Can AI Understand a Distressed Credit? - Harneet Kaur, Axim Intelligence
https://www.youtube.com/watch?v=WOszaC1WWbQ
AI Engineer · 14 min · 148 sentences · sentence-level transcript from YouTube's caption track; each line begins at the time shown

[00:13] Thank you to everyone who came.
[00:14] I am very pleased to deliver this report.
[00:18] So, a brief introduction about me.
[00:22] My name is Harnit.
[00:23] I am the founder and CEO of Axon Intelligence.
[00:28] At Axon, we create an intelligent layer for investing in distressed loans.
[00:33] This is also what we will talk about today.
[00:37] The title of my talk: Can AI understand problem loans?
[00:45] I think the first question that arises for everyone who is not familiar with loans is: what does a problem loan actually mean?
[00:55] Credit is pretty simple.
[00:57] It is simply a bond or loan from a corporation.
[01:01] A Coca-Cola bond is a loan.
[01:06] And what does it mean to be under pressure or in a difficult situation?
[01:10] Is n't that right?
[01:14] Pressure arises if a company has some underlying risk that could cause them to run out of money or be unable to repay their debt.
[01:25] Every year, a certain portion of the credit market faces some level of pressure or problems.
[01:34] This may be caused by cyclical problems.
[01:38] Let's say the construction market is in decline, the housing market is in decline.
[01:42] This may be caused by operational issues.
[01:45] For example, if you are a restaurant business and there is a large-scale product recall.
[01:49] This may also be caused by long-term structural changes .
[01:53] If you are in software today and are facing the impact of AI.
[01:58] Some companies overcome difficulties and recover.
[02:01] Some companies face challenges and find themselves in difficult situations.
[02:06] What does a difficult situation mean?
[02:08] In reality, it's not that simple.
[02:10] Bankruptcy is one of the consequences of a difficult situation.
[02:17] Liability management measures are also possible—I can explain these terms in detail, but just know that this is another option, as well as an out-of-court restructuring or an agreed exchange.
[02:30] So, next slide .
[02:32] As you can see, it's not that simple, is it?
[02:39] Historically, the issue of predicting problems has been quite binary.
[02:43] Bankruptcy could be predicted : either you predict default or you don't.
[02:51] Today it has become much more difficult.
[02:57] Many traditional machine learning methods simply can't keep up with this, nor can the assessments that were important 10, 15, or 20 years ago.
[03:10] Um, uh, the main problem is that Altman's Z-score has become very static in a situation that changes, you know, sometimes monthly, quarterly, or annually, and in a much more opaque
[03:22] market.
[03:24] So, if you've been following the credit market, the growth of private lending and some other sectors has led to, you know, a lack of pricing.
[03:33] On the other hand, there is a very high-quality method of understanding and analyzing problem loans, when you, as an investor in problem debt, can analyze financial statements, understand what is happening on the part of the lender, and read loan documentation.
[03:49] Basically, you know, all the things a typical hedge fund or private equity investor would do to understand credit.
[03:57] This process is extremely manual and, of course, very difficult to scale.
[04:01] Isn't that right ?
[04:03] Now the question is , you know, part of the founding of Axon and the question that I've been thinking about is how far AI can go in understanding distressed
[04:16] credit and what paths can be taken.
[04:19] And is it possible to systematize what is actually a rather complex process, right?
[04:27] So, what does it mean to understand a non-performing loan, right?
[04:34] As investors, we have many questions that we are trying to find answers to.
[04:41] Things you think about from a financial perspective, right?
[04:47] Do you know what liquidity is?
[04:49] What is the capital structure?
[04:51] And then there are legal questions.
[04:54] How do you know what the structural priority of the debt is ?
[04:57] Isn't that right?
[04:58] That is, do you have a higher or lower priority?
[05:02] And then there are operational questions, like: how do you understand business?
[05:05] What are the revenues there?
[05:07] Will revenues go up or down tomorrow, based on fundamental, you know, fundamental business analysis.
[05:14] Is n't that right?
[05:16] Right now, what we're seeing in the market, if you follow Pro Go, Harvey, or some of these big AI companies, they're actually thinking about how to compare all these different
[05:27] issues, right?
[05:29] For example, how well does AI answer the question of which portion of debt is structurally senior?
[05:37] And then do thorough comparative testing and make it, let's call it, a big financial benchmark.
[05:45] But I wanted to think about it a little differently, didn't I ?
[05:50] Because I was thinking this way: the real understanding of credit, especially distressed credit, is whether you can predict what will happen next?
[06:00] Isn't that right ?
[06:02] Can I figure out and get a probability distribution of which way this will go in the future?
[06:08] Um, so I thought this was a pretty interesting exercise.
[06:15] The first step, you know, what you should do is take all the, you know, I chose the company.
[06:20] So, I chose iHeartMedia, a company that filed for bankruptcy in 2018 .
[06:27] So she, you know, she went through all these bureaucratic hoops.
[06:31] Um, she also had another stressful transaction in 2024.
[06:36] So this is a real-time situation.
[06:39] This is happening today.
[06:41] Nobody has a real, you know, answer to what the future holds, right?
[06:46] Um, there is no absolute truth.
[06:49] But what I did was this: let's throw this into the smartest, you know, models.
[06:53] I used Astra 6 on Max Intelligence and Opus 5.5. And the conclusion that I found very interesting is that they disagreed.
[07:06] They didn't agree on what the probabilities of the future would be, given the structure or context of the data, right?
[07:17] And so, if you look here at the agreed restructuring, Astra thought it was the best option for bankruptcy, and Opus thought it was the best option, right?
[07:28] Now, how do you tell who is right and who is wrong?
[07:34] You can't, because I don't know what the probabilities are, right?
[07:38] I don't know what will happen.
[07:40] I don't know if it will be a bankruptcy.
[07:42] I don't know if this will be an out-of-court restructuring.
[07:45] I do n't know if this will be a refinancing.
[07:48] I have my own judgment on these things.
[07:50] I would say I'm leaning towards Astra 6, but, you know, we're all human.
[07:56] And that's a pretty interesting conclusion, right?
[07:58] Like, okay, wait, they disagree.
[08:03] Um, so the second thing I thought about was: okay, how can I at least improve their process assessment, right?
[08:11] So I can see whether I agree or disagree with different assumptions.
[08:18] Um, and so, essentially, you've created a kind of benchmark where we test every level.
[08:29] Obviously, mistakes can be made at every level of the LLM, right?
[08:33] Regarding the level of data extraction, I would say that with public companies this issue is already pretty much resolved.
[08:39] I would say with private loans, and also when you have Excel and PDF documents, it can still be a challenge, right?
[08:47] Then you have the reconciliation or data output level.
[08:50] So, extraction is, for example, obtaining the EBITDA indicator.
[08:52] Uh, the conclusion would be : okay, now you have EBITDA and you have debt, so you have a credit load.
[08:58] And then you have judgment.
[09:00] Okay, now you have a credit load.
[09:02] Is this a problem loan or not?
[09:07] What is your Oh.
[09:19] Uh, is there any...okay, I'll just, I'll just keep talking.
[09:27] So, uh, actually, the question is , is it like my computer has shut down?
[09:43] Uh, excuse me, just one second .
[09:47] I'll tell you I have a charger, but it's not here.
[09:57] Uh, yes.
[10:05] Okay, I'll just, I'll just talk.
[10:08] This is normal.
[10:13] Oh, okay.
[10:14] It works.
[10:19] OK.
[10:21] So, therefore, you have a level of judgment , right?
[10:25] And then, uh, then, then you can start, okay , if you can evaluate each level, right?
[10:32] When you get to the prediction level, you've made sure the data points are right, you've made sure the, uh, level of agreement is right, you've made sure there's some kind of reasonable judgment,
[10:45] right?
[10:46] And then , once you're confident in those things, then you can move to the forecasting level, which is really how, you know, you as a human analyst think about all of these steps, right?
[10:57] You, you, if you base a judgment on a bad data point, it's not going to be a good judgment, and that's, that's just how it works, right?
[11:07] So, uh, Op.
[11:14] Here we are.
[11:17] So, overall, what we found, and maybe this isn't surprising, but, uh, I think it was interesting, is that once we created a structured reasoning process, you , you actually got both intellects,
[11:29] both models, to actually converge on what they thought the future probabilities for the business would be, right?
[11:42] So now, instead of talking about a 62% bankruptcy with Opus and a 25% with Astra, now both of these models say that there is actually only a quarter, or a 25% chance of bankruptcy, and both models agreed that the path of concerted
[11:56] restructuring is actually the more likely path for the business.
[12:02] So, what are the main conclusions here, right ?
[12:07] Uh, one, I would say, first conclusion: if you give questions that cannot be verified , right?
[12:14] With the same structured context, organized context, as humans, I mean , do these models have an advantage in prediction?
[12:25] And you have to say "no", right?
[12:28] Because they are, in fact, even agents— they are people, right?
[12:33] In some sense of the word, and I won't get into an argument about that .
[12:37] Uh, the second thing is, but a structured reasoning process can force two models or, you know, different intellects to actually come to the same vision of the future and a similar understanding of the business and the path it
[12:55] can take.
[12:58] Hmm, but is this the right way ?
[13:02] We don't know, do we ?
[13:04] Um, and is that really the single most accurate measurement, right?
[13:07] Otherwise, any historical analysis would create a certain distortion by predicting the future.
[13:13] Um, but I think the really interesting thing that I was thinking about is that convergence is also quite interesting, right ?
[13:22] Because if you can systematically build a process that actually converges to the same probabilistic outcomes.
[13:30] This, in fact, is the market price.
[13:33] Um, and that's why I thought that was an interesting conclusion.
[13:38] Um, so this is the last one.
[13:42] And now briefly about me.
[13:43] So, my name is , as I said, Harini.
[13:46] Um, building an Axon.
[13:48] Um, a little bit about my background.
[13:49] I spent 10 years in finance, um , the last few years as a distressed and special situations investor at a, uh, private equity/ distressed asset hedge fund.
[14:00] Um, here's the QR code if you're interested in Axon.
[14:04] Um, and that's all I had.
[14:06] Thank you.