The muni market may be several years away from fully integrating artificial intelligence, but firms are taking small steps already.
“AI is a big deal if you want to personalize your approach, differentiate your products,” James Pruskowski, chief investment officer of 16Rock Asset Management, said during MuniTech NYC, a conference hosted by Munichain and Spline Data that brought together a gathering of technology-focused market participants, on Sept. 28.
Susan Joyce, head of municipals trading and fixed income market structure at AllianceBernstein, said the market is still in the early stages of AI, as it is still evolving.
Despite a slower evolution than other markets and industries, Joyce said various firms are already starting to implement AI.
AllianceBernstein, for instance, is using AI to explore how it employs quantitative, writing abilities, but the firm is not “completely AI-focused” at the moment, Joyce noted during a panel entitled Data and Automation of the Buy Side.
AG Analytics has been focusing on using AI in credit research, with the firm using it for content generation by combing data, said Abhishek Lodha, director of strategy and innovation at AG Analytics.
The firm is using natural language processing and machine learning on data generation, as the muni market has 50,000 issuers and hundreds of thousands of bonds outstanding, all of which have their own documents, audits and continuing disclosures, he noted.
“It’s very hard to form patterns, especially in fragmented markets, unless you have a lot of data to train these models,” Lodha said.
And accessing the data to train these models faces some obstacles.
As it exists, data is hard to get and can be expensive to access, Pruskowski said.
Another hiccup is getting the data to one centralized location.
Information is not always located in the same place for different deals. One issuer may have some information listed in the official statement and closing documents, while another may have that same information on the issuer’s website, noted John Murphy, director of municipal investor relations at PFM, at the conference’s second panel, Primary Market Technology.
Furthermore, there’s also the issue of how data can be brought together without having to switch between platforms.
If some of these issues are ironed out, however, Lodha said, “it makes the process a lot more seamless without having that added effort or workload on the research team.”
However, simply implementing AI is not as easy as it may appear; it’s more complicated, said Steve Winterstein, managing partner at SP Winterstein & Associates, at the second panel.
For natural language processing, he said “you have to train the machine [which] is difficult because you need people who can train the machine.”
“You need someone experienced not just in natural language processing and machine learning, but people who understand what an OS is, and to be able to be able to work through it and identify peculiarities of different kinds of sectors and so on,” Winterstein said.
All of this, he said, can take time, but some firms are willing to put in the effort.
One firm “recently got into the reference data business where the goal was to use natural language processing and machine learning,” Winterstein said. “What it took were a lot of QA people to work through it.”
It’s also a costly endeavor, and not every firm will shell out the money to invest in AI.
The fragmented nature of the muni market and the cost of standardizing something like this is so high, Lodha said. Due to this, he said firms may choose humans over “thinking about the long term and building out a core good infrastructure and technology to scale up” as profitability on technology can be hard to achieve.
But that thinking is short-sighted, market participants say.
“Anything that is developed — new quantum methods, AI or anything else — has to be integrated into your trading strategy,” Joyce said.
And while AI will still take time to become a part of the muni market, it simply cannot be ignored, said Matthew Gerstenfeld, Munichain founder, at the second panel.
“You have to have a strategy around how you integrate technology,” Lodha said. “It’s not a back office function anymore. It’s a core business.”