The Exchanges

Q When you dive into how you go about implementing this, that first layer of data sources, what's the core set of information that you've wanted to train? And then where are there alternative data sources you've accumulated over the years?

A We started from the information that was hiding in plain sight, hiding because there was so much of it that was hard to get to the insights, even though they were available to every professional in the market. So all the SEC filings, global filings from every country with a stock exchange, earnings call, transcripts, conference presentations at every investment banks, and then of course, press releases, news, and then broker research, getting Wall Street research on a platform where you could now compare what is the company saying, what is an analyst saying about any topic, any company. And then that was still information that people could get access to on other platforms. One big step for us was acquiring a company called Stream, where they had built an expert transcript library. That's allowed us to start scaling and generating high value proprietary content that you couldn't get anywhere else. And we could really point that system to generate information on specific companies. What are their customers saying? What are their suppliers, partners, former employees, executives saying about things that really matter? Before this, you had to rely on what is the company saying? What are they putting out in the press releases or saying in public forums and filings? But you really had to go talk to management to question that or get alternative points of view. And the expert intervie…

AI assessment note: “all the SEC filings, global filings... One big step for us was acquiring a company called Stream”

Q What was the vision for what you first wanted to build?

A In the early days, we even called it, I forgot if it was us calling it, or clients first calling it, Google for finance, or Google for analysts. It really built a semantic search engine that would understand what an analyst really is looking for when they are reading a financial filing, or an earnings call, or a research report from Wall Street. We built that system that understood millions of terms and linked them to core concepts Understood that revenue is the same as top line across the whole vocabulary of finance. So as we built a system that was able to do all that and look at an earnings call happening in Japan or SEC filing in the US, all the information around the world, all companies using different vocabulary and reliably find every single data point about every single topic or theme that our analyst was researching. That was somewhat revolutionary at the time. That just wasn't available. People were still control of searching for individual terms one at a time in PDF reports. Shocking, but that was how work was done, and we had magnified the speed and efficiency and reliability of finding that information. That was the vision that it started to execute on, and the product we built got pretty quick traction in the hedge fund world where there was this strong thirst for information and efficiency and speed to insight. That I had experienced as an analyst, and we were n…

AI assessment note: “we even called it... Google for finance, or Google for analysts.”

Q So what you described, there's some building blocks of quantitative, publicly available information, then you have company reporting information, and then this huge library of what's called expert opinions. If you were in that use case, the hedge fund analyst, how would you describe to someone who hasn't used this system, what it is they're seeing so that they can pull out whatever information they want to pull out?

A Today, the easiest comparison point is something like a ChatGPT, where you're asking a language model, a human language question. The system is now able to precisely understand what you're looking for from your prompt, and now it goes across all the half a billion documents in our system and is able to find the most relevant ones and then dig deep into them and ask those same questions from every single document and see, is the answer here? Is it here? And do that hundreds of times, thousands of times for the most relevant documents for the user to look at and provide a narrative format answer that is granularly cited to those documents. A crucial difference to what people are used to with these chatbots we all use as consumers is that we focus on taking users to those underlying documents. Our users are serious professionals that care about reading and getting deep into the context that is Stated in a SEC filing or research report or expert interview, and we have made our user interface such that it's very easy on the same screen to see all those citations in the narrative format answer, but also dig deep into the underlying document and really understand the context and go deeper and lodge additional queries from there. So you can get a very strong confidence in what you're reading because you know where it's coming from. It's coming from high quality sources. You can see exa…

AI assessment note: “on the same screen to see all those citations in the narrative format answer”

Q What are some of the initial responses you give to someone who's asking you how to create a better prompt into AlphaSense?

A I'd go back to how do you ask an analyst that's working for you? How do you make sure that you convey all the information that the analyst needs to know so that you can be sure that your request has been understood? It's the same thing with the machine. If you keep it too vague, it might misunderstand the question or it may make assumptions that you don't like. So if you want to be very clear about what you're looking for, then it pays to be detailed in what you're asking about. But you can also be iterative. We've built our system so that you have different modes. You can ask questions in fast mode generative search where the average answer comes back in six seconds. There's a mode where you let the machine think longer. It takes maybe one or two minutes of chain of thought reasoning. It runs dozens of search sheets, synthesizes an answer, and brings it back. And then there is deep research where you let it work for 10:15 minutes, and it comes back with a 10 page report. And it's going to have gone much deeper. When you're doing that longer cycle work, you're going to be more careful with your prompts. You don't want to wait and then realize that you weren't precise enough. But when you're iterating quickly on a six second cycle, it's cheap to ask lots of questions. So if you don't think the first question got there, then you can ask again. We feel it's our job to make sure we…

AI assessment note: “I'd go back to how do you ask an analyst that's working for you?”

Q So business started mostly with hedge funds. You did mention, oh, there's a little different use case for corporations. How has it evolved from that initial user base to who your customer base is today?

A We always had the idea that one day the corporate world should like this too. They need information. They're making big decisions. They're deploying capital. They're acquiring companies. They're making investments and launching new products, entering new markets. They should need this same information. And that was the thesis. We learned that at least this thesis played out well in investor relations at first. They were hearing from hedge funds that they're using this new great tool called AlphaSense. We started to really spread like wildfire through word of mouth in the investor relations community across public companies, and then started to map out all the other big pockets of knowledge workers in those companies from corporate strategy to competitive intelligence to corporate development to strategic marketing, product management, even engineering these days, and then going back to the CFO's office and across the C-suite It's really gotten across dozens of personas within corporations where they are just trying to be as much on top of the information as the investment world is, but more narrowly focused on their industry or different forces affecting where that industry is going. That's today pretty broad, diverse landscape of corporate users. And of course, then everybody else that's in the knowledge worker universe, from consultancies to bankers, There are similarly now u…

AI assessment note: “We learned that at least this thesis played out well in investor relations at first.”

Q What's the process for figuring that out? Which model is the right one for the right task?

A It's a very systematic engineering led process where you're really just testing all of them all the time. We have teams of people evaluating outputs. We have LLMs evaluating outputs. So there's a system that keeps running standard checks. And when you have a new model, you can compare it to the baseline and see, okay, what does this new black box deliver? You can't evaluate it until you've tested it. You kind of look at the external test results and there are some of these Metrics, some real metrics, some vanity metrics, hard to really draw conclusions from what you read. You have to just deploy them and test them and see what is the effective performance. Try to test numerical metrics and see how often or what percentage of the time do human analysts agree with the output. But ultimately, there's also a style test. Do I like the way that this LLM speaks? Is it too verbose? Is it speaking specifically in a finance type language where We've ended up even giving financial services, the buy side, a different model, and the corporate world, a different model. We've learned that there are different stylistic references as to how you want the LLM to be speaking to you. So lots of aspects that you have to test, and some of them are qualitative in that way.

AI assessment note: “systematic engineering led process where you're really just testing all of them all the time”

Q You've done a series of acquisitions along the way. How does that fit into the business strategy for AlphaSense?

A There is no specific acquisition focus. We don't feel like we need to do acquisitions. Everything is somewhat opportunistic as to what is out there. Now, if there was another Tegels-like asset, I'd be very excited. These are few and far between. This was quite perfect, and it made a lot of sense, and it was a huge bet for us, but we had a lot of confidence that this is the right thing to do. We made a much smaller investment at first, and then we're able to make this very big investment, almost a billion dollars that we deployed in that Tegus deal. But we felt that as this sort of intelligence factory, the more content and proprietary insight that you feed it, the more value you're able to deliver to customers. So it felt If we have a great user interface and it's the better interface to deliver this qualitative content, then acquiring and plugging that content and data into the system is going to make it so much smarter that the combined value proposition is one plus one equals five. And that's what we have seen. So that is the thesis. I'd say we're forced to be opportunistic because these kinds of fantastic content assets aren't available all around. They are pretty rare.

AI assessment note: “Everything is somewhat opportunistic as to what is out there.”

Q You started this business hoping to solve this frustration that you had with the efficiency of gathering all financial information for the decisions you need to make. There's so much evolved over the last 17 years that you've been at the front foot of increasing that efficiency. What excites you about what will happen for the next 17 years?

A It's certainly hard to see that far, but it's a very exciting time given what this technology revolution with AI and language models has enabled. It's an incredible time for creative building. As a startup founder, CEO, I'm sure many in the same position would share this. The most fun thing to do is just to wake up every morning and think about what can I create next? And there's just the incredible sandbox of what we can do to create the next generation Indeligent machine that the whole financial and business world can use and be smarter and make smarter decisions when you're making big bets and be more agile with better confidence, better data. So it's a very fun place to be building all that. As I see, not that far forward, but into the next months and years, it's about building this always on machine that is working for all of the investment firms Public and private markets and banks and consultancies and corporations across every industry. If you think about every user having kind of thousand analysts in their pocket to think about AlphaSense that way, well, what if all our clients had that available through our system? That means our system is churning through this machine intelligence day and night, every minute, and running things on their behalf when they ask, even when they don't ask. If you're a buy-side firm, you've got a portfolio, well, Our system can be monitorin…

AI assessment note: “it's about building this always on machine that is working for all of the investment firms”

Q What were your frustrations at the time doing the work you were trying to do without the tools needed?

A I had a pretty strong work ethic. I'd work the usual all-nighters and trying to do a really good job doing the analysis, but the pace was really fast and you'd end up not having enough time to do a really good job. I remember some client boardrooms where I'm sweating and barely awake, but also afraid of what did I miss and what am I going to be called on by the CFO or CEO of that company where I just missed something in my analysis. And that feeling has stuck to me, frankly, still every day when I walk into a boardroom, I have flashes from those situations. That really was because of the lack of technology to help an analyst who needed to consume so much information and trying to catch up on a new industry that you didn't know and new companies you didn't know and the sort of cross-sectional information across different industries that you really should have known to be smarter with your analysis, with your viewpoints. And to be able to talk to some really experienced business professionals working on about to bet billions on a deal. Certainly we had the data terminals that a refinance professionals has still today. That was a big part of the frustration that you could go and manually look for data, but it was very hard to consume it at the scale and speed that was needed. And back in those days, you already started to have technology for consumers where I had Google to search …

AI assessment note: “That was a big part of the frustration that you could go and manually look”

Q As you worked with the LLMs and as an example of building the AI interviewer, what have you seen as some of the challenges that you had to overcome to make these work the way you wanted them to?

A There are lots. If we're thinking about technical challenges, it starts from what is the leading edge LLM for this particular task that we're trying to solve? And can we actually get it to do that work? And what is the right combination of can it absorb enough context? In this case, the AI interviewer needs to read a whole deep research board of maybe dozens of pages in some cases to really get expertise on any topic. Can it hold that in its memory and also take The speech of the expert it's interviewing, and then also when it hears something unexpected, can it do quick research and now adjust what's in its context and pivot and ask a new better question that took a lot of iterating. You've got to test different models. You've got to test different configurations. We've built this, our system as kind of an LLM agnostic orchestrator that is working with just about every one of the leading edge models in the market. And we're deploying them in different ways. And depending on the task at hand, you'll end up using this model now and maybe another model in a few weeks when a new breakthrough happens. So the team has had to build these capabilities of tracking and testing and very effectively staying up to speed on every new breakthrough to understand where is the right combination of these things. There isn't one perfect model out there as models that have been optimized for given …

AI assessment note: “what is the right combination of can it absorb enough context?”

Q What was the best advice you ever received?

A One thing that I have been told, and I dismissed it at first, it sounded fluffy. Follow your passion. If you're going to be an entrepreneur and launch a company, do something you're passionate about. I've actually learned that that was incredibly good advice, having spent a decade and a half on this company. If that wasn't true, if I was building some, I don't know, some accounting software, I can't imagine being equally passionate about it, but I'm doing something that I wake up excited about every day, and I have this personal passion to go and solve this problem that I still feel in my bones thinking back to my analyst days and feel I can go and help the whole industry do these things so much more better and more efficiently that it keeps me going. So following the passion actually has turned out to be really good advice, not at all fluffy as I thought at first.

AI assessment note: “Follow your passion. If you're going to be an entrepreneur and launch a company”

Q I'm really curious about the unit costs of the business in the sense that some of the data sources you talk about are probably data sources you have to buy And then on the other side, you're curating and providing this information that's essential for huge decisions. How do you figure out what someone's willing to pay for that?

A The biggest variable today in that is how much intelligence do you apply and what does that cost? More steady pieces of that apply to any data business, but the raw LLM intelligence and chain of thought reasoning, if you take that to the Maximum of where that is heading, you're gonna have a system running 24 seven and running processes both initiated by users, initiated by APIs that clients are running. Clients are building agents for their own internal workflows, triggering these generative search and deep research reports through those APIs. And then the system is initiating work to generate more information. Where that is heading is recognizing What information gaps exist and going through our expert network and finding experts and launching calls and bringing back new information into the system. So there's this sort of intelligence factory that processes millions of tokens for every decision that happens in these financial and business workflows. We are having to look at what benefits the overall system. What can we amortize across all customers? Where is that token usage so high for individual client that we need a pricing model where They're leveraging the API in a massive volume, and they get a lot of value. That is a very evolving world right now where you have to do whole new kinds of math and estimations on what that intelligence is going to cost, how that cost is ev…

AI assessment note: “Where is that token usage so high for individual client that we need a pricing model”

Q As you look at the business today and where you're headed, what are the most important metrics that you're reviewing to gauge your progress?

A There are the standard financial SaaS company metrics where the recurring revenue is probably the primary one that you stare at and kind of have big targets. And beyond that, we're fortunate that we've been able to build a business with great SaaS metrics that we feel both private and ultimately public market investors will really appreciate. So it allows us to use that spectrum of metrics and make sure that we're gradually tuning and turning the right knobs to Improve the performance, but there isn't anything that we feel, oh, we've got to really, really focus on this. We have good growth margins, unlike some companies leveraging language models that have big challenges there. We are able to deliver enough value that we're able to continue investing a lot in the intelligence and the token capacity and still absorb that cost and not worry about our overall metrics. Beyond revenue and growth, which is primary, to me, it's all about growth. We were accelerating our growth every quarter and that's pretty rare. It's great to see in this environment where language models have both created a lot more client demand and create a lot more value in our product. So there's this great tailwind and momentum. If I was going to crystallize it down to one thing, it's about growth because we're trying to build something really big and getting there faster is the primary objective.

AI assessment note: “recurring revenue is probably the primary one that you stare at”

Q What's been most surprising to you in this process of testing the LLMs?

A What's been hard to do is knowing where the cutting edge is. Knowing what each model is capable of in practice. Even the system, you can't really build it, design it, build it, it's ready. No, it's a very iterative process. You have to go and keep evolving it and seeing how much human evaluation you could do, how much LLMs can actually effectively evaluate each other, and that changes as their capabilities evolve. It's a process that keeps you on your toes. You can't Claim to master it at any time, or if you feel like you've mastered it, some new breakthrough happens, and now you have to re-challenge your assumptions again. Learn that we just have to have teams that keep on doing this, and we have to be ready to pivot when something changes. That readiness to pivot and the flexibility is perhaps one of the bigger learnings from this, that you just have to be on top of this all the time and invest the time, including myself. As the CEO, I got to understand what's going on These are critical choices for our product. Product is what adds value to users. I feel like I need to be reading a lot and trying to have tentacles through our team to making sure I'm up to speed and that applies to everybody that's part of that chain. It feels like an around the clock, very intense process of staying up to speed with all the development.

AI assessment note: “What's been hard to do is knowing where the cutting edge is.”