Hebbia founder and CEO George Sivulka discusses the potential for reasoning models and AI agents to supercharge knowledge-worker productivity — and the global economy along with it.
In this episode of AI + a16z, a16z partner Alex Immerman sits down with Hebbia founder and CEO George Sivulka to discuss the potential for reasoning models and AI agents to supercharge knowledge-worker productivity — and the global economy along with it. As George explains, his customers are already saving significant time and and effort on important, but monotonous, tasks, and improved models paired with savvy users will continue to reshape how industries including finance, law, and other professional services operate.
Follow everyone on X:
George: A lot of my peers, if they were smart and they were lucky, they were going into financial services jobs. They would go and become an investor. And I realized that the smartest people in the world, incredibly smart kids, were going and doing the stupidest tasks.
Alex: So many tedious things.
George: A lot of...
Alex: On repetition, boom, boom, boom, night after night.
George: And I saw pain. I noticed in my friends that they literally hated their lives. And at Stanford, maybe the only other thing that I took from the entrepreneurship community there was build a company where there's pain. And my 22-year-old brain said, well, there's a lot of pain here, and there's the most important technology in the world that can solve it. That's where I'm going to go and build a business before anyone else catches onto it.
Derrick Harris: Thanks for listening to the a16z AI podcast. Today, we're exploring the intersection of professional services and AI agents via a conversation between a16z partner Alex Immerman and Hebbia founder and CEO, George Sivulka. Although we're very early on in the development of agentic workflows, Hebbia can safely stake its claim as an early application of them thanks to its embrace of reasoning models and a user interface that does away with the reliance on chatting.
In this discussion, George and Alex explore the rationale behind some of these design decisions. They talk about building products for an AI-native user base and dive into the industry-changing benefits of AI, especially in areas like financial services, investing, and law, where smart people spend way too much time on tedium and checking boxes.
Oh, and George also gives his opinion on DeepSeek and dishes on his favorite AI tools. All that and more after these disclosures.
Disclosures: As a reminder, please note that the content here is for informational purposes only, should not be taken as legal, business, tax, or investment advice, or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any a16z fund. For more details, please see as16z.com/disclosures.
Alex: George, it's really fun to have you here. I've enjoyed getting to know you the last couple of years. I'm excited for the world at large to hear more from you.
George: I'm excited to be here.
Alex: Amazing.
Well, we're going to open with a lightning round getting to the heat of AI.
George: Great.
Alex: Let's rip it. Do you think scaling laws are going to hold?
George: It's a good question. I think that there's two types of scaling laws. You have scaling laws for training, which is what I think you're referring to. And then more recently, people have started to talk about scaling laws at inference. And I think they're both effectively mathematical properties of the universe. I don't think they're just an experimental observed thing. You just know as you add more data and add more compute, these models get better for training, that is. And I think that they will always hold if there's enough data. And that is the question. But I do think that GPT-5 will be significantly better than GPT-4.
At the same time, you're starting to see models like kind of 01, 03, doing this, like, reasoning at inference, scaling and...
Alex: DeepSeek.
George: DeepSeek as well. And this is a technique, scaling at inference, that was first actually pioneered at Hebbia. And we quickly noticed another scaling law where if we ran more models and basically more compute at inference time, you could get much better results for very complex tasks. And I think that scaling law has already proven to kind of extend the runway of AI. And I think it will continue to hold as a mathematical property in itself.
Alex: All right. So scaling laws are going to hold. We talked about DeepSeek briefly there. Is DeepSeek a nothing burger or is it a big deal?
George: I am a believer that like it was more on the nothing-burger side of things. I think China has shown time and time again that they're able to take technologies that are invented oftentimes in the United States and make them more efficient. And I think China has also shown time and time again that they're willing to obfuscate the truth or omit certain things when talking about, you know, certain technologies or the science behind a variety of different things. You know, regardless of the truth with DeepSeek, I generally also think it's really an American technology that's cheaper. And Americans can also make it cheaper. And it was invented here in the United States and China has not shown that they can actually continue to play ball pushing the frontier of AI. This is not an example.
Alex: Yeah, I think the one piece on it was also like pushing open source in the U.S. here and making us rethink potentially things on the regulatory side there.
George: It's a good point. I think open source is important. And I think that as software becomes open source, it stops being a geopolitical strength of just like the closed source nation that has invented it. But at the same time, you know, America, I believe, is so far ahead. And I think that it will continue to be so far ahead. And these technologies, you know, every single year are so much exponentially better than the last year that we have the headstart. And I think that we'll continue to have the headstart. And we have the world's best scientists, researchers and minds thinking about this and working on it.
Alex: Absolutely. And then we also have the best companies building on the app layer who are going to benefit from these models getting cheaper and cheaper.
George: One hundred percent, especially if they scale at inference. So...
Alex: Exactly. So next one, what's your favorite AI tool that's not named Hebbia?
George: I think my favorite tool right now is Deep Research, which was launched yesterday or the day before yesterday.
Alex: It's pretty new.
George: Very new. It has changed the amount that I use AI. And I use AI for everything. I try to automate everything.
Alex: What have you used it for? What searches?
I was looking at microplastics' impact on children. That's, like, a hot topic these days. Deep Research goes deep. It's pretty inconclusive though.
George: It's inconclusive. Really. And we can get to the difference between Deep Research and ultimately what other agents over private data will be working on. And I think that this level of agentic deep research is going to be one of the most exciting things that I can't quite disclose now that Hebbia has proven out and is working on over private information.
I think Deep Research allows you to experience what an agent can do for the first time. And it uses information retrieval over the web, which is very good. But I've used it for things like looking at a list of names of executives I'm about to meet and understanding everything about them. So almost like client prep. I've used it for building a competitive strategy around how I should approach a variety of different things strategically with partners in our ecosystem. And it's not as good.
Alex: Deep Research has been busy for you.
George: It's our number one employee. Don't take that wrong, Hebbia. It's not as good as the stuff we'd create internally. But it starts to be more fulsome. It's more than just a single step. And that's a really important piece of the future of AI applications.
Alex: Well, you keep alluding to Hebbia. So I think it's time we get there. Let's rewind the clock a little bit.
George: Sure.
Alex: You're a PhD student. You're studying neuroscience, engineering, applied physics, all the things. And then you decide you're going to build a company that sells into financial services. How did we get there?
George: At Stanford, there's a class which is CS 330, which I think it's probably still around. But it's like an introduction to meta-learning and multitask learning. And when I was in grad school, I was completely captivated by this class. The idea and the promise of meta-learning and what that means is teaching machines to learn how to learn, was to me going to be the most important technology of all time. It was just the coolest thing in the world. And I was in my PhD. And I think around June of 2020, looking at all the research, keeping track on meta-learning, multitask learning, OpenAI releases a paper. And it said "GPT-3, large language models are..."
Alex: Pretty important.
George: It was very important. But the title of the GPT-3 paper was actually "Large Language Models are Multitasking Meta Learners." Or something along those lines. And I remember thinking, okay, that would have just completely botched this entire field of research that is encapsulating all of my academic interest. And I played around with it. And I remember... and GPT-3 wasn't ChatGPT. This is actually over 12 months before ChatGPT was...
Alex: Yeah, like a year and a half almost.
George: Yeah, 18 months, I guess. And I'm playing around with it. And I'm like, holy shit, this is way better than anything I'd ever seen. And I remember I had one aha moment playing around with it where I was like, okay, it's definitely a multitask learner. Another aha moment. And I was like, it's not going to be anyone working on this stuff at Stanford that is going to invent the most important technology of the next century. And I just became obsessed with understanding how to apply it and to build what I thought would be the most important product of the next century. And at the same time, and this is kind of, it's like, how do you get from that to building an AI platform for knowledge workers, for financial services? A lot of my peers were, if they were smart and they were lucky, they were going into financial services jobs. You know, they would go and become an investor. And I realized that the early, like, incredibly the smartest people in the world, incredibly smart kids, were going and doing the stupidest tasks.
Alex: So many tedious things on repetition, boom, night after night.
George: And I saw pain. Like, I noticed in my friends that they literally hated their lives. And at Stanford, maybe the only other thing that I took from, like, the relative entrepreneurship community there was build a company where there's pain. And, you know, my 22-year-old brain said, well, there's a lot of pain here and there's the most important technology in the world that can solve it. Like, that's where I'm going to go and build a business before anyone else catches on to it. Yeah.
Alex: You were talking about GPT-3. It came out in summer of 2020. Fast forward a bit. ChatGPT came out in November 2022, but not specific for financial services. Right? So as you were building your product, what made you realize Hebbia is going to be much better for financial services use cases?
George: There's a piece of the ChatGPT experience, which is very magical, where it's very good at doing a single task. You can have it write a poem or you can have it do your kid's homework. And it feels like the world's going to end and everything's fixed. And we've reached Nirvana AGI mode. And then you try to start to ask it to do complex tasks, things that would require multiple steps or be differentiated. And the response is always, as a large language model, I can't do that. And I think they've now changed the way it responds. That was always the response. And if you think about why it was broken and it was good for things that were maybe fun, creative, generative, and bad for things that were requiring serious work, there's...
Alex: Accuracy.
George: It's not even just accuracy, though. There's some piece of it where it's trained only on public data. And the stuff that matters for financial services, but also for law, for all of knowledge work, is offline unstructured information. And there's a piece of it where the process that is required to work is a little bit too complex. And if you look at ChatGPT and the limitations of like you give it to a bunch of investors and you ask it to draft an IC memo for a company like Hebbia, every single VC would get the same IC memo out.
Alex: More or less. Yeah.
George: I mean, it's not differentiated. It doesn't have custom process, custom data.
Alex: No alpha.
George: No alpha. And the idea was, hey, financial services is a place where the onus of putting the right information and getting the right result and having your custom process and it being transparent, i.e. doing knowledge work, was the most important. And if this technology was fundamentally as transformative as it should be, a tool that would fix the lives and workflows of investors would actually go to market far faster than anything that would just be a value add in a more ephemeral or like single-step way.
Alex: Yeah. I think an important piece there that you mentioned is the private data. So if I'm using Hebbia, I could upload all of Andreessen Horowitz's investment memos. And with that information, you could give me a very different result than if you were just searching on the web.
George: That's right. Yeah. So you could put every single former IC memo in, structure that, create a library over, Hebbia has got like an infinite effective context window. And then when you look at a new opportunity, you could filter to a variety of other opportunities that were similar and actually begin to piece out how it's different, how it's similar, and all the pieces that would constitute an Andreessen Horowitz investment memo. You could start to recreate.
Alex: And then with Hebbia, I think if you were to compare it to ChatGPT as you just did, but many of the most commonly used AI applications today, Perplexity, Claude, maybe DeepSeek recently, they're all chat interfaces. You've taken a different approach, maybe share a little bit more about what you've done on the interface side.
George: I think it's the biggest change in AI interfaces since launch of ChatGPT where everything became a chatbot and everyone copied that. But Hebbia allows AI to do work and show its work. And to do that, we let it really put the sources first where it builds a grid out. And that grid has every single document as a row and every single column as a prompt or some sort of agentic action.
And so you could imagine that as an AI is going over data or pulling numbers or pulling different quotes from documents, when you're asking a bunch of different questions, it actually starts to build this piece of collateral that shows its work. And that collateral is effectively an agent orchestration system, an ability for humans to manage agents for every single row and every cell is an agent. And it shows the complexity of what's happening with repeated tasks by an AI agent. They're very simple and elegant and human first way. And it's almost like the mind of the AI is opened up. And you can see...
Alex: So you can see how it's thinking.
George: You could start to change the cells. You could start to act... it's almost like a sensitivity table. And so you're not only kind of telling the AI what to do and seeing its output, but you're working alongside it or co-working and collaborating with it. And right now, the limiting factor for much of AI, especially what we do in financial services, legal or knowledge work more broadly, is information retrieval. And so it's very document-centric. But we're already exploring and have built a multitude of interfaces where you can see inside the AI's mind doing different things. And that is a human-first approach. It's human and AI kind of fusion together.
Alex: So as you think about the new capabilities, there's computer use, there's reasoning. I think those are really instrumental in being able to drive more of these agentic workflows that you're referring to. How are you thinking about leveraging computer use in Hebbia's product?
George: I think that the way we've designed our product is not to be just the interface to AI. But actually to be a tool that we would hope AI or AGI would choose to use. And so that's a bit of a meta-statement, actually. It's saying, well, rather than try to build the AGI, how about we build the AI platform that is so good that if AGI were to complete a task, it would choose to use Hebbia to do that task? And to that end, I think computer use and a variety of these other tool-use phenomena from the large language model providers are just creating universal APIs to other tools.
Products like Instacart might still be the best way to deliver groceries, right? For if you're AGI, right? You don't want the AGI to go build a robot and go get it itself. And products like Hebbia would be probably the best way to do structured and unstructured data processing, the knowledge work tasks with sub-AIs. You wouldn't want an AGI to jam 100,000 documents into its context window and take an infinite amount of time to [inaudible 00:15:35.983] a very large amount of time to process that. You'd rather use and orchestrate a bunch of sub-agents. And that's a little bit of the philosophy behind what we do.
Alex: And as you think about building for this world with financial services, a lot of what you just talked about is more around the infrastructure, the architecture of what you've built. There's also the interface layer, the application layer, the interfaces, the data that you're leveraging, the integrations, the workflows. Maybe like go a layer deeper there. What are you specifically doing to make the, whether it's an AI agent or today a human that's in the loop, most effective as a financial analyst?
George: One of the big pieces and unlocks that Hebbia experience was coming to the realization that nobody knows what to use AI for. People pretend to go and create all these amazing demos, but nobody knows what to do with it. And you can actually look at our product today and you could argue that it could do almost any task that a junior analyst could do better than a junior analyst. But even knowing what tasks it should do becomes a problem. It's actually no longer a technology problem. It's really a sociology and change management problem.
So there's an integrations piece. There's a process piece of making sure it works for specific workflows. But the real differentiator is actually knowing what to use it for, knowing what is best in class, and helping evolve the firms and do the change management required to create an AI-centric workforce. We're starting to see the first year or second year out of college analysts come out and be like Hebbia-native where they, instead of doing it and reading every file by hand and then putting things into an Excel or making different documents, they're actually starting to use AI and incorporate it into their workflows. And it's making them like that much better.
Alex: That all makes a ton of sense. For most AI applications, there's two vectors of the value proposition. There's one faster speed to get the same result. So what used to take four hours now takes four minutes. And then there's the second vector, which is net new results. Things that maybe with infinite workers, with infinite memory, you could figure out. But those are like the magical experience moments. What are the common use cases you're seeing today?
George: You're right to put it into two buckets. There's the time savings piece and then there's like the stuff we wouldn't have done if we didn't have AI. And I'm way more excited about the second piece. I think everyone is versus here, your analyst is 80 times faster. On the buy side, we work with a lot of the largest asset managers in the world. And a very common use case is they get an hourly data room. It's got 40,000, maybe 100,000 files in it. They put it in Hebbia and it automatically builds out a variety of different analyses. Without having to do anything, without having to even open your computer, you have a lot of the different things you're looking for. Customer concentration. You have an understanding of all of their expert network calls. You know, it actually does the analysis for you. That's a time saver. I mean, it saves probably 20 to 30 hours on a deal process, depending on how disgusting these VDRs can get.
But there's an even more beautiful piece of the puzzle, which is the net new piece. On the net new side, I think a lot of the reason people become investors or work in finance is because they want to be discovering. They want to be doing the act of almost like a Sherlock Holmes looking through the data, looking through what the public markets are saying.
Alex: You're constantly learning. You're constantly finding something new. That's like part of where you get your energy.
George: It's scientific and it's actually an empowering act if you're really like investing. And my favorite thing on the net new analysis side are when people are looking over more data than they otherwise would have looked over and drawing a journey. They will go into a bunch of expert calls and pull out the fact that maybe the customers are unhappy with one specific feature. Maybe based on that, they'll go into a variety of technical documentation and pull out what's changing about that feature and notice that there's something in the supplier agreement that needs to also change and that is bottlenecking that.
And they'll go on these journeys through the supplier agreements and through all the data, public and private, about a company and actually come up with theses or hypotheses that they otherwise could not have done. And the feeling of discovery is, I think, the best part of being an investor. And I think we now impart that with a variety of our use cases. But I want to make it very tangible in terms of the use cases. And I'll just give you two examples where I'll go into an asset manager and I'll say, hey, these are two things we know people do with Hebbia.
And the first is, hey, people are reviewing, associates, analysts are reviewing marketing materials every single week, sometimes like for entire days of the week, just going through opportunities, trying to figure out if something matches your investment criteria or not. And the most common outcome of reviewing a marketing material in private equity would be a SIM, in credit might be an offering memorandum, is you come out and you say, it's not qualified. It doesn't meet our investing criteria. And so you spend all this time researching a company, reading through this, like, thing that a banker really tried to obfuscate to like tell you that it's [crosstalk 00:20:42.699].
Alex: It's worth your time.
George: It's a beautiful opportunity. And the majority of a junior analyst's life could end up going and saying, hey, not worth our time, not worth our time. And the deals just die.
Alex: You can focus on something that actually is going to have legs.
George: And what Hebbia does is the minute you get marketing materials, just a simple use case, you put it into the platform. It'll compare it to every other opportunity you've looked at. It'll spread it out across your investing criteria and give you a go, no go. So something that used to be the bane of your existence, it's literally meaningless labor, you're just pushing the boulder up the hill, is now instantaneously done. And you can spend more time, as you mentioned, on the deals that matter. So that's one use case, screening. We can screen 137% more opportunities, like in a given period of time, which is like, you know, and it's with the same depth.
Another use case, maybe a little bit later in the diligence process, every time there's a lot of documents or a lot of the same document, like a lot of credit agreements or a lot of expert network calls, you'll often find juniors or even VPs scouring and reading through every single one to understand and synthesize an insight across the entire thing. And with Hebbia, in seconds, what used to take 200, 300, 400 pages of reading, you can instantaneously get the insights on a per-call basis and then over every single call or on a per-credit agreement basis and the entire capital structure of a company. And when people see that and they experience that, it's not only like an emotional experience, they just realize that they've spent 20 or 30 hours, every single deal process, doing something that can now be done fast and their eyes light up. It's amazing.
Alex: One of my favorite parts when I'm using Hebbia is just leveraging the templates. And of course, you know, I can customize them and maybe there's some questions that I want to ask that not everyone else does. But if it's my first time doing an earnings call review using Hebbia, I can do it very quickly out of the box, like time to value.
George: Yes. There's a library or a thousand different templates. I think it's, like, now closer to 2000 different templates, which are all effectively agents. One of them is a credit agreement agent. One of them is a earnings call agent. One of them is a expert network agent or a screening agent. And those are the use cases. And that's actually why it's so measurable.
If last year was a year of everyone experimenting with AI and not knowing what to use it for, now we have like a screening template where we can measure the amount of time it takes you with Hebbia and before Hebbia and prove to a financial services firm this is tangible ROI.
Alex: All these examples that we've gone through primarily on the buy side. I think you guys work with a bunch of advisors. What are their use cases look like?
George: I can't speak to individual customers, but in general, if you can go in and understand not only from the materials that a client gives you, but then also the entire world, the internet of information, and are able to pitch their company better, create better marketing materials, that's a massive use case.
There's also just pieces in a standard deal process when you're responding to the buy side and responding to a due diligence questionnaire, being able to answer that. That's a perfect agentic workflow, but with checkpoint steps for every single piece. And we can automate a really large part of that, which is traditionally one of the largest time sucks of a junior first-year out-of-college investment banking analyst. And eventually, we're going to put the logos on the slide in the right format, and they'll be perfectly centered and everything else. But a lot of the most tedious parts of the job in banking, completely automated.
And on law, people are trying to use things like our infinite effective context window to actually look over entire libraries of formerly negotiated agreements. And live during a negotiation actually come with better terms or a better understanding of what is market over previously very obfuscated data where you'd have to be a partner with many years of experience to understand exactly what the right thing was to land on for a client.
Alex: So in 2023, certainly through much of 2024, AI was about experimentation. The board was fixated on what are you doing in AI? Today, we're in 2025, we've gotten to a place where it's kind of a given. We're assuming all of our companies are leveraging gen AI in some form. And so the question is becoming, where is the value? What is the return I'm seeing on all this AI investment I'm making?
George: Yeah, you're right. I'd say that 2023 and 2024 were really the year of experimenting with AI. And 2025, it's now like, hey, the boards are saying what's the P&L impact of the, you know, sometimes $100 million investments that we're making on this experimental technology. And it's put the onus on Hebbia and the way we engage with our customers to always have value cases and to always prove out value cases. And you start to get stories to go back to the credit agreement example, where firms are saving, you know, if every single credit agreement, you know, they're reviewed multiple hours, and it costs $2,000 per hour for a lawyer to review them. And now they can review them in-house, they're saving on tens of thousands of dollars on a per-deal basis, maybe hundreds of thousands of dollars for a significantly complex deal on credit agreements. And they're looking at hundreds or even thousands of deals a year. That's meaningful ROI.
There's other examples where Am Law 50 clients are onboarding new customers for their own clients. And they're basically saying, hey, what used to take five to eight hours a customer to onboard and pull up all the right data, we now can understand instantaneously, right? Lawyers sometimes will have to look at their clients' data room, figure out all the red flags, problems and everything else. And now they can do that in Hebbia and know everything about their customer out of the box. And there's examples of private equity associates that are saving, you know, four to eight hours a week because they're building out libraries of their portfolio, companies' board decks, and investor updates. And they're able to benchmark their portfolio much more accurately.
Alex: You just mentioned the private equity associate. To get ROI, humans are in the loop. Humans are working with the software. They're involved in Hebbia, they're there every day. Talk about the relationship between human and software and where that is today, where that's going.
George: I think ultimately, there's a lot of software and a lot of AI that seeks to replace humans. And the way that we design, the way that we conceive of new interfaces, and I think the way that I hope the larger industry does is in ways that empower humans, in ways that make humans better. And ultimately, I think Steve Jobs always talked about computers as bicycles for the mind, right? It was a way to just go down and basically explore or do whatever you wanted to do. It's a simple machine that was a very elegant solution. And I think that it's easy to stop at chatbot and say, okay, this is our bicycle for the mind. But when you can start to explore other interfaces and still empower the user, that is empowering. That is software that is in service of the human and the human that's appreciating the software for what it can do.
Alex: So as we think about humans in the loop, those humans could be senior investment professionals who may be reading someone else's work, but the majority of your users, they're early career professionals. They are what you referred to earlier as AI literate, this next generation. How are you building your products for them? How is it designed? What's special for that demo?
George: I think that the demographic of folks that are early in their career, they're the ones that for the first time are AI-native. They are the first analysts that are learning how to do the role where they're actually using technology that can do the mundane parts of it. And I think that it's funny because you'd think that only Gen Z would be really AI-native, but there are even some MDs, like the one-off MD that will go in and actually check their analysts' work with AI. So they'll create a matrix and pull out red flags or inconsistencies over something their analyst sent. But I think it's special because when you catch people early on in their career, obviously the banking analyst sometimes will go and become an investor and the investor will sometimes go and work in a large corporation or maybe they'll become the founder of their own firm one day. And I think that the change starts with the people that are the early adopters, with the people that have the most mundane workflows, and then it has to bubble out through the organization. But that can only happen with a lot of the C-suite executives and MDs and CIOs that support Hebbia. I mean, they love it. They use it themselves and they also have a change management mandate. And so it's kind of two-sided. And our best deployments are those where both the senior folks and the junior folks are aligned on using AI.
Alex: As you think about, like, the finance industry in 5 to 10 years, Hebbia AI, a lot of change coming the industry's way. Do you dream the dream? What does the industry look like in a decade's time?
George: I think this is my favorite question in the world because I believe fundamentally a few different big things will happen. First, when AGI is here, there will be a massive correction in the financial markets. That is actually my Turing test. Is if AGI is here, will it actually be able to make significantly more money than a human investor, i.e. better than human investing? And I think part of that will be uncovering fraud at massive scale, uncovering market inefficiencies and human behavior that is mispricing assets. And you'll actually start to see what happened with quantitative investing sort of happen. And really are about all of the qualitative alpha that exists or does not exist in the world. And everything will just turn to beta. Levered beta, if you will. If you have leverage.
On the other side of things, I also think that private investing will change way more than public investing.
Alex: Why is that?
George: If you look at public markets, you've got like a Bloomberg terminal, right? You can go in and you can look at every single number and it's all perfectly formatted. And every quarter you have to file with the SEC and you have all your numbers, perfect. In the private markets, you have a data room. And the art of private market investing, especially in the later stages, especially private equity, private capital, is basically building a Bloomberg terminal for the private company. You want all the numbers. You want the charts. Your IC memo is effectively the Bloomberg terminal for the private company. And there might be some qualitative pieces, but you're trying to structure and structure information. And what Hebbia is trying to build is the Bloomberg terminal for private companies. Not like a pitch book or a crunch base, which is really just for sourcing, but actually to go out and take all of the information private and public and pre-structure that so the minute you get to a new company you can actually perfectly understand what it should be valued at, all of the numbers that you would look at on a Bloomberg terminal, pulled out and ready to go for you for the private company. I think that will change the way from a speed perspective, from an accuracy perspective, from the ability for LPs to want to invest in the asset class, from the ability or the proclivity for private companies to want to go public. All of that, I think, will change. And I think AI will really transform the capital markets.
Alex: Does that mean that older firms who have the most data will be advantaged? How important is data, like private data, towards winning?
George: It's interesting from the perspective of firms that have lots of data, like megafunds. When they come to Hebbia, they're able to say, hey, I have this history of 10,000 SIMs. And so...
Alex: Since 1975.
George: Since 1975. And Hebbia can OCR them, pull out all the... look at all the graphs and tables.
Alex: When you used to run times per head.
George: Yeah, in an HP 12C. No, but it can basically turn into this beautiful SIM library. And everyone's like, wow, this is amazing. But at the same time, I also think that as the market changes, you'll actually start to see new opportunities that don't have comps. The historical stuff won't actually matter. And so perhaps spending too much time thinking about the historical deals that were pre-ChatGPT or pre-generative AI or pre-LLMs are not actually going to help you with the deals post-LLMs. And it'll be even better if you don't have that much information.
Alex: It's like if you had all the software deals from license maintenance mode to switching to subscription, the license maintenance ones are helpful to some degree, but what you actually care about is the subscription business model. And maybe in this transition we're undergoing right now with AI, there's questions around business model. Subscription is a default. A lot of companies are charging per seat. But some are starting to charge different things on an outcome basis, on volume. For Hebbia, you've got one model today. How do you think that evolves over time? How have you thought about pricing?
George: Hebbia right now prices per seat because we're trying to build a product that incentivizes usage. And I think that as you start to have different AI agents that can do different things, it's very clear that you'll start to have people that are paying AI agents salaries or people that are paying AI agents on a consumption basis. And there'll be lots of pricing arbitrage of how you want to wrap and package agents and agent SaaS, if you'll call it. And I think that we're still too early to do that arbitrage. And the reason is that you need to get to adoption first. You need to get to the usage first before you charge on consumption. Because the minute you charge on consumption or you charge on your number of agents and times of salary, you actually are disincentivizing the change that is required to build an AI-native company, an AI-native firm. And that's a bad thing. So...
Alex: I think in San Francisco, like everyone's talking about different pricing strategies for AI agents. And I'm like, outside of San Francisco, no one knows what an agent is at this point in time. So Hebbia today, it's a B2B company. You work with some of the largest asset managers you mentioned. As you think about where to build features, how do you maintain a great UI versus becoming the next Salesforce?
George: I commonly feel like a tension between just building what the customer wants right now and building for the future. Building the product that I think solves all the customer's needs. But maybe the Henry Ford quote of, do you want a car or do you want a faster horse? And everyone would always say they want a faster horse. And Salesforce is a faster horse. It's got components and a library for everything and everyone. And everyone at my company will tell you that when we prioritize features or when we ask for a feature, it's not something that we add on. But we actually look at the entire piece of software as a painting or as a composition and we figure out how it fits together. And we continuously redesign versus append. A lot of B2B enterprise AI apps will have a billion different apps. And they'll just be, here's this feature, here's this feature, here's this feature. It's already starting to look a little bit like Salesforce. We're taking a very different approach. We're saying, how does this feature integrate into this other feature? And how does that make sense to the user [inaudible 00:35:41.708]?
Alex: All right. We got to wrap things up. Final question. As we think about AI into the future, what does success look like?
George: For the industry writ large, there's a lot of models that are creating disinformation. They're generative and so they're naturally creating a lot of noise. And I'm very hopeful. And I think, I hope to be a voice of AI that doesn't only create information, but finds the signal in the noise or actually helps distill and make more concise and more cogent the information that we receive.
At the same time, I think you'll start to see the workforce change drastically. You'll start to see, one of my big predictions is that over 50% of the global GDP will be contributed by AI agents sometime in the next decade. And I don't think that's work that's replacing humans. I think that's actually net new value creation. I'm excited for that. And I think that you'll start to see humans, and my hope is that you'll start to see humans love work more. If you look at the jobs where people are the least satisfied, it's almost always these jobs where there are robotic automatons doing a task over and over again. And I think that humans, when they're taken out of the mundane execution and put in a seat of thinking and deciding and creating and discovering...
Alex: Yeah, moving from the objective to the subjective and focusing time there.
George: Exactly. I think that is my goal for the larger industry. Yeah. And I'd be happy to live in a world where AI saves 1% of the world's population 1% of their time. And if I can contribute to that, I think it'll be much more than 1%. I'll be a very happy man.
Alex: Prosperity. Thanks for spending time with us, George.
George: Thank you for having me.
Derrick Harris: With that, another episode is in the books. If you enjoyed it, please do rate and review the podcast or at least share it among your friends, colleagues, and network. We'll be back with a new episode next week.