
What this covers
This Lightcone episode has the hosts, including Jared Harge, Gary and Diana, argue by historical analogy. They treat LLMs) as a new computing paradigm, much as cloud and mobile were in 2005, and ask what that implies for startups. The opening stretch of the episode is a history of SaaS): Paul Graham's Viaweb in 1995, why it struggled without XMLHttpRequest, and how AJAX in 2004 became the catalyst for the boom. Salesforce comes up as the first true SaaS company, and the hosts discuss the skepticism it faced about whether sophisticated enterprise software could run in the cloud.
From there the conversation looks at who wins in each wave. The speakers claim that obvious consumer ideas such as docs, email, calendar and chat went entirely to incumbents like Google. Non-obvious ones like Uber and Airbnb went to startups. The reason incumbents stayed out is contested: regulatory risk and the prospect of personal prison time for founders are offered as the explanation. They also argue there is no "Microsoft of SaaS," since each vertical demands its own domain knowledge. They cite more than 300 SaaS unicorns and over 40% of venture dollars in twenty years.
The later sections cover headcount and examples. The hosts note that SaaS firms at $100-200M ARR often carry 500-2,000 employees, and they discuss whether agents change that, with Coase's theory of the firm as a limit that applies to human and AI workforces alike. They also describe three durable software price points of $5, $500 and $55,000 per seat. The examples centre on repetitive administrative work: Salient in collections, Outset in surveys, and MurmurAI in QA testing. A separate stretch covers voice-calling companies and infrastructure such as Vapi, and the episode closes with advice on choosing a vertical as a founder.
Vertical AI agents will create $300 billion+ in new company value by automating specific business functions more effectively than horizontal SaaS, following a similar pattern to how SaaS disrupted enterprise software but with greater leverage through replacing entire teams rather than just software.
- Every major SaaS category will have a vertical AI equivalent that's potentially 10x larger because AI replaces human workers, not just software
- Boring, repetitive admin tasks are the optimal target for vertical AI agents, as demonstrated by companies like Salient (collections), Outset (surveys), and MurmurAI (QA testing)
- Vertical solutions win because no single horizontal platform can deeply understand every domain's needs, a lesson already proven by the 300 SaaS unicorns that emerged instead of monolithic enterprise software
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
Paul Graham invented the concept of SaaS in 1995 with Viaweb (a web-based store, predecessor to Shopify), but early SaaS had poor user experience because XMLHttpRequest didn't exist yet, forcing full page reloads on every interaction, which prevented adoption until 2005 when AJAX became widely available.
“Paul Graham actually uh shares in that lineage in that he was one of the first people to realize that he could take the HTTP request and then actually hook it up to a Unix prompt and you didn't actually have to you know have a separate computer program that would change a website so via web was a online store kind of like Shopify but way back in the day yeah it was basically like the first SAS app ever like like PG actually invented SAS in like 1995”
There is no 'Microsoft of SaaS'—no single company that dominates all SaaS verticals—because each vertical requires deep domain expertise and understanding of obscure issues specific to that domain, making it structurally difficult for any one company to do everything well.
“there is no like Microsoft of SAS like there is no company that somehow does like SAS for like every vertical and every product like for structural reasons it seems to be the case that like they're all different companies and that's why there so many of them”
Paul Graham invented SaaS in 1995 with Viaweb, an early online store builder that used HTTP and Unix prompts, but it flopped because browsers lacked XML HTTP request, forcing full page reloads with every action, creating a poor user experience until 2005 when AJAX adoption made the paradigm viable.
“Paul Graham actually uh shares in that lineage in that he was one of the first people to realize that he could take the HTTP request and then actually hook it up to a Unix prompt and you didn't actually have to you know have a separate computer program that would change a website so via web was a online store kind of like Shopify but way back in the day yeah it was basically like the first SAS app ever like like PG actually invented SAS in like 1995”
Coase's theory of the firm predicts that any given firm grows only to the point where it becomes inefficient to be larger, after which networks and ecosystems emerge, and this dynamic applies whether a firm uses human employees or AI agents to do the work.
“I mean we already gave three examples of customer support but there are very different verticals it's like de tool comp need very different kind of support that you need and the training set to marketplaces very different right yeah I guess whether you have agents or real human beings working for you you end up with the same problem which is every company bumps up against CO's theory of the firm which says that any given firm will grow only so much to the point where it uh becomes inefficient to be larger than that and then that's why they sort of networks and ecosystems and you know a full blown economy you know like every firm will sort of specialize to do what it is particularly good at and then the limits the outer limits of what those firms can be it's actually based on uh your ability as a manager”
LLMs represent a new computing paradigm similar to how cloud and mobile computing were new paradigms in 2005, creating an open question about where value and opportunities lie for startups, with subsequent massive value creation across multiple company categories.
“I see this llm thing as like actually very similar um it's like it's a new Computing Paradigm that makes it possible to just like do something fundamentally different and in 2005 when cloud and mobile finally took off there is this sort of like big open question of like okay well this new technology exist what should you do with it where is the value going to acrw where are the good opportunities for startups”
Salesforce was probably the first true SaaS company, and the core perception problem early SaaS faced was that people didn't believe you could build sophisticated enterprise applications via the cloud/SaaS, despite early pioneers like PG showing it was possible.
“I think Salesforce is probably like the first true SAS company um and i'd I remember Mark benov coming to speak at YC and he tells the story as just very early on people just didn't believe you could build sophisticated Enterprise applications like over the cloud or via SAS it was just so um there was just like a perception issue right”
Travis Kalanick, Uber's founder, was personally scared he would go to prison for building Uber in its early years, showing that regulatory risk was real and personally significant for founders pursuing these opportunities.
“I was listening to this uh talk by Travis and one of the things that he said that really stuck with me is that in the in the first years of uber he was very scared that he would was going to personally go to prison for like a long time like he was actually personally risking going to prison in order to build that company”
Whether you have AI agents or real human employees, every firm bumps up against Coase's theory of the firm, which states that any firm will grow only to the point where internal coordination becomes inefficient relative to market transactions, which limits firm size regardless of whether workers are human or AI.
“I guess whether you have agents or real human beings working for you you end up with the same problem which is every company bumps up against CO's theory of the firm which says that any given firm will grow only so much to the point where it uh becomes inefficient to be larger than that and then that's why they sort of networks and ecosystems and you know a full blown economy you know like every firm will sort of specialize to do what it is particularly good at”
XML HTTP request (AJAX), added to browsers in 2004, was the literal catalyst for the SaaS boom because it enabled rich internet applications in web browsers that looked like desktop applications, allowing software to move from CD-ROM to web delivery.
“in 2004 browsers added this JavaScript function XML HTTP request which was the missing piece that enabled you to build a rich internet application in a web browser so for the first time you could make things in websites that looked like desktop applications and then that created Google Maps and Gmail and set up this whole like SAS boom essentially”
The Catalyst for the SaaS boom was AJAX (Asynchronous JavaScript and XML / XMLHttpRequest) added to browsers in 2004, which allowed developers to build rich internet applications that looked like desktop apps, enabling the shift from CD-ROM installed software to web-based software.
“in 2004 browsers added this JavaScript function XML HTTP request which was the missing piece that enabled you to build a rich internet application in a web browser so for the first time you could make things in websites that looked like desktop applications and then that created Google Maps and Gmail and set up this whole like SAS boom”
Triple Bite was 50 people but people thought it was hundreds of people because so much high-quality work was being done by a small team, especially in marketing and out-of-home advertising, indicating that team size doesn't determine output quality.
“I remember like the Cal trained station takeover that you did all the like out of home stuff that you did it was like really high quality stuff it stuck with it you could tell was not being done by some like VP marketing person um and that was all mik and like the comment I would often get when people would ask me around that time like how big is triple bite and we were like 50 people and so much yeah yeah people be like I thought this's like hundreds of people”
Zero startups won in the category of obviously good consumer ideas that could be moved to cloud/mobile (docs, email, calendar, chat); incumbents like Google, Facebook, and Amazon captured 100% of the value in these categories.
“the interesting thing is zero startups won in those categories 100% of the value flow to incumbents right like Google Facebook Amazon they own all all those businesses folks forget that like Google Docs wasn't the only company that tried to bring Microsoft Office online there were like 30 companies that tried to bring Microsoft Office online but they all lost Google one”
A trend toward hiring engineers to automate functions was already underway pre-LLM: when building marketing/customer acquisition at Triple Bite, hiring a smart MIT engineer who understood paid advertising was more effective than hiring a traditional marketing executive or VP of Marketing.
“I think what you're saying is like a trend that was already underway pre llms like I remember when I was running triple bite for example we needed to like build marketing or cust like user acquisition basically um and especially after we raised a series B the like traditional way you were supposed to do that is to like hire a marketing executive and build out like marketing team and um and just like basically spin up this machine to do like sales and marketing but I'd actually met like a y founder um Mike who was his company was basically building like a smart frying pan sounds like bizarre but like he was a MIT engineer yeah you remember this um he's an MIT engineer and to sell the smart frying pan he had to get really really good at understanding like paid advertising and um uh Google ads and just a whole bunch of stuff and so he he taken this Engineers mindset approach to it”
Uber-like companies (which had regulatory risk) didn't get cloned by incumbents like Google even after they became obviously successful because incumbents didn't want to endanger their existing business models with regulatory risk—this is the primary reason incumbents don't compete in certain categories.
“this is the classic innovators dilemma at the end of the day I mean you could argue going back to what you said about Uber or Airbnb these were actually really risky things from a regulatory standpoint so if you're Google and you have basically a guaranteed you know giant a pot of gold that you know sort of comes to you every single month like why would you endanger that pot of gold to sort of pursue these things that uh might be scary or might ruin the pot of gold I think that's I think that's like probably the primary reason why the incumbents didn't end up building those products”
Software has three durable price points: $5/seat (consumer), $500/seat (SMB), and $55,000/seat (enterprise), and this maps directly to different buyer types and buyer incentives; enterprise software is traditionally terrible because the purchaser (high-level executive) is not the end user who actually has to work with it daily.
“well there's only uh what three price points in software it's uh $5 per seat $500 per seat or $55,000 per seat and uh that Maps directly to Consumer SMB or Enterprise sales and then I think time in Memorial has taught us that in the past and this is less and less true uh with new software thankfully but Enterprise is terrible software because it's not the user buying it you know some high up Mucky muuk inside Fortune 1000 is the person who's getting whin and D for this you know Mega seven figure contract and you know they're going to choose something that maybe isn't that good actually for the end user the person who has to actually use the software day-to-day”
The primary reason incumbents didn't clone or build Uber and Airbnb even after they became obviously successful is that these companies faced severe regulatory risk and potential personal prison time for founders, which highly paid Google executives wouldn't accept when they had guaranteed revenue streams.
“this is the classic innovators dilemma at the end of the day I mean you could argue going back to what you said about Uber or Airbnb these were actually really risky things from a regulatory standpoint so if you're Google and you have basically a guaranteed you know giant a pot of gold that you know sort of comes to you every single month like why would you endanger that pot of gold to sort of pursue these things that uh might be scary or might ruin the pot of gold”
The second category of successful startups was non-obvious, unpredicted mass consumer ideas like Uber, Instacart, DoorDash, Coinbase, and Airbnb; incumbents didn't compete because the ideas were so non-obvious that they didn't even try until too late.
“then there was a second category which was like Mass consumer ideas that were not obvious that nobody predicted um that's like uber instacart door Dash coinbase th Airbnb those ones those ones came out of left field like the the dot dot dot between XML HTTP request and Airbnb is like very not obvious yeah and so the incumbents didn't even try competing in those spaces until it was like too late and so startups are able to win there”
Salesforce was the first true SaaS company and faced massive skepticism because people believed you couldn't build sophisticated enterprise applications over the cloud, viewing it as a fundamental category error rather than just a technology problem.
“I think Salesforce is probably like the first true SAS company um and i' I remember Mark benov coming to speak at YC and he tells the story as just very early on people just didn't believe you could build sophisticated Enterprise applications like over the cloud or via SAS it was just so um there was just like a perception issue right it was like no like you don't you buy like your box software and that's like the real software that you run the way we always do it”
Zero startups won in obviously good consumer product categories (docs, photos, email, calendar, chat) when cloud and mobile enabled them; 100% of value flowed to incumbents like Google, Facebook, and Amazon instead.
“there's a first bucket that people started with which was like I would call them Obviously good ideas that could be Mass consumer products um so that's like docs photos email calendar chat all these things that like we used to do on our desktop with that obviously could be moved to the browser and mobile and the interesting thing is zero startups won in those categories 100% of the value flow to incumbents right like Google Facebook Amazon they own all all those businesses”
Voice infrastructure companies like Vapi have made it easy for founders to get started building voice AI agents, reducing friction and enabling rapid experimentation and deployment of voice-based applications.
“the space feels like it's moving very quickly and that we have incredible companies that are voice infra companies like vapy and then people can sort of get started right away”
Consumer applications (email, chat, maps) established themselves before SaaS became prevalent in enterprises, and people who used these tools as individuals then advocated for them in their companies, which facilitated SaaS adoption.
“if you think about the history of SAS the consumer things worked first like 2005 to 2010 was mostly consumer applications like email and chat and maps and people got people as individuals got used to using these tools themselves and I think that made it easier to sell SAS tools to companies”
The early LLM app wave (Winter 2023) consisted of simple text generation tools like ChatGPT wrappers and content generation tools like Speedy Brand, which mostly got replaced by the next generation of models, demonstrating that simple wrapping of AI models doesn't create defensible business value.
“the apps were essentially just things that spat out some text and not even like perfect T Rockit talk that's about it yeah sort of more like copy editing marketing edit email edits it was just a kind of more like just like incremental yeah like I I had a company I mean the one that sticks in my head is a company Speedy brand and all what they did is make it very easy for like a small business to just generate a Blog and spit out content marketing um it's like very obvious idea and it wasn't perfect but it was pretty cool at the time”
Over 40% of all venture capital dollars in the past 20 years went to SaaS companies, and more than 300 SaaS unicorns were produced in that 20-year period, which is more than every other category combined.
“over 40% of all venture capital dollars in that time period went to SAS companies and we produced over 300 SAS unicorns in that 20-year time period which is way more than every other category”
Different verticals require very different kinds of support: customer support for a dev tool company is different from support for a marketplace, which is different from support for SaaS companies, which is different from support for hardware companies, validating the need for vertical specialization.
“I mean we already gave three examples of customer support but there are very different verticals it's like de tool comp need very different kind of support that you need and the training set to marketplaces very different right”
The early web and early SaaS apps sucked and required visionaries like Paul Graham to understand that the browser would keep improving and eventually be good, which is reminiscent of today's skepticism about AI agents (that they hallucinate, aren't perfect, are just toys).
“it was it was quite contrar cuz the early web app sucked they were like via web where you had to be a Visionary like PG and understand that the browser was going to keep getting better and that eventually it'd be good which feels like quite reminisent of today right where it's like the yeah the same thing like oh no like you won't be able to build like sophisticated Enterprise applications that use these llm or AI tools because they hallucinate or they're not perfect or they um they kind of like just toys”
Competition among foundation models (with Claude as a major new contender alongside OpenAI) creates a fertile marketplace ecosystem where consumers have choice and founders have viable opportunities, rather than a monopoly scenario.
“lot of the foundation models are kind of coming head-to-head there used to be only one player in town with open AI but we've been seeing in the last batch this has been changing Claude is a huge Contender thank God it's like competition is you know the the soil for a very fertile Marketplace ecosystem uh for which consumers will have choice and uh Founders have a shot”
As SaaS revenue scales, the standard pattern has been for headcount to scale proportionally, with companies that reach $100-200M ARR frequently having 500-2,000 employees, but LLMs are changing this dynamic by potentially enabling significant headcount leverage.
“one of the more Salient things that we've seen for both SMB and enterprise software companies is that or all software companies all startups period is like you know there's a sense that as Revenue scales the number of people you have to hire scales with it and so when you look at unicorns uh even in today's YC portfolio uh it's quite routine to see a company that reached a hundred or $200 million a year in Revenue but they have like 500 a th000 2,000 employees already”
The shift from box software to SaaS eliminated the need for expensive custom integrators and implementations, which opened the gates for vertical SaaS solutions by making it possible for companies to build specialized point solutions rather than having to support every possible feature.
“in the old ways of selling software again like you had this box software that was really like expensive to install and you had like a whole ecosystem around it and anytime you wanted something custom like the integrators would just say oh no like we can like just build a UA custom like payroll feature or something like that and then Salesforce comes along with like a SAS solution and it just seems like it could never be as powerful or sophisticated as like the expensive Enterprise installation you just paid for but they Prov that it totally was the case and I think that just like opened the gates for all all of these like vertical sash solutions to emerge”
Every 3 months the capabilities of AI systems have been getting progressively better, progressing from simple text generation to full-on vertical AI agents that can replace entire teams and functions within enterprises.
“every 3 months things have just kept getting progressively better and now we're at this point where we're talking about full-on vertical AI agents that are going to replace entire teams and functions and Enterprises”
Over the last 20 years, more than 40% of all venture capital dollars went to SaaS companies and we produced over 300 SaaS unicorns in that 20-year period, which is way more than every other category.
“we've mostly been producing SAS companies guys like that's literally been like most of what has been coming out of Silicon Valley it's over 40% of all venture capital dollars in that time period went to SAS companies and we produced over 300 SAS unicorns in that 20-year time period which is way more than every other category”
Obvious general-purpose AI consumer products (like a voice assistant that can do anything) will likely be dominated by incumbents (Apple, Google), similar to how general-purpose consumer products were dominated by tech giants in the cloud/mobile era.
“there's a bunch of categories that are like Mass consumer applications that are obviously huge opportunities but probably the incumbents will win all of those so that's something like a per like a general purpose AI Voice Assistant that you you know you can ask it to do anything and it'll like go do that thing that's an obvious thing that should exist but like all the big players are going to be competing to be that thing right W Apple's a little slow on that one why is Siri so stupid still”
There are only three price points in enterprise software: $5 per seat (consumer), $500 per seat (SMB), or $55,000 per seat (enterprise), and these map to different sales models and user decision-making dynamics.
“there's only uh what three price points in software it's uh $5 per seat $500 per seat or $55,000 per seat and uh that Maps directly to Consumer SMB or Enterprise sales”
Humans have a Dunbar number (about 150 people that you can have meaningful relationships with, or 300 with looser relationships), which limits manager span of control and therefore firm size, but AI tools could extend this Dunbar limit by enabling managers to process more information about more people.
“there a limit of how much uh humans we can have meaningful relationship there's like the whole thing with the D Mar number it's about 300 people that you 150 that you can have a meaningful relationship with”
Foundation models are now coming head-to-head in competition; there used to be only OpenAI in the market but now Claude and other models are serious competitors, which is positive for the ecosystem because competition drives innovation and gives founders choice.
“what's interesting to see is we discussed this in the last episode is a lot of the foundation models are kind of coming head-to-head there used to be only one player in town with open AI but we've been seeing in the last batch this has been changing Claude is a huge Contender thank God it's like competition is you know the the soil for a very fertile Marketplace ecosystem uh for which consumers will have choice and uh Founders have a shot and that's the world I want to live in”
Instant delivery marketplaces like Zepto (competitors to each other) could all potentially use GigML because marketplace customer support is a well-defined category, even though the solution is built for Zepto.
“but if you are you know any of the other Marketplace companies you're probably going to use it cuz like that's a very well-defined kind of marketplace that's you know instant delivery Marketplace”
It's common for YC unicorn startups reaching $100-200M ARR to already have 500-2,000 employees, but LLMs are changing the hiring advice founders receive: instead of recruiting experienced executives to build large teams, founders should hire talented software engineers who understand LLMs to automate bottlenecks.
“to date one of the more Salient things that we've seen for both SMB and enterprise software companies is that or all software companies all startups period is like you know there's a sense that as Revenue scales the number of people you have to hire scales with it and so when you look at unicorns uh even in today's YC portfolio uh it's quite routine to see a company that reached a hundred or $200 million a year in Revenue but they have like 500 a th000 2,000 employees already and I'm just going to be very curious like uh even the advice that I'm starting to give companies that are you know a month or two out of the batch uh it's a it's feeling a little bit different than the kind of advice I would give last year or two years ago”
A plausible future scenario is a unicorn company that operates with only 10 employees, where founders focus on writing evals (evaluation tests) and prompts for LLM systems to drive the core business operations.
“we talked about this in a previous episode we talked about there will be a future unicorn company that's only run if we take it to the limit with only 10 employees that's completely plausible and they're writing the evals and the prompts”
Vertical AI agent companies can be 10x larger than the SaaS companies they're disrupting because they don't just replace software—they replace both the software AND the people who operated the software, capturing payroll savings in addition to software spend.
“the other point of why the bull case for vertical AI agents could be even bigger than SAS is that SAS you still needed a operations team or set of people to operate the software in order to get all the workflows to be done I don't know approval workflows or you have to input the data the argument here is that you will get not only replacing all that set of SAS software so that would be like one to one mapping but is also going to eat all of the a lot of the payroll because we look a lot of the spend for companies big chunk is still a payroll and software's Tiny exactly they spend way more on employees than they do on software so it'll be these smaller companies that way more efficient that need way less humans to do random data entry or approvals or click the software”
Unlike the SaaS transition, enterprises have already been trained to value point solutions and vertical solutions over broad platforms, so vertical AI agents don't need to wait for consumer adoption—they can gain enterprise traction immediately because companies already understand and prefer specialized tools
“I I think the answer might just be like this is this is all just a continuation of software and just there's no reason it has to reset back like llms don't have to reset back to a few general purpose like Enterprise llm platforms doing everything because Enterprises have already been trained on like the value of Point Solutions and vertical Solutions um and like the user experience not going to be that different these things will just be a lot more powerful and so if Enterprises have already built the muscle of believing that like startups or vertical Solutions can be better than like Legacy broad platforms they are probably going to be willing to take a bet on a startup promising a very good vertical AI agent solution today”
There is no 'Microsoft of SaaS'—no single company that provides SaaS for every vertical and product—because for structural reasons each vertical requires different companies, which is why 300+ SaaS unicorns were created instead of one dominant horizontal player.
“there is no like Microsoft of SAS like there is no company that somehow does like SAS for like every vertical and every product like for structural reasons it seems to be the case that like they're all different companies and that's why there so many of them”
Vertical AI agent companies selling to enterprises must go top-down (to CEO/executive level) rather than to the team being replaced, because if you sell to the team whose jobs will be eliminated, they will sabotage the product implementation.
“Enterprise and SMB software often is sold based on a particular person who who is the key decision maker and um you have to go high enough in the organization so that the people you're selling to are not afraid that their whole their job Andor their whole team's job is going to go away totally that's kind of the move that I seen that a lot of companies that sell need to do because if you're going to go and sell to the team that's going to get replaced by AI they're going to sabotage it man it just does not work”
The best vertical AI agent opportunities come from finding boring, repetitive, administrative tasks where you have direct experience or relationships, because that's where the highest-impact and highest-defensibility opportunities live.
“people are watching and thinking about starting a startup or maybe have already started and uh they're hearing all of this how do you know what the right vertical is for you you got to find some boring repeative admin work somewhere and that seems to be like the common thread across all of the stuff is if you can find a boring repetitive admin task um there is likely going to be a billion dollar AI agent startup if you keep digging deep enough into it but it sounds like you should go after something that you directly have some sort of experience or relationship to that is a common like there there's definitely a Common Thread I've seen in the companies that are that I'm seeing promis with”
LLMs represent a new computing paradigm similar to AJAX and cloud/mobile in 2005—a fundamental technology shift that creates an open question of what value opportunities exist and where startups should focus
“I see this llm thing as like actually very similar um it's like it's a new Computing Paradigm that makes it possible to just like do something fundamentally different and in 2005 when cloud and mobile finally took off there is this sort of like big open question of like okay well this new technology exist what should you do with it where is the value going to acrw where are the good opportunities for startups”
Most startup founders, especially young ones, don't fully appreciate how big vertical AI agents will be because they tend to see the startup industry through the lens of consumer products they use, and most SaaS tools are built for companies, not consumers.
“I think in a in a similar fashion people don't understand just how big SAS is because most startup Founders especially young ones tend to see the startup industry through the lens of the products that they use as a consumer and as a consumer you don't tend to use that many sass tools because they're mostly built for companies and so I think a lot of people have missed the basic point that if you just look at what Silicon Valley has been funding for the most”
Mukech is an AI agent for QA testing that can actually replace entire QA teams (not just augment them), allowing sales to pitch directly to engineering leaders without QA opposition, whereas Rainforest QA faced constant sales friction because it only made QA teams more efficient without fully replacing them.
“a company I'm working with u MCH that's sort of essentially an AI agent but for at least where they're starting is like QA testing um they're getting really great traction right now and it's interesting because you remember a decade ago um why can't we worked with rainforest QA like rainforest was a QA as a service company and that they had this exact tension of where they couldn't actually replace your QA team and so they needed to build software that made the QA te more like efficient but really that obviously meant trying to replace as many of them as possible they couldn't replace the whole team and so they were always on the sort of like tight rope between trying to sell the software to like the head of engineering as like this will mean you'll need less QA people and great but then you also have to go sell that to the QA team who don't want to be replaced”
If writing is thinking, then there's significant work involved in figuring out who's an effective communicator and what are the key things to focus on in a company, raising the question of whether LLMs that read and summarize employee input are actually doing some of the thinking/leadership work that was previously done by the CEO.
“I think what you're saying is like a trend that was already underway pre llms I could also see that being another case of um hyper specialization or hyper verticalization I wonder with like that particular tool um just like it's not it's going Beyond just like reading and summarizing like this this is the argument of like if writing is thinking then like there's actually just a huge amount of work that's involved in the effort of figuring out like who's an effective communicator and like what are the most important things to be like what what are the key things to be focused on as a company and I just wonder if that at some point do the llms do like they go beyond just like summarizing and reading and doing actual thinking at which point like who's actually running the the organization”
For every SaaS unicorn, there could be a vertical AI unicorn equivalent in a new universe, where the AI version replaces both the software AND the people that previously used the software.
“here's my pitch for 300 vertical AI agent unicorns literally every company that is a SAS unicorn you could imagine there's a vertical AI unicorn equivalent in like some new universe cuz like most of these SAS unicorns beforehand there were some like box software company that was making the same thing that got disrupted by a SAS company and you could easily imagine the same thing happening again where now basically every every SAS company builds some software that some group of people use the vertical AI equivalent is just going to be the software plus the people in one product”
We are right at the beginning of the AI agents revolution, and it's plausible that there will be a future unicorn company run by only 10 employees who write evals and prompts.
“I think we're right at the beginning of that Revolution right now I mean we talked about this in a previous episode we talked about there will be a future unicorn company that's only run if we take it to the limit with only 10 employees that's completely plausible and they're writing the evals and the prompts”
Vertical AI agents often face a specific sales challenge: they must sell to high enough in the organization (CEO/leadership) to avoid being sabotaged by the team that will be replaced by the AI, because teams whose jobs are threatened will oppose adoption if they're involved in the purchasing decision.
“um because Enterprise and SMB software often is sold based on a particular person who who is the key decision maker and um you have to go high enough in the organization so that the people you're selling to are not afraid that their whole their job Andor their whole team's job is going to go away totally that's kind of the move that I seen that a lot of companies that sell need to do because if you're going to go and sell to the team that's going to get replaced by AI they're going to sabotage it man it just does not work so I think this is an interesting way that a lot of these are top down and you have to go through at some point even get the CEO to sign off on it”
In the box software era, there were initial vendors trying to convince people to use software at all with general-purpose 'does everything' products, before it evolved into more sophisticated and specialized vertical players, following a similar trajectory that LLMs might follow.
“I wonder if like in like the box software world you started off with just like a few vendors who just basically were trying to convince people to use software at all and it was just like it does everything um and then it gets more sophisticated and higher resolution and you get lots of like vertical SS players”
Parker Conrad at Rippling has a different perspective: LLMs' ability to read and process information (not just talk or write) lets one person extend their context window far beyond the natural Dunbar number of ~150 people they can maintain relationships with, potentially increasing the scale of firms a single manager can run.
“that that part a little bit breaks my brain because you know when we spend time with Parker Conrad at ripling uh one of his favorite points is actually well you know everyone's very obsessed with with the fact that the rocks can talk and you know maybe they can draw but the more interesting thing for him you know running HR It software that uh you know he spends a lot of time thinking about HR like actually the coolest thing about the LMS is that the rocks can read and from his perspective like he he's I think he has 3,000 employees he still runs payroll for all 3,000 employees through Rippling so I think he spends a lot of time thinking about like how can one person extend their ability as a manager and uh I think we're going to see a lot more there that would be an a reverse argument that if we're at this moment where uh tools for managers and CEOs are going to get much more powerful um oh it could it could increase the scale of the firm that you can run right”
Enterprises are uncertain about which AI agents they need, so some founders (like Brett Taylor from Facebook) are building platforms that allow enterprises to deploy custom AI agents rather than pre-built vertical solutions, creating a meta-platform approach
“Enterprises in general right now are a little unsure about what exactly they like what agents they need and one approach I've seen from especially more experienced Founders like um Brett Taylor the CTO of Facebook started his company Sierra I don't know all the details but as far as I can tell it's essentially more like broadly about letting Enterprises like deploy these AI agents and spinning them up like custom for the Enterprise versus like oh hey we have like this specific agent to do this”
Voice AI infrastructure companies like Vapid enable rapid prototyping of voice agents, and the question of competitive sustainability is whether these platforms can keep raising the ceiling of capability to prevent customer defection to underlying APIs like OpenAI's voice APIs
“the space feels like it's moving very quickly and that we have incredible companies that are voice infra companies like vapy and then people can sort of get started right away and Retail also I mean these companies that have reached pretty fast scale just because it's one of the more exciting like mindblowing things that you can get up and running within I mean literally the course of hours um and then some of the question that you know remains unanswered and we hope they figure it out is how do you hold on to them especially as you uh run into things like the new open AI voice apis um you know do you go direct like you Pro it's probably way more work to try to use the underlying apis off the bat but these uh platforms are clearly low bar and then the question is can you keep raising the ceiling so that you can hold on to customers forever”
Large consumer applications like general-purpose AI voice assistants are obvious opportunities where incumbents (Apple, Google, Microsoft) will win because they can leverage existing distribution and have less regulatory risk than non-obvious categories like Uber or Airbnb.
“when I think about the parallels with LMS I could easily imagine the same thing happening which is that there's a bunch of categories that are like Mass consumer applications that are obviously huge opportunities but probably the incumbents will win all of those so that's something like a per like a general purpose AI Voice Assistant that you you know you can ask it to do anything and it'll like go do that thing that's an obvious thing that should exist but like all the big players are going to be competing to be that thing right W Apple's a little slow on that one why is Siri so stupid still what year is it it makes no sense”
Customer support AI agents is a famously crowded category with supposedly 100 competitors, but almost all are doing simple zero-shot LLM prompting that can't actually replace real customer support teams with complex workflows, leaving the market almost completely open with only 3-4 companies attempting genuine replacement.
“I also worked with a customer support like an AI customer support agent company called Power help well actually we both did um last batch and I learned a couple interesting things from parel um the first is customer like AI agents for customer support was like the category that's like famously crowded where there's like supposedly like you know a 100 of them and if you go and you Google like AI customer support agent you'll get like a 100 results on Google um but what I learned through working with parel is like it's actually kind of like like almost all of those companies are doing very simple like zero shot llm prompting that can't actually replace a real customer support team that does a lot of really complicated workflows it just kind of makes for like a nice demo like to actually replace a customer support team for like an at Scale company that has like 100 customer support reps that does lots of complicated things every day you need like really complicated software that does all the stuff that like Jake heler was talking about and there's there were only like three or four companies that were even attempting to do that”
A weekend project created a voice chat system that could call all 1,500 of a company's employees, provide personalized messages from the CEO, and extract meaning from employee responses to create executive summaries for the CEO, demonstrating AI's ability to extend managerial span of control.
“I think someone had made a voice chat like just weekend project as a CEO but it would call uh all 1,500 of their employees yeah and uh you know it was you know very short call like kind of sounded like it was from the CEO just asking kind of personally”
Rippling has over 100 founders working at the company as specific people who run entire SaaS verticals inside Rippling, suggesting an alternative approach to building massive platforms by recruiting founder-entrepreneurs rather than traditional managers.
“I found out about this recently off a an interview with Matt mcginness his coo that uh there are more than a hundred Founders who work at ripling now as sort of specific people who run like an entire SAS vertical inside Rippling it's super cool the way he's built the team”
A key unanswered question for voice agent startups is whether they can maintain customer lock-in as foundational APIs (like OpenAI's voice API) improve, or whether they'll need to continuously raise the ceiling on top of commodity infrastructure.
“the question that you know remains unanswered and we hope they figure it out is how do you hold on to them especially as you uh run into things like the new open AI voice apis um you know do you go direct like you Pro it's probably way more work to try to use the underlying apis off the bat but these uh platforms are clearly low bar and then the question is can you keep raising the ceiling so that you can hold on to customers forever”
It's possible that the vertical AI agent equivalent will be just big enough on its own and won't need to go horizontal/broad, unlike historical SaaS companies that eventually broadened to survive, creating an alternative path to massive scale.
“there there's two case it could be that the vertical Point solution could be just big enough and you don't need to do that bro breath thing right it that could be a nice scenario”
Early LLM apps like Speedy Brand (simple blog content generation) were obvious ideas that got crushed by improvements in GPT itself, but this doesn't invalidate the thesis because those early wrappers were fundamentally limited—real vertical agents require deep customization.
“I had a company I mean the one that sticks in my head is a company Speedy brand and all what they did is make it very easy for like a small business to just generate a Blog and spit out content marketing um it's like very obvious idea and it wasn't perfect but it was pretty cool at the time and that's what we've talked about a bunch of the show but that's like the chat gbt racer turned out around that time hey like this is what an llm app looks like it's just a chat GPT rapper it does very basic spits out some text like it's going to get crushed by openi in the next release like and it did”
Enterprise software from incumbents like Oracle and NetSuite has terrible user experience because they try to be 'jack of all trades, master of none,' covering so much ground that they sacrifice usability, creating a 10x worse experience than vertical B2B SaaS companies focused on single use cases.
“the other problem is that with a lot of this enterprise software if you're a user of Oracle and a netw suite because they're they have to cover so much ground the user experience is actually pretty bad they're trying to be jack of all trades but master of none yeah so ends up being a bit of a kitchen sync type of experience and this is where if you go and build a B2B SAS vertical company you could do literally a 10x better experience and more delightful because there's this Stark difference between consumer products and Enterprise user experience”
Triplebite founder found that hiring an MIT engineer (Mike) to handle marketing with an engineering mindset was far superior to hiring traditional marketing executives, achieving 10x better results with leverage from treating marketing as an engineering problem, prefiguring how LLMs will enable similar leverage across roles.
“I remember when I was running triple bite for example we needed to like build marketing or cust like user acquisition basically um and especially after we raised a series B the like traditional way you were supposed to do that is to like hire a marketing executive and build out like marketing team and um and just like basically spin up this machine to do like sales and marketing but I'd actually met like a y founder um Mike who was his company was basically building like a smart frying pan sounds like bizarre but like he was a MIT engineer yeah you remember this um he's an MIT engineer and to sell the smart frying pan he had to get really really good at understanding like paid advertising and um uh Google ads and just a whole bunch of stuff and so he he taken this Engineers mindset approach to it”
Triplebite had the same sales friction as Rainforest QA because the software was designed to reduce recruiting team size but not replace it entirely, requiring buying from both engineers (who benefit) and recruiters (who see a threat), creating tension that limited growth.
“I'm seeing this interesting now um like in recruiting too I had this exact same issue with triple vet where to build the software um to build software that makes it easy to like screen and hire software Engineers you need buying from both the engineering team that they're joining but also the recruiting team and effectively the software we were building was trying to replace the recruiters but we couldn't completely replace the recruiters but now with NYC and so the recruiters were always like oppos opposing it CU it was a threat to them yeah so it just always like friction on like how um on like how far you can get when the customer you're trying to sell to is worried about being replaced”
Rippling launches new verticals that hit multi-millions in ARR on day one, demonstrating that once you have a toehold in an enterprise (HR/payroll), the sales and marketing investment is already sunk and can be leveraged across new products with much lower CAC.
“I mean it's definitely very focused on uh recruiting Founders and I mean Parker like Rippling is essentially the the case against vertical like verticalization trying to uh horizontalization like lots of value and he wants to recruit Founders and teams that build on top of the platform like it's almost a little bit more sort of like amazones whereas like shared infrastructure um yeah I think every product that they've released I mean things like time tracking and whatnot I mean basically they launch a thing and it hits like multi-millions of dollars in ARR on day one of launching and that's exactly what we were talking about earlier like once you once you have a vertical once you have a tow hold what you're saying is Well I have to spend this money on sales and marketing anyway can I uh you know basically get higher LTV and hold my CAC constant and uh that's sort of what you if you look at all the top uh software companies today it's like that's what Oracle is that's what Microsoft is that's what Salesforce is ripling knock on wood going to be the next”
Early LLM-powered apps from winter 2023 were mostly text output (content marketing, email editing), but voice AI apps represent a major evolution showing the capability expand dramatically every 6 months, from 2023 toys to 2024 production systems.
“I think even if you went 6 months back it felt like the voices were not realistic enough yet the latency was too high like there was it felt like we were probably a ways off having AI voice apps that could meaningfully like replace like humans calling people up and like here we are and yeah I was just zooming out thinking back to the first YC batch where llm powered apps first came in was probably winter 2023 you know almost 2 years ago now and the apps were essentially just things that spat out some text and not even like perfect T Rockit talk that's about it yeah sort of more like copy editing marketing edit email edits it was just a kind of more like just like incremental”
In the 2005 cloud/mobile transition, consumer applications like email and chat developed first (2005-2010), and this helped people get comfortable using SaaS tools, which then made it easier to sell SaaS to enterprises because employees were already familiar with the tools
“if you think about the history of SAS the consumer things worked first like 2005 to 2010 was mostly consumer applications like email and chat and maps and people got people as individuals got used to using these tools themselves and I think that made it easier to sell SAS tools to companies because you know the same people are both employees and consumers”
Different customer support verticals (like instant delivery marketplaces vs. SaaS) require fundamentally different training data and workflows, so a solution built for one vertical won't transfer to another
“I mean we already gave three examples of customer support but there are very different verticals it's like de tool comp need very different kind of support that you need and the training set to marketplaces very different right”
Murch can focus its sales pitch solely on the fact that companies won't need QA teams at all, allowing it to sell directly to engineering without needing buy-in from QA teams, which was the friction point that limited Rainforest QA's growth.
“but now like mtic with AI can actually just replace the QA people so their pitch is not oh this like makes your QA people faster it's like this just means you don't need a QA team at all so they can just focus the sell onto like engineering and Engineering doesn't need buying from QA at this point”
For startup founders or potential founders, the core advice is to find boring, repetitive admin work that is tedious and then build vertical AI agents to automate it, as this is the universal thread across successful AI agent startups.
“people are watching and thinking about starting a startup or maybe have already started and uh they're hearing all of this how do you know what the right vertical is for you you got to find some boring repeative admin work somewhere and that seems to be like the common thread across all of the stuff”
Enterprises are currently unsure about what exactly they need from AI agents, creating an opportunity for two different approaches: (1) broad platforms that let enterprises build custom agents (like Sierra from Brett Taylor, or VectorShift), or (2) specific vertical agents for particular use cases.
“one thing might be just Enterprises in general right now are a little unsure about what exactly they like what agents they need and one approach I've seen from especially more experienced Founders like um Brett Taylor the CTO of Facebook started his company Sierra I don't know all the details but as far as I can tell it's essentially more like broadly about letting Enterprises like deploy these AI agents and spinning them up like custom for the Enterprise versus like oh hey we have like this specific agent to do this it's something I've seen from one of my companies called um uh Vector shift that funded about a year ago they're two really smart like Harvard computer scientists”
All software starts vertical and becomes horizontal as industries develop—early players focus on specific point solutions, then eventually try to build platforms doing 'not just a point solution but things that sort of work together' to reach the majority of the market
“I think we're just early in the game right like all software sort of starts quite vertical and then as the industries actually get much more developed um then I mean I just answered my earlier question it's like you know why does company end up having a thousand employees it's actually that uh you know early early in the game everyone's making these specific point Solutions and then at some point you've got to go horizontal like you're already doing this crazy spend on sales and marketing and then the only way you can actually continue to grow once you sort of get 100% or you know some large majority of the market is you actually have to do like not just a point solution but things that sort of work together”
Someone built a voice chat bot that called 1,500 employees sounding like it was from the CEO, asking them to ramble about their week, and extracting the meaning into a summary for the CEO, which is 100x better than pre-LLM pulse survey tools but raises questions about whether LLMs are reading/summarizing or actually thinking.
“yeah I mean large language models can talk and can have conversations and then to what extent can uh you know this power actually extend the capability of one or a few people to uh understand what's going on I I heard about that yuk it got it definitely got me thinking because as understood the product is something like it just it will call up all your employees and then your employees can just like ramble about what they've been doing and it will just extract the meaning out of it and give the CEO of like a like bullet point summary of here the most important stuff and there were a bunch of like SAS companies that attempted to do these sort of like weekly pulse Pulses from employees using like traditional SAS software but like that version is is literally a 100 times better than the pre- elm version of this idea”
Companies are getting faster traction in enterprises for vertical AI agent solutions right now than they've ever seen before, suggesting that Gary's hypothesis about enterprises' readiness for point solutions is already bearing out in the current YC batch.
“if Enterprises have already built the muscle of believing that like startups or vertical Solutions can be better than like Legacy broad platforms they are probably going to be willing to take a bet on a startup promising a very good vertical AI agent solution today and I feel like we're all seeing that in the batch now with some of our companies are getting faster Traction in Enterprises for these vertical AI agents than like we've ever seen before”
Outset is building an AI agent for the survey/Qualtrics space using LLMs for text analysis; Qualtrics won't build the best LLM reasoning system itself, and surveys are fundamentally language processing, creating a market opportunity for specialized vertical AI agents.
“your former uh head of product Aaron Cannon is working on a YC company called outset that I worked with and uh basically they're taking llms uh to the surveys and qual Trix space so qual Trix is almost certainly not really going to build the best of breed uh large language model with reasoning and then the funny thing about surveys is you know who's it actually for it's for people who run products for marketing teams it's for people who are trying to make sense of like what do our customers actually want and what are surveys like guess what that's language so um and then I feel like these types of businesses um actually have to thread this needle”
Rippling is attempting to build a suite of HR tools that eats multiple billion-dollar SaaS companies into a single giant platform, making the case that horizontalization is still possible for large venture-backed companies with strong execution and founder talent.
“if he wins he's going to eat a whole bunch of billion dollar SAS companies and like one one giant company it's very interesting point Gary I think what made me think about this is that with having all these AI SAS tools it's going to give the ability to all these leaders and all these Orcs to basically open the aperture of the context window of how much information they can parse because there a limit of how much uh humans we can have meaningful relationship there's like the whole thing with the D Mar number it's about 300 people that you 150 that you can have a meaningful relationship with but with AI because all of these rocks now can read I think think we will be able to extend that dumbar limit that we have”
Salient has achieved very high accuracy in its AI voice calling and has gone live with many large banks, demonstrating that voice-based AI agents can perform complex real-world tasks at scale.
“and this was a company from last year and demonstrating that that part of it that they were able to get in because they sold through top down I guess the space feels like it's moving very quickly and that we have incredible companies that are voice infra companies like vapy and then people can sort of get started right away and what Salient has done is has been able to actually get very very accurate and it has been going live with a lot of big Banks which is super exciting”
Traditional SaaS pulse survey companies attempted to create weekly employee feedback summaries using traditional SaaS software, but AI-powered versions are 100 times better because they can handle natural language from employees (rambling) and extract meaning, compared to traditional surveys requiring structured input.
“there were a bunch of like SAS companies that attempted to do these sort of like weekly pulse Pulses from employees using like traditional SAS software but like that version is is literally a 100 times better than the pre- elm version of this idea”
Every product Rippling releases hits multi-millions of dollars in ARR on day one of launching (like time tracking), demonstrating that once you have a vertical foothold and strong sales/marketing infrastructure, expanding into adjacent products is extremely profitable.
“I mean every product that they've released I mean things like time tracking and whatnot I mean basically they launch a thing and it hits like multi-millions of dollars in ARR on day one of launching”
Salient (a vertical AI agent for debt collection in the auto lending space) automates collection calls to people who owe money on their cars, which is a 'butter-passing job' (dull, routine work with high churn) that is perfect for AI automation.
“I have a company that I work with called Salient that basically does AI voice calling to automate a lot of that collection in the auto Ling space which tradition so they like call up people and they're like hey you owe $1,000 on your car”
Companies now emerging like Priora are building full-stack recruiting automation (initial recruiter screen + technical screen) and getting great traction because they completely eliminate the need for recruiting teams, avoiding the friction that earlier recruiting software faced.
“but now with NYC and so the recruiters were always like oppos opposing it CU it was a threat to them yeah so it just always like friction on like how um on like how far you can get when the customer you're trying to sell to is worried about being replaced um but yeah I think now it's still early days but now with AI you can build things that do the whole stack like of recruiting we have a company we worked with last batch like Nico work with them a priora which is actually just doing like the full like technical screen the full initial recruiter screen and getting great traction so I think as those things keep going like they won't have they have the same thing you won't have the friction of oh I need to convince recruiters to use this you're probably just like not build a recruiting team in the same way that you used to”
Parker Conrad runs Rippling with the philosophy that managers should be able to manage more employees by using AI and tools, and Rippling itself runs payroll for all 3,000+ employees through its own platform, demonstrating his belief in managers extending their capability through tooling.
“the more interesting thing for him you know running HR It software that uh you know he spends a lot of time thinking about HR like actually the coolest thing about the LMS is that the rocks can read and from his perspective like he's I think he has 3,000 employees he still runs payroll for all 3,000 employees through Rippling so I think he spends a lot of time thinking about like how can one person extend their ability as a manager”
Google might still win search, but Perplexity is definitely giving them a run for the money, suggesting that even Google's fortress categories are vulnerable to new entrants in the LLM era.
“the very obvious thing is search and maybe Google will still win um on search but perplexity is definitely give them run for the money right”
Dev tool companies have the same support/documentation burden as other SaaS companies, but capeAI (a customer support chatbot) built AI agents that respond to technical questions by ingesting documentation and YouTube videos from dev tool companies, resulting in smaller dev support teams for companies using their product.
“I work with this company called cap. AI that basically buildt one of the best chat Bots that responds to a lot of the a lot of the technical details that are hard to answer and I think a lot of the companies that started using them they actually ended up having Dev rail teams that are a lot smaller because it ingests a lot of uh the developer documentations even the YouTube videos that Dev tools put up and even a lot of the chat history so it just keeps getting better and better and it's like gives really good answers actually it's one one of the best I've seen”
Outset (founded by Aaron Cannon, former head of product) is taking LLMs to the surveys and Qualtrics space, where Qualtrics won't be able to build best-in-class language models, but surveys are fundamentally about language (understanding customer intent), creating a large opportunity.
“your former uh head of product Aaron Cannon is working on a YC company called outset that I worked with and uh basically they're taking llms uh to the surveys and qual Trix space so qual Trix is almost certainly not really going to build the best of breed uh large language model with reasoning and then the funny thing about surveys is you know who's it actually for it's for people who run products for marketing teams it's for people who are trying to make sense of like what do our customers actually want and what are surveys like guess what that's language”
Murch (an AI agent for QA testing) is getting great traction now, which is notable because QA testing is one of the few jobs that can be completely replaced by AI (you don't need QA people anymore at all), unlike roles that typically can only be augmented.
“a company I'm working with u MCH that's sort of essentially an AI agent but for at least where they're starting is like QA testing um they're getting really great traction right now”
Parel (an AI customer support agent company) operates in the famously crowded AI customer support category where there are 100+ competitors, but most of those competitors use simple zero-shot LLM prompting that can't actually replace real customer support teams doing complex workflows.
“I also worked with a customer support like an AI customer support agent company called Power help well actually we both did um last batch and I learned a couple interesting things from parel um the first is customer like AI agents for customer support was like the category that's like famously crowded where there's like supposedly like you know a 100 of them and if you go and you Google like AI customer support agent you'll get like a 100 results on Google um but what I learned through working with parel is like it's actually kind of like like almost all of those companies are doing very simple like zero shot llm prompting”
Rippling aims to build a suite of HR tools where if successful, it would eat many billion-dollar SaaS companies and consolidate them into one giant company, representing a reversal of the vertical-to-horizontal consolidation pattern.
“if he wins he's going to eat a whole bunch of billion dollar SAS companies and like one one giant company it's very interesting point”
Rippling has over 100 founders working as internal leaders running distinct SaaS verticals (product lines) within the Rippling organization, similar to Amazon's structure with shared infrastructure and internal entrepreneurship
“there are more than a hundred Founders who work at ripling now as sort of specific people who run like an entire SAS vertical inside Rippling”
Uch can sell to companies that don't currently have large QA teams, allowing them to scale without needing to replace existing QA staff—the company uses Uch from the beginning rather than replacing an existing team
“you can also go in I mean to start with you can go and sell to companies that don't even have big QA teams at the moment they just use something like mtic and then it will just like keep scaling with them scaling and they'll just never build a QA team ever”
A recent YC company (Priora) is building an AI agent for recruiting that handles both the full technical screen and initial recruiter screen, getting great traction because it fully replaces recruiting processes rather than trying to augment recruiters.
“now with NYC and so the recruiters were always like oppos opposing it CU it was a threat to them yeah so it just always like friction on like how um on like how far you can get when the customer you're trying to sell to is worried about being replaced um but yeah I think now it's still early days but now with AI you can build things that do the whole stack like of recruiting we have a company we worked with last batch like Nico work with them a priora which is actually just doing like the full like technical screen the full initial recruiter screen and getting great traction”
Cap.AI built a chat bot for developer tools that responds to technical questions by ingesting documentation and chat history, allowing developer tool companies to operate with smaller dev support teams because the AI agent answers most questions.
“I mean other example is even for de tool companies they have to do a lot of a developer support and I work with this company called cap. AI that basically buildt one of the best chat Bots that responds to a lot of the a lot of the technical details that are hard to answer and I think a lot of the companies that started using them they actually ended up having Dev rail teams that are a lot smaller because it ingests a lot of uh the developer documentations even the YouTube videos that Dev tools put up and even a lot of the chat history so it just keeps getting better and better and it's like gives really good answers actually it's one one of the best I've seen”
Salient is building AI voice calling to automate collections in the auto lending space, replacing low-wage call center workers with AI agents that make collection calls with accuracy comparable to human collectors.
“I have a company that I work with called Salient that basically does AI voice calling to automate a lot of that collection in the auto Ling space which tradition so they like call up people and they're like hey you owe $1,000 on your car yeah which actually up with that actually this kind of job is one of those butter passing job it kind of sucks because a lot of uh these low wage workers work in all these call centers and it's like a terrible boring job so very high churn and giant headcount to run these because there's just so many accounts with these banks that have to do that and this is a perfect task that AI could automate”
Sweet Spot found the idea for automating government contract bidding by noticing a friend whose full-time job was refreshing a government website looking for new proposals, identifying a boring admin task that an LLM could automate.
“there there's definitely a Common Thread I've seen in the companies that are that I'm seeing promis with and another one just pops into my head sweet spot I think I mentioned on this before like they're basically building an AI agent to bid on government contracts and the way they found that idea and a year ago was they just had a friend whose full-time job was to sit there on like a government website like refreshing the page like looking for new proposals to bid on and they they were pivoting they're like ah like that seems like something an llm could do”
Gig ML is building an AI customer support agent specifically for Zepto (an instant delivery marketplace) handling 30,000 tickets per day, replacing a team of 1,000 people with a hyper-specialized system built specifically for that use case.
“instead you know you're going to have companies like gig ml that you know it's doing it for zepto doing 30,000 tickets uh every single day and replacing a team of a thousand people and but it's very specific and it has you know it's not a general purpose demo Weare kind of thing like it's 10,000 test cases in a very detailed uh eval set that you know is basically just for zepto”
In the movie 'Her,' the scene zooms out to show that the OS is having conversations with thousands of people simultaneously, which illustrates the potential of AI systems to scale communication and decision-making beyond individual humans.
“I mean it's like a count to that is like the very obvious thing is search and maybe Google will still win um on search but perplexity is definitely give them run for the money right yeah this is the classic innovators dilemma at the end of the day I mean you could argue going back to what you said about Uber or Airbnb these were actually really risky things from a regulatory standpoint so if you're Google and you have basically a guaranteed you know giant a pot of gold that you know sort of comes to you every single month like why would you endanger that pot of gold to sort of pursue these things that uh might be scary or might ruin the pot of gold I think that's I think that's like probably the primary reason why the incumbents didn't end up building those products and didn't even clone them even after they got big and it was obvious that they were going to work who will never launch an an Uber clone they never launch an Airbnb clone um I was listening to this uh talk by Travis and”
Cap.ai's developer support chatbot is one of the best currently available, demonstrating that specialized vertical AI agents can achieve excellence that general-purpose systems cannot.
“it's one one of the best I've seen”