion companies in this category alone.
1. The success of the SaaS industry is the best evidence for the rise of vertical AI agents
Jared believes that the market size of vertical AI agents will be very large, and may even give birth to companies with a market value of more than $300 billion.
He believes that the success of the SaaS industry is the best evidence for the rise of AI Agents in vertical fields . The emergence of the SaaS (Software as a Service) model has completely changed the software industry. In the past, companies needed to purchase expensive software licenses and spend a lot of time and resources on installation and maintenance. The SaaS model hosts the software in the cloud, and users only need to pay a subscription fee to use it, which greatly reduces the threshold and cost of using the software.
Jared believes that vertical AI Agent, as an emerging B2B software, has the potential to surpass SaaS in market size because AI Agent can not only provide software services like SaaS, but also automate operations through AI technology, further improving efficiency and reducing costs.
2. LLM technology lays the foundation for the explosion of AI Agents in vertical fields
LLM (Large Language Model) technology provides new possibilities for software development. It can combine software with manual operations to create more powerful vertical AI Agents, replacing traditional SaaS software and manual operations .
LLM technology can understand and generate human language and can be used to build chatbots, automatically generate text, translate and other applications.
3. Why did big companies miss out on B2B SaaS?
The main reason why large companies miss out on the B2B SaaS market is that the B2B SaaS market is highly fragmented, and each vertical requires deep expertise and focus on specific problems . Large companies prefer to focus on a few large markets rather than spreading their efforts across many segments.
Gusto is a SaaS company focused on payroll management. It has been successful because it has a deep understanding of the various details and regulations in the payroll management field.
For a giant like Google, developing a product like Gusto requires investing a lot of time and resources to learn and understand the knowledge in the field of payroll management, which is not cost-effective for them.
4. How will AI Agent affect the personnel structure of enterprises?
LLM applications change the employment model of startups. In the future, companies will only need fewer employees (or even one-person companies) to achieve rapid growth .
In the past, startups would typically expand their teams as their business grew, but LLM can help companies automate and reduce their reliance on manpower.
In terms of AI interviews, Triplebyte is a company that recruits software engineers. Its software can automatically screen resumes, conduct technical tests and preliminary interviews, greatly reducing the workload of recruiters.
5. What is the market potential of vertical AI Agents?
The market size of vertical AI agents will be ten times that of SaaS, because they can not only replace existing SaaS software, but also replace a large number of manual operations .
Traditional SaaS software still requires manual operation to complete many workflows, while vertical AI Agents can combine software and manual operations to achieve higher efficiency and lower costs.
Momentic is a company that uses AI technology for QA testing. Their AI Agent can automatically execute test cases and generate test reports, thus completely replacing the traditional QA team.
6. Application cases of AI Agent in vertical fields
Four senior YC investors cited examples of AI agent companies in multiple vertical fields.
Outset: Improving the field of surveys and questionnaires.
Traditional survey and questionnaire software requires manual effort to design questions, collect data, and analyze results.
Outset’s AI Agent can automate these tasks, adjusting questions and answers in real time based on user feedback, thereby improving the efficiency and accuracy of surveys.
Powerhelp: Handling complex customer support workflows.
Traditional customer support requires humans to answer calls, respond to emails, and resolve issues.
Powerhelp’s AI Agent can automatically complete these tasks and provide personalized solutions based on the user’s questions and history, thereby improving customer satisfaction and efficiency.
Salient: Automating debt collection in the auto loan space.
Traditional collection work requires manual phone calls, communication with borrowers and recording of collection results.
Salient's AI Agent can automatically complete these tasks and adjust collection strategies based on the borrower's situation and repayment ability, thereby improving collection efficiency and success rate.
7. AI Voice Call Technology
AI voice calling technology has developed rapidly in recent years. With the advancement of AI speech synthesis technology and natural language processing technology, AI voice calling can be used in more complex scenarios, such as debt collection, customer service, marketing, etc.
AI speech synthesis T2V technology: converts text into fluent speech, allowing AI Agent to communicate with users like a real person.
Natural Language Processing (NLP) technology: enables AI Agent to understand the user's intentions and emotions, and respond accordingly based on the user's questions and feedback.
8. How to choose the AI Agent entrepreneurial direction that suits you?
Jared suggested that founders who want to start AI Agent companies should look for boring, repetitive and administrative jobs. These jobs usually require a lot of manpower and can be easily replaced by AI technology.
For example, Sweet spot saw that there was a lot of repetitive work in the government contract bidding process, so it developed an AI agent to help companies automate these tasks.
Interview Translation
Opening
Every three months, it gets incrementally better. Now we’re talking about fully vertical AI agents that are replacing entire teams, functions, and businesses. That progress is still exciting to me. A lot.
The base model is a head-on approach. There used to be only one player in town, OpenAI, but that has been changing in the last batch.
Thank God. It's like competition is a very fertile soil for a market ecosystem where consumers will have choice and founders will have opportunity. That's the world I want to live in.
Welcome to another episode of Cone of Light. I'm Kaijie (Gary) Yu. This is Jared Harge and Diana, and together we've funded startups worth hundreds of billions of dollars when they were just one or two people starting out. Today, Jared is a man on fire, and he's going to talk about Vertical AI Agents.
Passionate about vertical AI Agent
I'm really happy about this because I think people, especially startup founders, especially younger founders, don't fully realize how big vertical AI agencies are going to be. This is not a new idea. Some people are talking about vertical AI agencies, we've funded a lot of them, but I don't think the world has realized how big it's going to be. So I'm going to make the case for why I think there will be over $300 billion companies founded in this one category, which is great.
I'll do this by making an analogy with SAS, and I think in a similar way, people don't understand how big SaaS is because most startup founders, especially younger founders, tend to view the startup industry through the lens of the products that they use as consumers. As a consumer, you don't tend to use that many SaaS tools because they're mostly built by companies. So I think a lot of people are missing the basic point, which is if you look at what Silicon Valley has funded the most over the last 20 years, it's like we've primarily produced SaaS companies, guys, like most of what's come out of Silicon Valley. Over 40% of all venture capital funding in that period has gone into SaaS companies, and we've produced over 300 SaaS unicorns in 20 years, which is more than any other category.
Software is awesome. I was thinking back to the history of this thing, because we always like to talk about how the history of technology influences the future, and what the real catalyst for the SaaS boom was. This was a, do you guys remember the XML Http request? Oh my goodness, I would argue that was really the catalyst for this ass explosion.
Like Ajax, yes. In 2004, browsers added this JavaScript function XML http request, which was the missing piece that enabled you to build rich internet applications in a web browser. So for the first time, you could make things on the web that looked like desktop applications, and then Google Maps and Gmail were created, and like SAS built this whole thing. Essentially, the key technology of lock-in was that software went from something you got off a disc and installed on your desktop to something you used through a website and on your phone.
Paul Graham actually shares that lineage because he was one of the first people to realize that he could take an Http request and then actually connect it to a Unix prompt. And you didn't actually need to, you know, have a separate computer program to change a website. So through the web was an online store, kind of like Shopify. But in the old days.
It was basically like the first SaaS application ever, just like Pg actually invented SaaS in 1995. It's just that those first SaaS applications were kind of terrible because they didn't have XML Http requests. Every time you clicked a button, it had to reload the entire page, and it was just a hard experience. So it didn't really catch on until 2005 when XML Htp requests became white, right? Anyway, I think this lum thing is actually very similar. It's like a new computing paradigm that made it possible to do something fundamentally different. And in 2005, when cloud and mobile finally took off, there was a big open question of, like, okay, this new technology, what are you supposed to do with it?
Similarities between Traditional SaaS and LLM
Where will value accrue and where are the good opportunities for startups?
I was going through the list of all the billion dollar companies that were created, and I kind of realized that you can break down the different paths that people take into three buckets, and this is the first bucket that people start with, like, I'll call them obviously good ideas that could become mass consumer products. So it's like documents, photos, email, calendars, chat, all these things that we used to do on the desktop but obviously can move to the browser and mobile. And what's interesting is that there's one category in the startups where 100% of the value flows to the incumbents, right? Like Google Facebook and Amazon, they own all these businesses.
People forget that, like Google Docs wasn't the only company that tried to bring Microsoft Office online. They were like 30 companies that tried to bring Microsoft Office online and they all lost to Google 1. And then there's the second category, which is like the mass consumer ideas that weren't obvious, that no one predicted. It's like Uber, Instacart DoorDash, coinbase, AirBNB, these, the ones that came out of left field, like the dot, dot between XML Http request and AirBNB, it looked very non-obvious. So the incumbents didn't even try to compete in those areas until it was too late. So the startups were able to win there.
And then there's a third category, which is all the B2B SaaS companies, which is about 300 of them. Just like the number of logos, there are more billion-dollar companies created in the third category than in the first two categories. I think one reason this happens is that there's no SaaS company like Microsoft, there's no one company that's like SaaS in every vertical for every product, because of the structure. It seems like all the different companies are the same, and that's why there are so many of them.
I think Salesforce was probably the first true SaaS company, and I remember Marc Benioff came to YC and spoke, and he said the story was that very early on, people just didn't believe you could build complex enterprise applications through the cloud or SaaS. So it was like a perception problem, right? It was like, no, just like you don't buy your boxed software, that's like real software.
You always do. It's pretty hard to do this because early web apps were terrible over the web, and you have to be as visionary as Pg and understand that browsers are going to get better and better and eventually it'll be good.
It's very reminiscent of the same thing today, oh no, you won't be able to build complex enterprise applications that use these llum or AI tools because they're hallucinating or they're not perfect or they're kind of like toys, but yeah, that's like the story of early SaaS.
So when I think about the parallels with OLENS, it's easy to imagine the same thing happening, which is that there's a bunch of categories like mass consumer applications that are obviously huge opportunities, but probably the incumbent companies will win all of them, so it's like a general AI voice assistant that you can ask it to do anything and it'll just like do that one thing that obviously should exist, but like all the big players are going to compete to be that thing.
Right? Apple is a little slow in that area, why sir, so stupid, it still has no perception.
I mean, there seems to be a counterpoint to that, which is the obvious thing is search, and maybe Google will still win on search, but confusion will certainly give them a run for their money.
Yeah, and that's the classic innovator dilemma that ends up in your life. I mean, you can counter your argument about Uber or AirBNB, those are actually very dangerous things from a regulatory perspective, so if you're Google and you have essentially a guaranteed, you know, giant pot of gold coming to you every month, why would you jeopardize that pot of gold to go after these things that might scare people off or might destroy it?
I think that's probably the main reason that the incumbents didn't end up building these products, or even cloning them after they got big, it was clear that they would work hard. Google never launched a U clone, they never launched an AirBNB clone. I was listening to this talk by Travis, and one of the things he said that really bothered me was that in the early years of Uber, he was very afraid that he was going to pursue, go to jail for a long time, like he was actually risking jail time to build that company. So yeah, no highly paid Google executive would do that.
Why big companies can’t do B2B SaaS
Why do you think the big guys haven't gotten into B2B? Is SaaS partly to blame? A lot of the use cases are very widespread?
Is that a good question? I love hearing what you guys think. My take on it is that it's just too hard to do so many things as a company, just like every B2B SaaS company needs to have like people running product in the business who focus very deeply on one area and care very deeply about a lot of very nebulous problems to act on with gusto. For example, why didn't Google build a serious competitor? Well, nobody at Google really understands payroll and has the patience to deal with all the nuances of all these stupid payroll regulations. It's like it's not worth it to them. It's just easier for them to just focus on a few very large categories.
In the BCB SaaS world. It's kind of the software unbundling vs bundling debate that I think comes up a lot as well and it's expanded. Why are all these vertical B2B SaaS products evolving and not like Oracle or Sap or Netsuite? Yeah, Netsuite is like having everything. I think that's probably another thing that's also caused by the move to the old way of selling software like SaaS and the internet.
Again, like you had this box software, which was really expensive, and you had a whole ecosystem around it. Any time you wanted something customized, like the integrators would say, oh no, like we can build a UA customization like a payroll feature or something. And then Salesforce came along as like a SaaS solution that never seemed to be as powerful or as sophisticated as the expensive enterprise installation that you just paid for. But they proved to be exactly that. I think that just like opened the door for everybody. These are like vertical SaaS solutions that do exactly what you said they would.
The other problem is that for a lot of enterprise software, if you're an Oracle and Netsuite user, because they have to cover so many areas, the user experience is actually pretty bad. They're trying to be a jack of all trades, but a master of none. So it ends up being a kitchen sink experience. That's where if you go and build a B2B SaaS vertical company, you can provide a 10x better experience and a more delightful experience because there's a clear distinction between the consumer product and the enterprise user experience.
Software only has three price points? $5 per seat, $500 per seat or $5000 per seat. That maps directly to consumer SMB or enterprise sales. And then I think time immemorial taught us that, and thankfully that's becoming less and less true for new software. But enterprise software is bad software because it's not bought by the user, you know, some high level muki dmack in a fortune 1000 company who eats and drinks this big 7 figure contract. And you know, they're going to pick something that might not be that great for the end user, the person who actually uses the software every day. I'm kind of curious to see how that changes in the LMS.
I mean, one of the things that we've seen more prominently in SMB and enterprise software companies so far is that all software companies, all startups, period, like, you know, as revenue scales, there's a perception of the number of people you have to hire relative to its size. And so when you look at unicorn companies, even in the YC portfolio today, it's very common to see a company that's doing $100 million or $200 million in revenue a year. They've got like 501,000 employees. I'm just very curious, like as I start giving advice to companies that are a month or two old, it feels a little different than the advice I would have given last year or two years ago.
In the past, you might have said, let me find the absolute smartest person in all the other parts of the organization, like customer success or sales or something like that. I want to find someone who I've worked with before and I know is great. And then I'm going to go sit on their doorstep until they quit and come work for me. I want them to be able to build a team for me and hire a lot of people. That may still be true, but I'm starting to sense this, the meta-shift is kind of like you actually might want to hire more really great software engineers who understand large language models and can really automate the specific things that you need that are the bottlenecks in your development. So this could lead to, you know, a very subtle, but, you know, significant change in the way startups grow their businesses, kind of post-production market adaptation. It means I'm going to build LM systems that lower my costs so that I don't have to hire 1,000 people. I think we're at the beginning of that revolution right now.
I mean, we talked about this in the previous episode, we talked about how there's going to be a unicorn company in the future that's only going to be able to operate with only 10 employees. That makes perfect sense.
They are writing evals and prompts.
I think what you're talking about is like a trend that was already starting before lum. Like I remember when I was running Triple Byte, for example, we needed to like build out marketing or user acquisition. Especially after we raised our Series B, the traditional way you should do it is hire a head of marketing, build out a marketing team, like start this machine to do sales and marketing.
But I actually met a guy who was like Mike, the founder of YC, whose company was basically building a smart frying pan. Sounds weird, but like he was an engineer from MIT. Yeah, you remember this? He was an engineer from MIT. And to sell a smart frying pan, he had to be very, very good at understanding paid advertising and Google ads and a bunch of stuff. So he took an engineer's mindset, and I remember I just talked to him about this, and I realized that if there was an MIT engineer working on our marketing, that would be much better than any marketing candidate I had ever talked to. He was able to scale us to the point where we were spending about a million dollars a month just on marketing and various events.
In the triple of great marketing, like I remember, like the Caltrain station takeover, you did all the things you do at home. It was like really high quality stuff. It stuck with it. You could tell he wasn't done by marketing like that.
That guy, that was all mic, like the comment I always get when people ask me at that time, like how big is Triple Byte? We were like 50 people and we did better. It was like hundreds of people. And I was like, no, it's all because if you take a really smart engineer and do something like that, they just figure out how to get it done, they find the leverage, and now like LLC, the leverage that can go beyond pure software.
Okay, this is the talk I gave for the 300 vertical AI agent unicorns. Really, for every company that's been assessed as a unicorn, you can imagine there's a vertical AI unicorn equivalent, just like some new universe like most of these SaaS unicorns, there are some companies like box software companies that are building the same thing that got disrupted by SaaS companies. You can easily imagine the same thing happening again. We have basically every SaaS company now that builds some software that is used by some people. The vertical AI equivalent is just software plus a person in a product.
One thing is probably just that enterprises are generally a little bit unsure right now about exactly what they like, what agents they need. I've seen one approach come from particularly experienced founders like Brett Taylor, who was Facebook's chief operating adviser, who started this company. I don't know all the details, but from what I understand, it's basically more about having enterprises deploy these AI agents custom-tailored to the enterprise, rather than like, "Oh, hey, we have this specific agent that does this."
This is something I've seen from a company of mine called Vector Chef, which got funded about a year ago. They're two really smart Harvard computer scientists who figured out that they were trying to build a platform that would make it easier for enterprises to build their own, like using no code or SDKs to build their own internal lum-powered agents. But enterprises often don't know what they want to do with this stuff. So bringing it back, I'm wondering if like in the box software world, you start out with like a few vendors who are basically just trying to convince people to use the software. Like, it can do anything, and then it gets more complex and higher resolution, and you get a lot of players that are like vertical SaaS players. We went through the same period with llds, where the early winners were probably just these generic, like we like make it easy for you to do LLD stuff. And then the vertical agents will come in over time. Or do you think what's the reason that it's different now, that the vertical agents will take off on day one?
Yeah, it's interesting because if you think about the history of SaaS, initially like 2005 to 2010 it was mostly consumer applications like email, chat, and maps. People were used to using these tools themselves. I think that made it easier to sell SaaS tools to companies because, you know, the same people were both employees and consumers.
Yeah, I think the answer is probably like, it's just a continuation of software, there's just no reason it has to reset, just like llms don't have to reset back to some general purpose, like enterprise llum platforms, doing everything that the enterprise has been trained to do, like the value of point solutions and vertical solutions and like user experience won't be that different. Those things will just be more powerful. So if the enterprise has built enough power that they believe a startup or a vertical solution can be better than a traditional broad platform, they might be willing to bet on a startup that promises to deliver a really good vertical AI agent solution today. I feel like we're all seeing that right now with some of our companies getting traction in the enterprise with these vertical AI agents faster than we've ever seen before.
I think we're just early in the game, right? Like all software, it starts out pretty vertical. And then as the industry actually becomes more developed, I mean, I just answered my earlier question, which is like, you know, why does a company end up with 1,000 employees? And actually, you know, early in the game, everybody's making these specific point solutions. And then at some point you have to go horizontal, like you've spent like crazy on sales and marketing. And then once you get to 100%, or you know, the vast majority of the market, the only way you can actually continue to grow is if you actually do not just a point solution, but something that's common.
Maybe another reason why the bold case for vertical AI agents could be bigger than SAS is that SaaS still requires an operations team or a group of people to operate the software in order to do all the workflows. I don't know about approval workflows, or you have to input data. The argument here is that not only can you replace all of the SaaS software, so it would be a 1-to-1 mapping, but it would also eat up a lot of payroll because you find a lot of companies where a large portion of these expenses are still payroll and the software is very small.
To be exact, they spend much more on their employees than on their software.
As a result, these smaller companies are more efficient and require less manpower for random data entry, approvals, or clicking through software.
I agree, I think the vertical equivalence is likely 10x that of the SaaS companies they are disrupting.
I mean, there are two scenarios. The vertical point solution might be large enough that you don't need to do the broth breathing thing, right? That might be a nice scenario.
Examples of AI Agent vertical applications
Should we give some examples? I feel like we've all had the pleasure of working with so many verticals. AI agency companies, we have news from the front lines of how it's actually developing.
Here's an example. Aaron Cannon is working at stray', a YC company I work with, and basically they're bringing LDS to the survey and Qualtrics space. So Qualtrics is almost certainly not really building the best large language model with inference capabilities.
And then the interesting thing about surveys is, you know, who is it actually for? It's for the people who run the product. It's for the marketing team, it's for the people who are trying to sense what our customers really want and what do the survey results look like? Guess what, that's the language. So, I feel like these types of businesses actually have to thread the needle because enterprise and SMB software is often sold to specific people who are key decision makers. You have to go high enough in the organization so that the people you're selling to aren't scared that their entire job or their entire team's job is going to completely go away.
This is a move I see a lot of sales companies need to take because if you’re selling to a team, that’s going to be replaced by AI.
So I think it's an interesting way that a lot of it is top down and you have to get it through at some point and you can get the CEO to sign off on it.
The company I'm working with is Mee, which is essentially an AI agent, but at least where they started is like QA testing. They're getting really good traction now.
It's interesting because if you remember ten years ago, why companies that we work with like Rainforest QA, like Rainforest is a QA services company, they had this exact tension of, you can't really replace your QA team. So they needed to build software that made QA more efficient. But in reality, that obviously meant trying to replace them as much as possible. They couldn't replace the entire team, so they were always walking this tightrope between trying to sell software to engineering heads like that means you need fewer QA people and great, but you also have to sell it to QA teams that don't want to be replaced. So I think that's always been like a friction in how that business can scale and grow, but now meics like AI can actually replace QA people. So their pitch wasn't, oh, this makes your QA people faster. It's like it means you don't need a QA team at all. So they can focus the cells on engineering, which engineering currently doesn't need to buy from QA, and you can also go in, I mean, first of all, you can go sell to companies that don't even have large QA teams currently.
They just use something like Metic and then it's like, scaling and scaling. They'll never build an 18Q. This is a real case study of why these vertical AI agency companies can be 10x bigger than SaaS companies.
I see this now and it's interesting, like in recruiting, I have the exact same problem in building software, to build software that makes it easier to screen and hire software engineers, you need buy-in from the engineering team that they join, but also the recruiting team. In fact, the software that we're building tries to replace recruiters, but we can't completely replace recruiters. But now there is New York.
So recruiters always push back, fight it because it’s a threat to them.
Yeah, so there's always like friction, like how far can you get when the client you're trying to sell to is worried about being replaced. But, I think it's still early days, but now with AI, you can build things that do the entire stack, like recruiting. We had a company that we worked with last patch like Nico that actually just does the full technical screen, the full initial recruiter screen, and that got a lot of traction. So I think as these things go on, like they won't have the same things, you won't have the friction. Even though I need to convince recruiters to use this, you might not be building recruiting teams the same way you did before.
I mean, another example is even for dev tool companies, they have to do a lot of developer support. I work with a company called capillo AI that basically built one of the best chatbots that can respond to a lot of technical details that are hard to answer. I think a lot of the companies that started using them, they actually had a much smaller devrel team than I did because there was just a lot of developer documentation and even YouTube videos that the dev tools released, and even a lot of chat history. So it just kept getting better and better, like giving really good answers. One of the best I've seen actually.
Yeah, I also worked with a customer support company, like an AI customer support agent company called Power Help. Well, actually, we both did the last batch, and I learned a couple of interesting things from Power Help. The first is the customer support category like AI agent is like notoriously crowded, there's supposedly 100, if you go to Google like AI customer support agent, you'll get 100 results on Google. But what I learned from working with Power Help is that it's actually kind of bullshit, like almost all of these companies do very simple, like llum tips with zero data can't really replace a real customer support team that does a lot of very complex workflows. It's just a nice demo that actually replaces a customer support team at, like, a scale company with about 100 customer support reps doing a lot of complex things every day. You have like very complex software that can handle all of the things like Jack Heller was talking about, but there's only three or four companies trying to do that. And, Curry, they cumulatively have less than 1% market penetration. So the market is completely open.
I can also see that this is another case of hyper-specialization or hyper-verticalization, like there won't be, I mean, maybe eventually there might be a general purpose customer support agent software company, but we're like in a win, you know that's like the eighth inning or the ninth inning kind of thing. We're really in the first inning. So, you know, instead, you know, you're going to have companies like Giga ML, you know, it does 30,000 tickets a day for zepo and replaces a 1,000-person team, but it's very specific, you know, it's not a general purpose demo where there's a very detailed eval set with 10,000 test cases in it, you know, it's basically just 4 zepo and zepo and stuff like that. But if you were, you know, any other market company, you would probably use it because it's a very well-defined market, you know, the instant delivery market.
I think that's the dynamic that leads to like $300 billion SaaS companies out there, as opposed to like the $10 trillion meta SaaS that's providing all the software stuff to the world, where it's hard to build a company that's like engineering for all of them, where customers need very customized solutions.
Exactly, we've given three examples of customer support, but there are very different verticals, like a developer tools company, where you need to be very different from the market in your training set, very different, right?
Yeah, I guess whether you have agents or real people working for you, you end up with the same problem, which is that every company violates the COS theory of the company, which is that any given company will only grow to the point where it becomes inefficient to be bigger than that. That's why they're networks and ecosystems, and you, a mature economy. Every company you like specializes in doing something that it's particularly good at. And then the limits, the external limits of those companies, are really based on your capabilities as a manager. So, that part kind of makes my heart ache because you know, when we were with Parker Conrad, one of his favorite things was actually, you know, everybody was really obsessed with the fact that rocks can talk, you know, maybe they can draw. But what was more interesting to him was running HR IT software, you know, he spent a lot of time thinking about HR, and actually the coolest thing about LMS was that rocks could read, from his perspective, like he had 3,000 employees, he still ran payroll for all 3,000 employees through Ripple. So I think he spent a lot of time thinking about, how does one expand their capabilities as a manager? And, I think we're going to see more of the opposite view, that the tools for managers and CEOs are going to become more powerful.
It can increase the size of the company you run, right? That's certainly what Ripple is trying to do, just like he's trying to build this HR tool suite where if he wins he's going to eat a bunch of billion-dollar SaaS companies in one giant company.
Very interesting point Gary. I think what got me thinking is that having all these artificial AI tools will allow all these leaders and all these organizations to basically open up the aperture of the context window or how much information they can parse because there is a limit to how many we can have meaningful relationships with. The Delmar digital thing is like the whole thing, you can go around 300 people, you can have meaningful relationships with 150 people. But with AI, because all these rocks can now be read, I think we're going to be able to expand the Dunbar limit that we have.
Yeah, I think there was an interesting post by Flo crevello on Twitter that went viral. I think someone was using voice chat as a CEO, like a weekend project, but it would call all 1,500 of their employees, you know, and it was very short calls that sounded like it was from the CEO, just asking in person. I mean, it kind of reminded me of that scene of hers, it zoomed out, actually, you know, you're following the experience of one person using her operating system, but her operating system is actually talking to 15 people, you know, thousands or tens of thousands of people at the same time, and many other 8,316. Yeah, I mean, large language models can speak and can have a conversation, so to what extent does knowing that capability actually scale the ability of one or a few people to understand what's going on?
I hear you, and it definitely got me thinking because from my understanding of this project or projects like it, it would just call up all your employees and then your employees would just stroll through what they were doing and it would just extract meaning from it and give the CEO a bullet point summary of what was most important to him.
There are also a bunch of SaaS-like companies trying to run weekly pulse operations by having employees use similar traditional SaaS software, but this version is literally 100x better than the pre-LM version of this idea.
But I wonder if that particular tool, like it's not, it goes beyond just reading and summarizing like this. That's the argument that if writing is thinking, then there's actually a ton of work involved in like figuring out who is effective in communicating, like, what are the most important things, what are the key things to focus on as a company? I just wonder if at some point lds like go beyond summarizing and reading and actually thinking, at this point, interesting thinking like who actually runs the organization.
I guess one of the other interesting things in Parker Conrad's mind, and this is something I discovered recently in an interview with Matt McGinnis at Proof of Origin, is that there are over 100 founders working at Rippling right now, and they are specific people that operate vertically within Rippling like the entire SAS.
The way he builds his team is really cool. Hal probably knows a lot about this because you've been interviewed a lot.
Yeah, I mean, it's definitely very focused on recruiting founders. I mean, Ripple like it's essentially going against the grain of what's going on vertically. And trying to take over all of HR and IT software horizontally.
Like the whole thesis is there's basically an underlying platform that has tremendous value and he wants to recruit founders and teams that build on the platform, just like Amazon's squeeze, like shared infrastructure.
Yeah, I think every product that they release, I mean things like time tracking, I mean, basically they release one thing and it goes to millions of dollars on the first day of release. That's exactly what we were talking about earlier. It's like once you, once you have a vertical, once you have a foothold, you mean, I have to spend this money on sales and marketing no matter what, can I? You can basically get a higher LTV and keep my CAC constant, and if you look at all the top software companies today, it's like Oracle, Microsoft, Salesforce are rippling. Knock on wood will be the next one, but it's an interesting alternative to go from 0 to 1 completely on your own.
AI Voice Technology (Company)
For you, I want to talk about some of the voice companies that we own, which I think is an interesting subcategory of these things that are really exploding right now.
I work with a company called Salient, which basically automates the collection and auto-closes the space through AI voice calls.
Traditionally, they would call people and they would say, hey, you owe a thousand dollars on your car, and yeah, actually you do.
The job is one of those butter-pass jobs. It kind of sucks because a lot of low-paid workers are working in all these call centers, and it's like a horrible, boring job. There's such a high turnover and a huge headcount to run these banks because these banks have so many accounts, and it's a perfect task that AI can automate. And what Salient does is it's able to get very, very accurate results. And it's already live at a lot of the big banks, which is very exciting. This is a company that's been around for the last year and proved that part of the reason they were able to get in is because they sold through a top-down approach.
I guess the space feels like it's moving really fast, we have some incredible companies like voice infrastructure companies like vappie. And then people can start right away. And retail, I mean, those companies have reached scale pretty quickly just because it's one of the more exciting, like, exciting things that you can get up and running in. I mean, literally the process of time.
And then some of the questions, you know, that are still unanswered, that we hope they figure out is, how do you hold on to that, especially when you have something like the new OpenAI speech APIs? You know, do you just go straight ahead like you did? It might be more work to try to use the underlying APIs offline. But these platforms are clearly low. And then the question is, can you keep raising the ceiling so that you can hold on to customers forever?
Heart, you brought up an interesting point earlier about how the applications that people build on LMS have changed from the beginning of 2023 to now.
The voice thing that we just talked about is a great example. I think even if you go back six months ago, it felt like those voices weren't real enough. The latency was too high. And just like that, it felt like we could be a way to have AI voice applications that could meaningfully replace, like, humans calling people. Like, we're there. Yeah, it just kind of zooms out, thinking back to the first YC batch where llen-driven applications first appeared was probably winter 2023, you know, almost two years ago. And those applications were basically just spewing out some text stuff, and it wasn't even like it was perfect.
It's more like copy editing, marketing, editing, email editing. It's just a more incremental approach. Yeah.
Like I had a company, I mean, the one that stuck in my mind was called Fast Brands, and what they did was make it very easy for small businesses to generate a blog and spit out content marketing. It was a very obvious idea that wasn't perfect, but it was cool at the time. And that's what we talked about throughout the show, but like the ChatGPT rapper said at that time, hey, this is what a big language model application looks like, it's just a wrapper around ChatGPT. It does a very basic thing, it spits out some text, and it's like it's going to get crushed by OpenAI in the next version, and it did.
Yeah, I don't know about that one, but the first wave of lum apps were mostly crushed by the next wave, GPT.
But I feel like we've had this bullying frog effect where, from our perspective, it was like every three months, things would incrementally get better. And now we're at a point where we're talking about full-blown vertical AI agents that are replacing entire teams, functions, and enterprises, and that progress is still exciting to me, like two years in, it's still relatively early days, and the pace of progress is like unlike anything we've seen before.
I think what's interesting is, and we talked about this in the last episode, a lot of the base models are kind of hitting head-on. It used to be that there was only one player in town, OpenAI, which we saw in the last batch. That's been changing. Claude is a huge competitor.
How founders choose the vertical field of AI Agent
Like competition is a very fertile soil for a market ecosystem, consumers have choice, founders have opportunity, and that's the world I want to live in. So people are watching and thinking about starting a startup, or maybe already started, and they hear all of this. How do you know what's the right vertical for you?
You have to find some boring, repetitive administrative work somewhere, and that seems to be the common denominator for everything, if you can find a boring, repetitive administrative task, there’s a good chance there’s a billion-dollar AI agent startup out there if you keep digging deeper.
But it sounds like you should pursue something you have some direct experience or relationship with.
That's very common. That's definitely a commonality that I've seen in companies that I've seen commitments to, another one that just came to my mind, I think I mentioned it before, they're basically building an AI agent to bid on government contracts. The way they found the idea is, this was a year ago when they just had a friend who was working full time sitting there, like on a government website, like refreshing the page, like looking for new bid proposals, and they were turning to, and they were like, this seems like something that a lum could do. A recent batch of companies turned to a new idea that got a lot of traction, like they're basically building an AI agent to process like medical bills for a dental office. The way they found the idea is one of the founder's mothers is a dentist. So he decided to work with her for a day and sit there and see what she did. And she said, oh, like all of this, like processing claims seems really boring, like a lum should be totally able to do that. And he just started writing software for his mother's dental office.
So I guess, I mean, in the robotics space, the classic adage is, you know, the robots that are profitable and that work are going to be the dirty and dangerous jobs. In this case, for vertical SaaS, look for boring butter delivery.
