Demand grows for Generative AI tools on the desk

Published on 10 July 2024

It’s been a bumper 6 months of 2024 for fixed-income trading volumes and primary market issuance—and equally elevated activity is expected for the second half of the year. The higher volumes have created a greater need for more sophisticated tools on the desk to help consume and analyze these large data sets.

Trader TV spoke to AJ Cass and Mary Beth Sweeney at LTX to discuss new developments in generative AI solutions, like LTX’s Bond GPT, including how they can be leveraged across the front office and how buy-side comfort levels have evolved when it come to the tech, particularly if models are using client data.

Interview Transcript

Jo Gallagher Welcome to Trader TV your insight into institutional trading I’m Jo Gallagher. Today I am joined by AJ Cass and Mary Beth Sweeney to discuss how generative AI solutions are transforming the trading desk. Welcome to the show.

AJ Cass Great to be here. Thank you.

Mary Beth Sweeney Thank you for having us.

AJ Cass I think the current trading environment, certainly for 2024, has been relatively healthy. Yes, we are at a place with higher yields for a little longer than most folks thought, but the environment has been very well supported by yield buyers and yield buyers for longer. But I think the strongest point around trading this year has been received from the new issue market. So new issues as of the end of April are almost $750 billion in new issuance and that’s between IG and high yield and that’s up about 38% year over year.

We’re also seeing daily volumes up well over 40 billion and ticket account averages are up to about 32% higher than last year at this time. So with this ever-growing amount of issuance and higher volume and more trading activity, it’s creating larger and larger data sets and data points. What we’ve heard from both our buy-side clients and sell-side clients is I need tools to help me manage all of this data. I need tools that are bigger, faster, and stronger, but most importantly, tools that are easy to use and fit in with my workflow.

Jo Gallagher What changes are you seeing in the way traders use data and analytics?

AJ Cass So a year ago when we first launched our Gen AI tool Bond GPT, the concerns most people had were about usage. Are the models taking my data to train? Do I have any control over my data leakage? I think we’re now at a place where we’re seeing much greater level of comfort and acceptance.

We provide a pretty wide mix of data sets, whether that’s pricing, bond characteristics. We have a bunch of models in there. We pull from 10-K and 10-q. So our clients are very happy with the data sets we provide. But just to let you know how far the pendulum has swung, we’ve now come to a place where our clients are asking us if they can contribute their data.

We’ve seen more of that from the buy side, and they’re willing to contribute things like portfolio characteristics, their recommended list, their restricted list overlaid with their compliance. So just think about the change we’ve seen in one year. It’s from a relatively difficult time accepting using these tools inside of their workflow. And now that fear is pretty much shifted towards ‘am I going to be left behind by not using these AI tools?’.

Jo Gallagher Where are trading desks maturing most rapidly in their use of innovative services such as AI?

Mary Beth Sweeney Well, I would say that AI tools have been around for a long time in the marketplace, but what we’re seeing is the evolution of those tools being used in the trading workflow. From the inception of LTX, we have a number of different AI tools such as similar bonds, relative value models, and even liquidity scores. But what we’ve seen with the introduction of our large language model and GPT is that introduction and that weaving of the AI tools into the trader’s workflow.

AJ Cass Let’s say I was sitting in the role of a portfolio manager, and I know I needed to fill a hole in my portfolio. So I would query Bond GPT and say, I’m looking for a single A-rated TMT bond with a spread greater than plus 200 at least ten years to maturity. I want to make sure that bonds index eligible and by the way, from a company that produces more than 5 billion in revenue. Bond GPT will produce a list of Cusips that match that criteria. We will give you a liquidity profile of those bonds with your counterparties, with the dealers you trade with, and depending upon the liquidity of the bond, you could right mouse click and send an order directly to each dealer. Or you can run a simple RFQ off the back of the inquiry.

Jo Gallagher And what’s coming down the line for LTX later this year?

Mary Beth Sweeney Over the last couple of months, we’ve been engaging with our clients about the next protocol that we should be adding to our toolkit and we’re continuing to develop innovative trading protocols in combination with our AI offerings. So I’m most excited to be overlaying our AI offering like bond GPT over our trading protocol such as List. So this summer, we’ll be giving users the tools that they need to generate a list of Cusips by asking Bond GPT a number of queries or a complicated query, and quickly translating that into the construction and execution of a list, understanding that seamless workflow is a key driver for our clients.

Our bond GPT product has grown so quickly through client feedback. Through this feedback, we’ve added 30-plus growth enhancements throughout this year alone. We’re hearing a ton of important input from our live POCs (proof of concepts), and we continue to grow the data set expansions, use cases, and I have no doubt that the list will continue to grow.

Jo Gallagher AJ, Mary Beth, thank you for your insight. This has been Trader TV.

Catch our other shows:

To watch other Trader TV This Week shows go to – Video

Or make sure to follow us on LinkedIn – Trader TV’s LinkedIn

Available as a podcast – here