AI-Powered Trading Alerts, AI and Data Analytics, Credit Trading
AI-Powered Trading Alerts, AI and Data Analytics, Credit Trading

Shut Out the Noise: A New Era in Credit Trading

I’m here for one reason and one reason alone. I’m here to guess what the music might do a week, a month, a year from now. That’s it. Nothing more.”

This quote is one of my favorites from Margin Call for two main reasons. Firstly, it cleverly alludes to Chuck Prince’s infamous comment about dancing until the music stops. As someone who formerly worked at Lehman Brothers, I must admit I enjoy a subtle jab at any institution associated with the ’08 Financial Crisis. Secondly, it encapsulates the ultimate goal of every trader or portfolio manager: anticipating market movements.

This two-part commentary is not about a new product release – it is about how I envision the evolution of AI in corporate credit. 

When I was in the financial world, I spent the vast majority of my time on data acquisition, monitoring and analysis. That took various forms including:

  • interpreting streaming information on Bloomberg
  • reading news articles and financial/analyst reports
  • creating and looking at various graphs and charts

But it was generally a filtering exercise in managing messy and inefficient data flows. The data I received was noisy and unfocused, it was often unclear how it may apply to my real revenue generating opportunities, and the tools I was using were poorly designed for the credit world…yet, missing out on any detail could mean overlooking significant opportunities.

The real question is, how can AI free up credit professionals in their day to day lives to “listen to the music of the markets”?. This involves looking at where we are now…and what the roadmap is to make every person feel like the CEO of Margin Call, with carefully curated and analyzed data at your fingertips for any decision you want to make next.

Current State of AI in Credit

My initial two years as an investment banking analyst provided a tangible analogy for the current state of AI in credit – and how quickly things will move over the next several months. Let me walk through how coverage of a new public Company would work to demonstrate that the workflow and data analysis within Investment Banking teams forms a nearly perfect example of a pyramid.

To start, an analyst would build a detailed financial model by slogging through all company financial reports and the underlying notes associated with those filings. The goal of this exercise is to properly identify and isolate all relevant drivers of the underlying business on a historical basis, which is generally a painful exercise involving extensive manual processing. If done properly – it should be noted that this is not your typical “Yahoo Finance” style set of financial statements – the number of hours spent to create and maintain a proper model is shocking. I remember working on the SBC/BellSouth acquisition in 2006 and the “owner” of those 2 models worked 100+ hour weeks for the entire year – think thousands of rows and hundreds of linked sheets -err hardcoded.

The ideal financial model should be extremely detailed and robust – breaking down the business as granularly as possible to “understand” historical performance and project future results. It should also be extremely flexible (to the extent possible in Excel) and setup for a wide range of scenario analysis. This model is the “foundation” on which all advisory work on that Company is built and it is by far the most time intensive task. At the end of the day though…the analyst is effectively a specialist in retrieving data, discovering the relationships within that data, and using those factors to project future business performance. It should be noted that this requires the integration of economic data/forecasts, company projections, and comparable business analysis on a historical and future basis in order to make the model as accurate as possible.

Deep Market Making’s Solution

Deep MM’s FMV(Fair Market Value) Model is our “Financial Model” for corporate bond pricing. Building it properly took our CEO over 6 years and it forms the “base” for all of the analysis that we will begin to layer on top. 

Why is our model unique? It is AI/Deep Learning driven and uses over 80 million parameters (tested up to 3.3 billion) and will ultimately include ALL relevant data to provide the most accurate price estimate possible. That includes publicly available sources such as TRACE, equities, rates, news updates and company filings – and on a customized basis, “proprietary” sources like messages, inquiry and trade history. Even better – it is an analyst on steroids, updating in real-time for every piece of data that comes in.

There are a few other “real-time” pricing models out there currently – but I would argue they take the “Yahoo Finance” approach in reaching their FMV. They are “AI” based, to the same extent that it could be argued that a chatbot 20 years ago was “AI” (compared to ChatGPT today). These models use a one-dimensional approach to pricing, using only TRACE and some “proprietary” inquiry data, which has limited value longer-term. People on the sell-side love to say “go trade with TRACE” when a client is using the same approach to real-time pricing: it can work in a non-volatile world, where relationships are static, but it misses the depth and complexity of what is driving pricing on the next trade. 

In the context of my investment banking analyst analogy – you could use basic financial results from Yahoo to understand historical performance….or you can have your AI analyst comb through every piece of data possible in order to truly understand the underlying relationships and business. Investment banking teams build their own models for a reason – your “foundational” data needs to be as accurate, detailed, and flexible as possible in order to move to the next stages of the team hierarchy. 

If you would like to learn more about AI in corporate credit, please email us at or hit the Book Demo link to find a time to discuss.

Book Demo