Asset management and trading businesses are being transformed by artificial intelligence. That was the message from a series of panels at the Fixed Income Leaders Summit. In a panel on trading, participants discussed how AI was changing the trading environment.
For example, platforms such as MarketAxess are now offering predictive real-time pricing for European corporate bonds. The AI training set includes US corporate bonds for which TRACE data is available, and also MarketAxess’s post-trade service Trax, which compiles proprietary client feeds with vast amounts of European bond data.
When an audience member half-jokingly asked if such tools meant that a European consolidated tape was no longer needed, Gareth Coltman, global head of automation at MarketAxess was quick to respond.
“We strongly support the consolidated tape because it will be good for everybody”, he said.
Buy-side institutions are also feeling the impact of AI. Takaya Sekine, deputy head of quant portfolio strategy at Amundi Asset Management, told the audience how his firm applied natural language processing (NLP) models to fixed income.
“The idea was that for investment grade credit, anything could happen in the world or the environment but it was not relevant for the pricing. We asked ourselves, is this even possible? We took very large database of global events, we synthesised what was really related to nature, etc, and then with the help of sentiment analysis, we grouped and identified a crisis point in different countries, and then we took the bonds and we did a very simple event study, and we saw that crisis could most likely spread to credit.”
But Sekine cautioned against using large language models as predictive tools. “I would say this is too large a question, much too uncontrolled”, he said. “Where we see good results from LLMs, once we have extracted thousands of context sentences, then to summarise or separate the subjects and help the human to understand, we found it does the job really well”.
Timothee Consigny, chief technology officer at French-based fund H2O Asset Management said, “The most interesting use case comes from the investment management team. Someone there starts using the tool to get the full transcript of all the investment meetings the team had on a weekly basis. And after a couple of months, we realised that we’re creating a brand-new data set.”
“We fed it into the AI, and we asked the AI to highlight any investment bias that we as portfolio manager can have. The AI found some times where we changed our mind in the investment meeting, but we didn’t change our position, or the other way around, you change your position but not your mind. It’s a new way of working, sparking a very different conversation among a lot of managers.”
©Markets Media Europe 2024
