Anastasia Kinsky: We are here in Lisbon at the Alpaca Partner Summit, and I am joined by Braden Dennis, the CEO of Fiscal.ai. Braden, thank you so much for joining us.
Braden Dennis: Anastasia. Thank you for having me.
Anastasia Kinsky: So tell us a little bit about Fiscal.ai and how you guys are sort of disrupting that data infrastructure part.
Braden Dennis: Yeah. Fiscal.ai is a financial data provider. So we help fintechs and buy-side with accurate, source-linked financials, the earnings calls, the slide decks, everything that they’d want in the complete picture in the public domain of information that they want to consume before making any sort of investment decision. So for fintechs, that’s, you know, providing it to their end customers, and then directly with our buy-side clients. And now with AI, this is very rich content for all types of the research process today.
Anastasia Kinsky: Totally tasty. And tell me a little bit, you know, in a world where, as a retail end user, I might be able to slightly sidestep the fintech space and build whatever I need to build by myself, what do I need to turn to people like you for, even the fintechs who are using your services? Like, what is the differentiator?
Braden Dennis: Data as a service and accurate financials is a tip-of-the-iceberg problem, right? Companies report their financials as reported. Each company is its own special snowflake. It reports completely different business segments. They have different industry templates that you might want to consolidate into. They have different accounting standards that they report into, and different languages, and all of these different things, including the actual source material that they disclose the information from, whether you’re in Japan, UK, Canada, US, all over the world. There’s all of these edge cases. And the edge cases start to build up. And then we’ve built this taxonomy of thousands and thousands of evals and tests and decisions that are very important to [unclear: “create the handle all edge cases. Well, yeah. Data structure.”]. So people come to us because, including OpenAI, like including the large language models are coming to us directly and going, this is a very hard problem to solve even with AI today. And that’s why they come to us for trusted information.
Anastasia Kinsky: Yeah. So central. And, you know, I can imagine as a stack, between Alpaca and Fiscal.ai, you have so many of the [unclear: “founding the blocks”] that you need exactly to build on. Tell me a little bit, because there are ways of differentiating this when it comes to AI and what you guys are providing. You are not using a specialized AI harness. Is that correct?
Braden Dennis: Yeah. So there’s kind of two ways that asset managers are purchasing investment research tools: like AI harnesses, or going directly to the models themselves and using those tools. And so the question is, like, do I want to build a specialized harness or do I want to use something out of the box? And we used to actually have our own specialized harness, right? And we made the decision to use our data connection because we own the content, [unclear: “we of the IP”], we can let our customers get it in the feed, but we can also let our customers connect directly in Claude, connect directly in Copilot, connect directly in Codex from OpenAI, for example. And so that actually produced much better results. And every new model, we’re always at the frontier. And so we can focus on what we’re good at, which is giving people accurate, source-backed financial content. And then people don’t need to change their screen when it comes to the AI tool. And, you know, the numbers for our business were much better. Customer retention much better. Satisfaction. Every number up to the right. And so it just made complete sense. And now there doesn’t need to be some new purchasing decision for these firms. They’re already working in these tools regularly. And now they get something that they know is going to be right.
Anastasia Kinsky: Yeah. Interesting to see how that appetite matured into something that, you know, you guys have provided, just the Goldilocks of this service. Talk to us a little bit though, you know, as you say, that accurate advice that you guys are giving, there are certain narratives that actually, you know, how accurate can AI be? Do we have pitfalls where actually the advice isn’t accurate but we can’t see where it is? Tell me a little bit about how you guys mitigate those risks.
Braden Dennis: Yeah, we say it is better, faster, cheaper than manual aggregation, right? Better in terms of quality. So more accurate. I’m stuck in the mud at about 1 in 100 numbers being wrong, even through a human QA process. It’s natural, especially when we’re talking about reported financials being in the billions and sometimes hundreds of billions. A lot of zeros and a lot of digits. And so can we actually get to 1 in 100,000 numbers, or asymptotically reach perfect? And the answer is yes, because now I can actually do agentic QA.
Anastasia Kinsky: Okay.
Braden Dennis: Now I can have agents pull the data and agents verify it many, many more times and many more iterations than just a traditional QA process. So that’s better. So can we do better? Faster? Cheaper? We can do better. Faster is the easiest to explain. Typically, the incumbent providers take three days to kind of consolidate the financials. We do it in three minutes. Right? So especially for the fintechs here, they really care about giving their end users, when they see the reaction in the market of a company reporting earnings, they want to see the latest quarter’s data, not last quarter’s data. And so that’s been huge for a lot of our fintech partners as well. And then cheaper. That’s a language we all understand.
Anastasia Kinsky: Absolutely. Absolutely. But tell us a little bit about the audit trails and the transparency that you guys provide, because I know that’s very, very important to what you’re doing and how you differentiate yourselves.
Braden Dennis: Absolutely. So on our platform, in the MCP connection and through our API, what we do is, if we say the number’s X, it comes with a paper trail of a link to where it came from, in, like, the SEC, for example, an image screenshot that the agent took of where it came from and the exact page in a PDF from the file that it came from. So all the different sources. And then our customers can choose which paper trail they want to show end users or have internally for their own use case. And so say I hover over that number and I see it’s X, and it’s actually highlighted in their statements.
Anastasia Kinsky: Where they.
Braden Dennis: [unclear: “Influence”] where it is in that table. So I’m not having to go search around.
Anastasia Kinsky: And try to understand how you understood that something. [unclear: “Understood something.”]
Braden Dennis: And every time, like, just before this, I was looking at a number that these are reported. Obviously, you know, the credit card company. And it was 40 billion on the [unclear: “.4000”]. And then like all the zeros. I’m like, that looks like a... It’s like a wrong number.
Anastasia Kinsky: Really hard to [unclear: “come up with that is specific.”]
Braden Dennis: And I hover over in the statements, and yeah, that’s the number that was reported in their quarterly financials. And so these types of things are important because I’m in the business of show, don’t tell, when it comes to data quality. I can tell everyone how great the data quality is. But you want to be able to show that. You want to be able to show evals.
Anastasia Kinsky: It’s so refreshing to hear, because actually I think that’s what so much digitalization has been missing. And this kind of new era of transformation is going to need that transparency to ground the trust in it. So thank you so much, Braden, for talking to us about that, and I hope you enjoy the rest of the summit.