Bedrock AI CEO Kris Bennatti on BNN Bloomberg: How AI Detects Red Flags at Hedge Funds
May 11, 2023, Repost with Excerpts from BNN Bloomberg Morning Markets -- Kris Bennatti, CEO of Bedrock AI, joins BNN Bloomberg to discuss how funds are using AI to buy and sell stocks, with successful stories in catching red flags on CoinBase, Credit Suisse, Lift and others.
Bloomberg: ChatGPT went viral at an unprecedented pace. It made AI a household tool. But businesses are also rushing to figure out how they can make it work. Hedge funds and money managers are no exception. A survey predicted that 9 out of 10 hedge fund traders will use artificial intelligence to achieve portfolio returns this year. Well, our next guest has a product that does just that.
Bedrock AI is built to detect red flags and help warn investors before we see huge plunges. And we've seen it work in the case of Coinbase and even Credit Suisse. Their software was detecting issues before big problems started to emerge in the market. Joining me now is Kris Bennatti. She's the CEO of Bedrock AI. Thank you so much for being with us.
Kris: It's so great to be here. Thank you.
Bloomberg: You're using AI to help unearth trends more often than not red flags. You're trying to prepare investors for the downside versus looking for something that might outperform. Am I right?
Kris: Yeah, so that's correct. We focus not on generative AI. We do use generative AI in very, very specific small use cases. But we're focused on language models that are expert in the financial domain. We've trained our models on terabytes of SEC filing data, et cetera.
Bloomberg: So that's what they're doing. They're crawling all over these filings.
Kris: Crawling these filings and focus on the understanding task rather than the generation task.
Bloomberg: So, let's use an example. We mentioned Credit Suisse. Obviously we know what happened post Silicon Valley Bank, but your models were flashing warnings before any of this happened, on what basis?
Kris: Yeah, so our flagship product, it's called Bedrock Forensic Intelligence. Our focus is on identifying narrative information in these filings that empirically has had an association with downside events, specifically related to earnings management malfeasance and fraud. So, we put out a piece in 2021, highlighting algorithmic red flags at Credit Suisse. We also provide overarching risk scores and Credit Suisse was in the top 7% of highest risk companies at that time. It has been since. But we were at the time highlighting all of these really severe red flags, especially severe in comparison to other large…
Bloomberg: On what metric?
Kris: So we're looking at a number of different things. In the case of Credit Suisse, pulling up 20 times more high severity red flags related to legal and regulatory issues that were indicators of potentially poor management integrity, a lot of red flags related to money laundering, bribery, corruption, scandals. FIFA bribery, Mozambique tuna fishing, spying scandals, all in the footnotes.
If you're covering Credit Suisse, you see scandals popping up here and there. You see occasionally a scandal at JPMorgan or Citi, but you're not necessarily paying attention to how the frequency and the severity of some of these other scandals. But then we're also looking at more sort of traditional accounting forensic type red flags. So at Credit Suisse, they were booking one-time gains through sales, including a one-time gain on a sale to an affiliate. So a related transaction, quite large outstanding loans, a $17 million loan to the CEO of their Pacific's business. So these aren't normal. So using algorithms, we're seeing all of this being disclosed, all out there. And then the overarching risk score that we also provide was saying, hey, this is very abnormal.
Bloomberg: One stock we were tracking is Icon Enterprises, which has been targeted by a short seller? Is this something that you could kind of put into your model and it would spit out kind of a red flag?
Kris: So we process SEC filings in real time and have been doing so since we were founded. So we have a history and if you go onto our platform right now, you can see that we've had a high score for ICON enterprises for a long time, in addition to many high severity red flags.
Bloomberg: And things like auditor switching, right? We often look at that, sometimes people just change auditors, and sometimes when they change auditors, it signals an issue. Is AI complex enough to understand the nuance or the difference, or how do you account for that?
Kris: Yeah, our models are more or less trying to replicate the brain of a forensic equity analyst. So they are looking at information like that. Obviously not all of the context is disclosed, but what our models are looking at is they're looking at changes in auditor, they're looking at control issues, they're looking at changes in presentation, other types of management turnover. And as all of those types of things tend to add up or snowball, when you see all of that at once, you can start predicting disaster.
Bloomberg: One of the stocks that we've recently talked about is Lyft. And you mentioned to me in the break that Lyft kind of just got like a little bit of a flash red flag. What was that?
Kris: Yeah, one of the reasons that that's interesting is because I haven't seen it reported much in financial media. I guess Lyft is less important these days, but they disclosed an SEC investigation in their most recent 10Q. That's not something that you're going to be actively looking for, but it's something that we flag as the number one thing you should be paying attention to in Lyft's 10Q, and it's also part of our newsfeed, our AI-generated newsfeed on our homepage.
Bloomberg: Lyft, of course, as you mentioned, people have been focused on other things, that they're losing market share to the likes of Uber.
See the BNN Bloomberg Full Interview