The Tech Update for Muni Industry

Transcription:

Lynne Funk (00:05):

Good afternoon everyone. Thank you for sticking around for the last panel of the day. It's the best one. So you know you're lucky. I'm excited to welcome you all to our technology panel. We have a great slate of guests here, our panelists who will be giving some great insights into what technological changes, technology, AI, machine learning, all the fun stuff that this market is embracing or is about to embrace. Okay, so I'm going to introduce our panelists. We have Greg Beanstock with Solve, and we have Chris Fenske. He's head of Capital Markets at S and P Global Markets Intelligence. We have William Kim, who's founder and CEO of Muni Pro. Abhishek Lodha, who's director of strategy at Ag Analytics. Gil Schulman who's CEO of Ficc.ai, and Dan Silva of Adaje, Inc. CEO. Welcome everybody. Thank you. So let's get into this. There've been some pretty large challenges in the market over the past two years, to say the least.

(01:22)

From a dramatic rise in interest rates, market volatility to plummeting bond volume, massive municipal bond outflows, which among other factors have put a lot of stress on this industry. So today our panel is going to discuss the various ways in which technology can help the industry navigate these challenges from simplifying workflows to freeing up employees, to focus on clients more directly to improving internal and external communications, to defining what exactly is technology heard earlier today that we need this panel because we've got to define it. So I'd just like to start with each of you to tell us kind of how your firm has helped the industry through these stresses. Maybe give an example of something specific solution that your firm is working on and Chris, you get to take it away.

Chris Fenske (02:10):

Okay. For starters, I mean, let me simplify it. Our products only do two things for our clients. They either help them make money or help them save money. So it's pretty simple starting with our book buying application. So 95 plus percent of all new issues go through our platforms. So basically instead of spreadsheets and emails going back between the syndicates, they go through our platforms. It's efficient real time. That actually produces the golden copy of all the data you see in the new issue market, historical issuance data, things like that. So that's for starters. In addition to that, we have the buy-side platform, which is real-time, new issue information, going to buy-side clients. Most of the largest asset managers use the platform right now. So instead of Bloomberg messages with offering memorandums, things like that, they're seeing real-time pricing of deals. They're seeing what phase of the deal it's in, deal documents, they're seeing all the different information they need to actually decide whether or not they want to execute on the deal.

(03:03)

Just this week we launched our Muni deal query application. So this is 20 plus years of municipal bond issuance data, terms, conditions, data. We have pricing data underwriters, co-managers, et cetera. So it's easy way to access a lot of information connected with some MSRB trade data, other type of data sources, credit and risk solutions. It's actually one of our partner groups within market intelligence that actually has an application that credit scores, municipal bond issues in addition to providing financial data on a large portion of the municipal pond, universe of issuers. And then lastly, we have a pricing division that price about 1.1 million municipal bonds two times a day across all the different credit spectrum team of credit analysts that do the credit bonds. We have evaluators focusing on the other 950,000 bonds, so full ecosystem but don't do everything and this, we're always looking to partner with others that fill in some of those gaps.

Dan Silva (04:00):

Thanks Lynne and good evening everyone. I'm with Adaje we are a municipal bond platform that streamlines debt modeling, marketing and management in the face of rising interest rates and volatility. Our primary focus at Adaje has been to really enable our clients to be more proactive and adaptive with respect to modeling and taking advantage of understanding what scenarios are out there for opportunities. This not only means modeling potential future scenarios effectively but also efficiently reacting to real-time changes. And while these market challenges are undeniable, there's some opportunities here at this conference. We've heard talk about strategies like positive arbitrage and escrows and tender offer strategies and at Adaje we are seeing our clients relying on our software to effectively evaluate these strategies. Whether that means understanding the economics for an escrow that is either funded with slugs or open market securities under various rate assumptions or evaluating dozens of tender offer structures and potential outcomes on the tender offer. What we're really enabling our clients to do is understand those economic impacts of those different scenarios under varying rate environments, under varying assumptions around what those scenarios could end up being. And so we're helping folks be a little bit more nimble in what is fairly market volatile market. We're helping folks identify and capitalize on those opportunities as they emerge from a structuring and economic financial modeling perspective.

Lynne Funk (05:52):

Great, thanks.

Abhishek Lodha (05:55):

So I work at Ag Analytics. It's a fully owned subsidiary of a Assured guarantee, but we're an independent arm, but we still have credit in our DNA, just like our parent company does. And a lot of our focus is around credit analytics. The pain point we try to really focus on, and it's been a consistent theme today and we spoke about it at the breakfast AI meeting. The investor panel on ESG is data's there, data's available, it's just not accessible and that's a problem we try to address from a credit analyst standpoint, how do you collect systemize and contextualize this data? How do you analyze it and then build your workflows on top of that so that at scale you can go through hundreds of thousands of documents and information that you need to deal with. And ours is an interesting market from an analytical standpoint now, right? An analyst is supposed to understand credit, but they're supposed to understand cybersecurity, ESG, climate risk, right? They're now experts of a lot of different things beyond what they were fundamentally trained and used to. So how do you bring all of this information together so that they spend more time analyzing and less time collecting? So that's really where our focus is.

Greg Beanstock (07:04):

Great. Thanks Lynne. Thanks for the Bond Buyer for having me. I'm Greg Beanstock. So I'm going to kind of segregate a little bit of what we do. So some of you may know our platform, the diver platform. So what we've done over the years is really two primary things, business solutions and regulatory solutions. So one of the things that we've done with technology is to help the market as new regulation came into our space, how do we help folks efficiently address those regulations for purposes of meeting what they're being asked to do, but then being able to efficiently run their business. So for example, in 15 C two 12, that's one of the things that we focus in on a few other areas that we've spent time there as well. We've also built what I'll call business solutions a little bit to Abhishek's point, which is the idea of there's a massive amount of data in the marketplace.

(07:51)

How do you aggregate all of that data and make that data available in a usable format? So we built a pricing platform for new issues, which is broken into really two component parts. There's a scale writer and then there's the idea of a scale viewer being able to access any historical deal, slice and dice the data. However you want to very quickly bring that data into your analytical framework and work. We're now part of an organization called Solve Solves flagship product is the idea of being able to take unstructured data in the marketplace, unstructured messages that are being communicated across different means and being able to take that parset and turn that immediately background into the clients who are traders and middle office folks with the idea of being able to give more transparency to the market and more information than otherwise is available. So that's kind of a 30 second or maybe two minute overview.

Gil Schulman (08:48):

Thank you Lynne and thanks everyone for coming. I'm very honored to be here on this panel. So yeah, we've been thinking a lot at ficc.ai about these problems. We think that price discovery is a big problem in monies. The way that we solve it is that we train AI models specifically newer networks to accurately price monies in real time. That means that you can put in the specific size of the trade, 25 bonds or 1 million bonds and whether it's customer buyer or customer sell and in real time we can give you the yield and the price. So that helps customers. Some use cases, traders can put their entire bid wanted list. It basically creates a rich cheap analysis in seconds and you can do it every minute on the minute. So you can also see how the market changes throughout the day. We don't just aggregate data, we actually analyzing it in real time for portfolio managers, they can see the actual value of their holdings, so that's good for risk managers as well. And SMAs customers can use it to price specific autos and to decide what they want to move on. They can take intelligent actions in real time. So basically we help the customers understand how the market changes throughout the day for thousands of CUSIPs that are in their portfolio.

Lynne Funk (10:32):

Great, Thanks. Will.

William Kim (10:34):

Thanks Lynne. Thanks to the mom buyer. So with the recent market volatility, we've supported our issuer clients with really looking at their secondary trading and so we'll look at their credit spreads over the year as well as across the curve and see how everyone thinks credit spreads are going to stay the same and stay solid. But as market volatility increases, you see those credit spreads also gap out as well. And so we help issuers with pre and post pricing books to see how are their comparable issuers in the market trading or have priced as well as if you have priced a deal, how has your break pricing occurred? Was it a good price? Did you get a good deal from your underwriter and whatnot? So we do a lot on the issuer side with dealing with the massive rate changes over the past few years, but we also work with our banking and municipal advisor clients to really give them that scale, right with the reduced volume, with the increased rates, there's significant needs for efficiencies and productivity and so we give them that ability to reach thousands of issuers instead of a couple hundred.

Lynne Funk (11:42):

Great. Thank you all for the overview. So a few of you mentioned data and let's talk about data. So data, there is actually a lot of it right in the Muni market, but the accessibility of it, the usefulness of it, the strength of it, how do we integrate all of that data and technology tools that various participants are working on in public finance space from pricing to trading to digesting credit. So anybody want to take that first?

Abhishek Lodha (12:20):

It's my accent. So I think there's always this question of how do you, to your point onboard technology, technology is always cumbersome to a lot of people. I think the first thing there has to be a cultural shift in how you think about it. Technology should be thought of as a value creator, not a cost cutting measure alone. It's definitely helping in cost cutting, but it's something that's going to help you generate more revenue as you were saying, Chris. The second is it's not an all or none solution. When you bring in technology, when you onboard any sort of new products or technology to make yourself efficient, it's not going to solve all your problems. Focus on those efforts where you can quickly repetitively at a low cost, bring in technology automated which essentially redeploys your workforce, your human resource time. And I think that's really a big part of culture shift that we need to have.

(13:18)

There's a beautiful book I read called Empowered on how tech enabled firms work today and the first thing that the book says is think of technology as a core business enabler, not a back office function. And I think that's really important. The second is as you're onboarding technology, think of it in terms of pilots and phases, right? You're going to do a POC, it's never going to be fully ready in the first iteration itself, so start small and then grow and build that momentum. We call it agile in software world. I'm sure everyone on this panel is more aware than I am about those technologies, but I think you want to think in those terms iterative and pilot. And lastly is build a good feedback mechanism. I feel what happens, and we were talking about it earlier today, I think Miles and I were talking from F people onboard a technology and then they don't use it for six months and then suddenly you suddenly have to ask the question, Hey, why aren't you using it? And they're like, oh, I got busy or never got to test it. So spending the time and building a good feedback mechanism to learn and improve your systems in and around the technology or the solution is absolutely very important.

Greg Beanstock (14:32):

I'll just add to that, I think a good segue, I think the point you just made, which is the idea of deploying technology and then it's sitting there and not being used. One of the most important things is making sure that when you think about integration, onboarding is integrating into the workflow, into the client's workflow. What is the problem? Not that you're a technology solve, but what is the problem that you're solving for that specific client and then making sure that you're able to deliver that solution to them. So I think that's one of the highly critical points. Lynne, going back to I guess the first part of the question, just around data in our marketplace, there's the demand from all of you and lots of others is more and more and more data. And so there is now more and more data in all different types of forms.

(15:18)

There's what Gil is doing, there's what each of us are doing in our own and what it comes down to for each of you is how each of us can either integrate that data into a single platform or make our data accessible in multiple formats for your specific needs. So whether that's going through our UI and us being able to integrate data from 27 different sources so that you only have to go to one place or alternatively when you look at our ui, you say, I like this specific component aspect, can I integrate that somehow into your platform or can I simply take a data feed of solve quotes, one of the things that we do right now and just plug that into your internal system. So as you start to think about your needs and it's how you communicate them back to us, it's really critical that we understand what the problem is that we're solving for you all. And part of that is being able to bring in the relevant data the right way so that you have the ease of use that you need.

Chris Fenske (16:19):

So our platforms connect to pretty much the entire sell side, most of the buy side, constantly getting approached by clients, by vendors and other technology platforms. Can we connect to you? So it's not as simple as just turning a switch and connecting. We might call coupon something different than you do. We're going to connect your platform, you have to have top priority security in your platform because we're going to connect you. We're not create a security risk for our users also. So part of the prioritization of this, you take these 40 or 50 people in our queue to connect to, that's millions and millions of dollars of integration costs because there's no common language, there's no common protocol, there's no Apple store to be a conduit to make sure that everything is coded the same way. There's no continuity. So that goes into our calculus as far as determining who we do connect to and how we prioritize whether it's one client that wants this type of technology connect to our platform or 20 clients want the same platform. So part of that too, to Greg's point too, we want to improve your workflow, but we got to use our resources wisely to get there.

William Kim (17:23):

I think one of the things we've noticed with our clients is they'll log on multiple times a day, but one client will have this one feature that they want to use, whether it's verified debt profiles by credit, it's like that's the only thing I care about. Another client will only care about secondary trading data and analysis. One of the things that we see is they start off with this core product that they really love and they use every day and we try to introduce them to the broader scope of options or products that we make available. And so that builds a relationship, but it does take time. So I think people sign up, they know what they want and I think the challenge or the opportunity is showing how much more we can do for them.

Dan Silva (18:07):

Mean to piggyback off something Abhishek said, I think the key is really trying to get the client to derive value from the solution as quickly as possible and making that as part of your process, both from a data integration perspective, but also, I mean for us, trainings very important. Making sure while there's a wealth of really great publicly available information sources like Emma, which does great in terms of making certain public information available, there's a lot of untapped value and potential. I think internally at a lot of organizations that have more data internally than I think they recognize and this data is siloed, unstructured, formatted. As part of our onboarding and processes, we go through great lengths to try to identify what's valuable from a data perspective and existing processes ad organizations to try to bring that into our workflow because again, there's a lot of value not just in publicly available sources of information, but the proprietary data that a lot of you are sitting on to integrate that to bring into our workflows around financial modeling. Then provide for better decision-making processes around your plans of finance that are available to you. So that's a key part of our onboarding and our data integration.

Lynne Funk (19:30):

Gil, did you want to add anything?

Gil Schulman (19:33):

I agree with all of you. I would say one thing to pay attention is how technology change. So things that used to be extremely hard like integration are easier today. So everything can be serverless and you can put everything on the cloud and at least at Ficc.ai, we have brought together a quick creation of comps, any way that you want to slice and dice them or looking at recent prints or recent prints for comps. So you can do a lot with public data today That was, I would say more costly to do in the past and now can more easily be put together, if that makes sense.

Abhishek Lodha (20:24):

Okay. I was going to, sorry, I wanted to say one more thing. One of the things you mentioned is very important. Integrations across the board. Every time I've worked with a customer, the problem is we come with a solution or a platform, how does it work with my existing systems and how do they all connect? Right? And I think there's again a common theme here that technology and API integrations or data feeds have come a long way where data and information can be delivered where you want. So technologies are there now to be incorporated within your existing workflow instead of thinking of them as overhauling or completely changing your workflow, and that's the enablement part of things out here.

Lynne Funk (21:08):

Excellent. So transparency in the Muni market, I think we talk about that's always a complaint. How do you all see technology aiding in making the Muni market more transparent?

Gil Schulman (21:26):

I can take this one. I think there are two ways in which technology can help with transparency. Showing what is out there is one component and then the other component is the inference, the ai. So for showing what is out there, again, technology can show you for any specific CUSIP for any specific trade, it can quickly generate comparable, it can show you how all similar CUSIPs have been traded. So there's a lot of advancement in similarity engines like the same way that when you're buying something on Amazon or when you're watching a movie on Netflix and they know to say, Hey, if you like this, you always like that. It's a technology that is called embeddings and at ficc.ai we're doing the same thing for bonds so we can show you, hey, for this bond here are similar bonds to it, but I think more important for transparency is the artificial intelligence component because using AI you can truly understand within a few basis points of error, you can really understand what is the price of this CUSIP right now and that's hopefully helped the entire market to become more efficient.

Greg Beanstock (23:02):

I'll jump in here. So I think on the transparency question, I just think intuitively about the organization that now I'm part of, so solve quotes, the idea of being able to take and bring to the market pre-trade price transparency and the idea of a trader who receives thousands of messages a day. She's not looking at all of those messages, but there's valuable information in there and so being able to parse that unstructured message, unstructured messages, the runs that she receives and being able to deliver that back to her instantaneously is extremely valuable and she can sort that data obviously however she wants to Gil's point the idea of identifying comps using structural and credit characteristics to identify comps in the marketplace, that's something that we've permeated through the diver platform as well as the quotes platform really critical to what you're doing because you're looking for a specific type of bond for a specific client that bond isn't trading. We know our market, right? There's bonds don't always trade, but being able to find comps in the marketplace can be really valuable. The one piece where you mentioned AI, and I know you guys are very big in that space, the place where we're looking at AI as a differentiator is the idea of building tools around relative value and being able to, whether it's a bond versus a bond or a sector versus a sector or an issue versus issue.

(24:31)

The idea of giving to you, and this goes back to comments each of you guys made around the idea of being able to use technology to get the answer that you want. So with the tools that we're building around relative value, it will be the idea of being able to you decide what the things are that you want to compare to determine the relative value and you'll be able to do that whether it's on a price basis or a spread basis or whatever basis you want to look at that and the frequency as well. So that's a piece that we're going to be incorporating that's coming in the next few months just from a relative value standpoint, which will be pretty exciting and where we'll be using AI.

Lynne Funk (25:10):

Okay, Chris.

Chris Fenske (25:11):

So I'll say, so our pricing platforms, we ingest about 30 million dealer quotes a day, and that's just not just Munis, it's across the Securitize corporates. One thing technology could do that people can't do is handle vast amounts of high velocity data. So in the case of Munis RFQs traded price, MSRB dealer quotes I just mentioned before, taking all this data points, sifting it out, grouping it together, clustering it using ai, that's something it could do much better than a person could do. If our pricing group was to do that without technology, we'd have an army of pricing people looking through these quotes every day. Secondly, I mean in my belief, in my experience, transparency does drive bid spreads tighter. The more price transparency, the more comfortable someone is to bid or offer bond that helps bring bid as closer, which is a good thing.

(25:56)

But then sometimes it gets to the point where bid ask spreads get too tight, you start knocking out market makers where you're not going to get as much upside in each trade you're doing. So maybe only the most sophisticated shops can actually survive in an atmosphere like that where they have to do volumes at the very tight bid ask. But at the end of the day, I mean for the issuer that transparency does improve your execution at the end of the day, they're going to be lower borrowing costs, more liquidity in their bonds when they want to come to market. So I think it's a win-win all around.

Lynne Funk (26:23):

Anybody else want to touch on transparency? And we're going to get into AI too. So I asked the next question.

Dan Silva (26:29):

I just have a short comment. I mean I think everyone here, technology at its core I think produces transparency by centralizing and organizing data that's all of us here are doing that in one way or another we're taking data and we're making inferences and insights come out of that data hopefully in a way that improves transparency for us, that involves simplifying the continuing disclosure process, reducing the burden that issuers have on an annual basis with respect to their portfolios as they already exist since we are already structuring transactions. Once those transactions are structured and Adaje what we're allowing folks to do is then just move dates forward to the next fiscal year and be able to know exactly what was outstanding, produce outputs around what your remaining maturities are, what your average coupons are, any other statistics that are useful in terms of covenant calculations, et cetera, that really end up being a burden on issuers on a repetitive basis. On a continued disclosure basis, we're simplifying that we're automating all of that again because of the structured model solution that we have. So from a transparency perspective, that's really where added shines is in the continuing disclosure financial statements, all those metrics that you have to report on a regular basis from expected on the market side.

Greg Beanstock (27:54):

I think you just actually hit on something really important just around transparency, talking about we all have different technologies, right? And the idea is to be able to take the massive amount of data and deliver an answer. I think one of the core components, and I know certainly in stuff that you do, but the idea is here's the answer, but also allowing the user to be able to see inside, right? It's not a black box, to see what the underlying data is and how we arrived at that solution. Because what we've learned from you all over time is that you want the answer, but you want to do it your way. And so everyone has their own secret sauce. I know Jay's laughing, he knows this, but everyone has their own secret sauce and so they decide they don't want to see secondary trades in a scale creation or they do want to see it, or they only want to see inter-dealer and greater than a thousand bonds. Everyone's going to pick their own secret sauce to do it, and you have to give them the ability to do that and when they look at that curve or when they look at that answer to the question, they have to be able to see what's under the hood so that they can make that determination.

Lynne Funk (29:03):

Great. Okay, so artificial intelligence. I know Gil, you have a lot to talk about on this area, but how is it being used in Munis and how is the market benefiting from it particularly? You wanted to kick us off?

Gil Schulman (29:24):

Yeah, sure, Lynne, happy to. I think there's a current void in the use of AI and machine learning in Moonies and I think all of us on this panel are trying to address it in different ways. As Dan mentioned, I think the value come from being able to analyze and synthesize huge amounts of data in seconds. In other words, no human can hold 1 million CUSIPs and all of the reference data in their head and the ability to price hundreds of thousands of CUSIPs in real time would not be possible without ai. That this is what allows for customers to capitalize on marketing efficiencies in a systematic way. I would say that's the shortest answer that I can. Others can add here.

William Kim (30:26):

Yeah, we have a different take on AI. So I think there's been a lot of talk about generative AI creating itineraries for you and all that. We look at it from the flip side, which is extractive. So what we do is we look at all these documents and we apply AI to improve our table detection rates, to get the information out of the OSS, out of the financials, out of all the disclosures that the Assure community meticulously puts together, and that's really improved their detection rates to the point where we can put together these verified debt profiles by credit for issuers. We can go tighter than that on a by purpose or by project view. So we really see AI as an enormous enabler in terms of extracting the data that's already out there and digitizing it in a way that has meaningful use for our clients.

Greg Beanstock (31:15):

I'll give you a third take on AI. So someone earlier I guess quoted me from a conversation yesterday where I think they said 75, I think I said something like 75% of what's called AI is really marketing. I think it's kind of the same thing. Good point. So when I think about AI, I think about fundamentally what we're talking about is the ability to create models, algorithms, and then to use machine learning and the machine keeps learning with both historical as well as current and market conditions. So it's bringing all that data and it's constantly learning and it's constantly regenerating and refining itself to become a better solution. That's what I think about and that's what we think about at solve when we think about how we're using AI in our platform.

(32:10)

We built a parser that's able to extract unstructured data from messages, any kind of messages, but that's a parser and we're able to turn that back to our clients. The modeling that we are doing around AI is around complex models that are then kind of keep regenerating over time by what the machines learn and how they learn and it keeps getting fed with that market data, market information, the unique aspect of what happened today in the market. What's happening this moment in the market is going to feed that model.

Lynne Funk (32:44):

Great. Go ahead, hop in Dan.

Dan Silva (32:46):

We like to say we've been doing AI before it got cool. Our machine learning and natural language processing predates chat GPT and there's a lot of machine learning out there that's not just generative AI as the folks here have alluded to, and it allows us to convert complex deal documents into structured data models from official statements to swap confirms. So we're not just extracting basic information either. We're extracting intricate detailed information around ratings, covenants, deal details that are pretty useful to know from keeping track of what's going on with your particular debt portfolio, your client's debt portfolio and the future is really interesting. We're starting now to look into going in the other direction, taking structured data models and create documents. So I think there's a lot of opportunity in terms of AI application from price discovery to data extraction to now creation. We're getting into detection of errors, detection of risks, detection of opportunities, all enabled by AI. So it's pretty exciting place right now and in terms of feature development and new things coming out, it's pretty rapid fire at this point.

Gil Schulman (34:13):

Yeah, I completely agree. I think now is the time now it's cheap, it's available, it's affordable, it's no longer takes a lot of time and a lot of resources to train a newer network to learn the money market.

Greg Beanstock (34:31):

You are not supposed to tell them it's cheap.

Abhishek Lodha (34:38):

No, I think all the points about how AI can be used are covered. I think the thing I'd say and something that I've repeated many, many times is any sort of technological solution including ai, just don't throw it at the wall and expect it to yield results and when it does not, then get disappointed that AI doesn't work. It has to be invested in. If you think about AI in general as well, general is the right word around it. If you have to narrow it and train it very well, it needs a lot of data for that. So you have to invest a lot of time and I think that's really important. AI or any other technology and the ROI may not always work for you, right? AI is not a solution for everything. Again, there are times where there are simple models and logics and on the data extraction side we call them RPA, robotic process automation.

(35:29)

There are ways you can do that and it still works. So as you're thinking of it, don't think of it as very, very novel but at the same time don't think it's as easy as just slapping it and working it. I think all of everyone here has spent a lot of time with the data they've collected to be able to invest and build the right solution for the market. Where I'm personally interested in something that Will said, and you as well, Dan Extraction10,000 issuances, 10,000 OSS every year, 160, 170,000 continuing disclosures around them. That's a lot of information stuck in PDFs, right? We keep saying data's there, data's there, but it's not structured and I think there's a huge amount of information that you can extract out of it. There was a research assistant comment today from someone like some companies building that out. I think there's a lot of benefits of this information being unpacked even before synthesizing what it means. So I'm very excited for that.

Dan Silva (36:29):

And I think Abhishek, one thing that you hit on is in terms of the training aspect, all ais are not created equally. I think we saw that the pop culture reference for that now is chat GPT, right? There were other LLMs before chat GPT, but they just were not as quite impressive. So what I'll encourage folks is to try out the different AIs out there. Just because one did not achieve the intended goal does not mean another one won't because they are trained very differently. There's a lot of domain knowledge that goes into that training, a lot of nuance in terms of how these ais are built.

Chris Fenske (37:05):

I'm actually very fortunate to work very closely with our data scientists and one of the challenges I'm finding is trying to separate the cool and interesting versus what we could actually sell to people. Like I said, what's going to save money? What's going to help our clients make money? So some of the things we're working on around capital markets hasn't actually been, in all honesty, a lot of muni centric type projects. We actually have a history of all muni corporate equity holdings for all the publicly filing mutual funds in the US goes back about 15 years. So one of the first things we're working on is a targeting application, basically taking in past trade behavior at the portfolio level, looking when a portfolio member buys and sells, what triggers them, is it rates, is it credit, is equity movements, things like that. And using that to help our clients target potential equity.

(37:49)

IPO buyers, we have a group in issuer solutions that actually is an IR focus group that targets investors. So hey, I want to list a list of all the companies that have equities similar to my companies. So things like that. Capital markets and allocation optimizer. So basically taking into all your past syndicate activity, optimizing it for the type of clients you want to hold your bonds, someone that's not going to flip your equities or flip your bonds. So doing things like that. We have another project in the way in the corporates where trying to predict issuance, so who's going to issue, when are they going to issue and how much are they going to issue? Obviously it's not going to be perfect or flawless, but if it gets accurate enough you could start figuring out technicals three or four months in advance. So that's some of our hard priorities, any of which can, I don't know if anybody will pay for them or anybody buy them, but I mean you have to learn somehow, but we're always looking for client feedback on what people want and how we can improve their workflows at the end of the day.

Lynne Funk (38:43):

Great. So this is a question I didn't necessarily prep you all on, sorry, but I just came to my head. One of my responses were from Chat GPT,. This is not a headcount question which is going to come eventually, but this is a question of how much with AI, the human element, particularly in the Muni market, how can it replace institutional knowledge? Can it go back and really look at what happened in say 1986 during the tax reform act? How much of the human element is still needed in this? Will you want to go?

William Kim (39:26):

Yeah, I think it's an enormous enabler for the human element. The best credit analyst who's been there through the eighties has seen it all might have been just stuck covering one or two credits just because the work is so detailed in terms of getting all that information. You apply AI at scale, that credit analyst who's at the top of their field or she can just see everything all at once and pump out way more relevant insights to their PM, to their banker, to whoever it may be. And so you're really giving the experts, the human experts more ability to reach issuers at scale, reach investors at scale. So I think it's a huge enabler.

Gil Schulman (40:09):

I could not agree more with what William said. It's spot on. Basically you take an analyst and they turn into a superhuman because they have all of their capabilities, but they're just sitting at a cockpit. They're like, Hey, I want to look at 100 comps and I want them sliced and diced like that, and then take these 1000 CUSIPs that are similar and price all of them for me right now and give me the average, the median, so much more synthesized data, exactly the data that you need without you doing the actual work. So you're so much more powerful in a way.

Greg Beanstock (40:50):

Yeah, I'm going to also agree with will, which is good, but I think of it a little bit differently. It's the genesis of AI that is the learning. So what happened in 1980 or 1981 or 82 and today there's a similar environment. The ability to take the lessons of what that analyst learned at that point in time based on the hard work that he or she did then and be able to have that information at your fingertips instantly and then be able to do additional analytical work based on the plethora of data that we have today is really, that's where you start to just really see the power of AI and machine learning and the bringing together of all of that, that really great stuff and how it'll make us as an industry that much better.

Chris Fenske (41:37):

I also think as a power to take some of the bias out of your decisions too. So you take all your actions, investments or whatnot, decisions you made put into the model and you could see, okay, maybe I tend to do this when this happens or when rates rise a little bit, I act too quickly and you could look backwards and try to figure it out there. The other part too, I think the market in general is going into this multi-sector type of way. Sectors emerging, but maybe somebody will cover munis and corporates at some point very effectively, very efficiently, maybe securitized. You need less people to cover a wider spectrum of diverse securities and AI could probably get you there one day.

Lynne Funk (42:10):

Anybody else? I was talking to somebody recently I was thinking about like, oh, rates are so high right now and they are, but it said something along the lines of in the eighties there was a 10% Muni, so we forget that. Can we talk a little bit about a few of you on pricing? Can you talk a little bit more how these technologies that you have are helping with pricing generally? I don't know Greg, you want to start or?

Greg Beanstock (42:42):

Sure. So I guess a couple of ways that we've kind of sliced this over time. So first is through the diver platform and we'll be actually integrating some of the solve data in over the next month. But we built, for example, a secondary pricing tool, which is the idea is that you drop a CUSIP in and be able to take the structural and credit characteristics and not only identify activity associated with that specific bond, but for comps as well. So it's a way to make a market participant smarter, the ability to deliver to users trade tickers that are completely configurable for any market segment, right down to a specific CUSIP all the way up to say, I care about California healthcare and I just want to see every secondary trade that's coming off and literally being able to deliver that to a user within seconds.

(43:28)

That's an example I talked earlier about solve quotes, being able to take pre-trade chatter that goes on in the marketplace and create information for users so that they have up to the minute information as to what's going on, not just for the bonds that they care about for comps as well. So that's all on the secondary side. And then the new issue side, again, it's what we've been talking about here, it's about aggregating data, it's about finding transactions like trades to be able to bring information to a user to say, Hey, instead of having to go to the desk every time, you can get an indicative scale and you can run that very quickly based on the market parameters that are of interest to you. So I think it's harnessing the data, it's using technology to harness the data.

Lynne Funk (44:18):

Anybody else?

Gil Schulman (44:22):

I agree with you. I think with a quick ingestion of all of the market data, you basically, you can just price everything. And again, you can price to the odd lot and you can price 1000 CUSIPs, you can price 10,000 CUSIPs. We have a quant user that price, I think it's 8,000 CUSIP every minute on the minute. They just want to see holistically how the market changed throughout the day. So again, nobody can replace the human in the loop and I don't think anybody's trying, but the traders have so much more power with pricing these days because of AI.

Chris Fenske (45:17):

I want to mention too you, I think one thing that's very overlooked too is that people do focus on MSRB trades a lot. It's relatively real time high velocity. Not a lot of people effectively put in the new issue market into that. There's no CUSIP to an issue bond. You have an entity. So trying to connect the primary and secondary markets in a real-time basis is really important because each one feeds on the other. You basically use trades to determine whether or not you're going to bid higher or lower on a primary, use the primary color to figure out how you're going to make your next trade in the secondary. So AI could definitely close the loop between the two because not a lot of people have actually been able to do that very effectively, at least with technology wise.

Lynne Funk (45:54):

Okay, so headcount, we're going to talk about this. So knowing that there has been a large drop in issuance, there has been a decrease in headcount at various firms outside of technology, some meaningful loss of business. Can you discuss the reasons why investing in technology now during these challenging times is so important in your view?

Abhishek Lodha (46:18):

I'll kick it off. I don't think of technology as a one-time solution as well. I think it's always evolving and it's something that you need to constantly invest in. I think that's important for us to all agree upon. You use outside vendors, you build technology internally. I think that's very important. And also think of technology investment as a long-term capital asset, not as an annual expense or a back office operation. I think that's very important. So the reason you do it in a market like this where fees are squeezing is really scale and competitive. Being competitive in the market, you have the largest asset managers now charging single digit basis points for managing money. There's a whole idea of SMA that's picked up. We all know algo trading and electronic execution making a huge splash. Now you've got external players who are never part of Muni market now trading such huge numbers of trades. I think for you to be competitive in this market and be future ready, that's why you invest in technology. I think that's as simple as it for me.

Dan Silva (47:27):

I also think a low activity environment presents a unique opportunity to implement new technology when operations are not busy. You want to do that when you're at a walking pace, not a full sprint. So I almost view it as an opportunity when things are a little bit slower to go and evaluate new technologies, try to implement them, try to get into the workflow when you're not just trying to push out work product all the time and trying to get things done. So I think where we are in terms of activity that could be helpful and then look, any effective technological solution at the end of the day should enhance the bottom line either through revenue growth or cost reduction period. And to EK's point, if you're not implementing new technology in a competitive market, you're just going to fall behind competitors that are,

Greg Beanstock (48:22):

Yeah, I think Lynne, you've asked that question in good markets too. So I think the reality is you start with the premise that we all talked about, right? There's more data in our marketplace than ever before. And so to process the data without technology, you need people, you need more people in an environment like we're in right now where there's the unfortunate reality in some organizations where headcount is shrinking or headcount is stagnant, right? There continues to be more data, there continues to be more demand from our respective clients. At the end of the day, there's more data and there's more demand, yet you have the same or limited resources. So I go back to the most fundamental part of what we do, which is we take all of that data, we create information for you so that your human resources, whatever those human resources are, whether they're the same level or a little bit less than last year, that you can get the most out of them for the most productive items.

(49:16)

One of you, I can't remember who it was, but made the point, right? It's technology is about taking in essence what is arguably a menial task and automating it so that the humans, all of us can do the more important jobs. We've always asked, is technology going to replace a person? No, look around employment is in this country. Technology keeps evolving and unemployment's really low. There's a reason for that because there's insatiable demand and there's more and more information out in the marketplace. So I look at technology as an enabler. As organizations look at where they're going to spend their dollars, if they're going to be cutting, especially on people, they still need to get the work done and technology is part of how they're going to get that work done going forward.

Chris Fenske (50:00):

I also think too, now you have less people, more work. Quality of life for the existing staff is horrible at this point. So give them some tools, invest a bit, make their life more easy. So you keep the good people on staff. So that in and of itself, keeping attrition low with what's left to your workforce I think is extremely important. Why you invest. Now in addition, we're about to hit the golden age of bonds, 6%, 5% yields people are going to start investing in droves at some point once these rates stabilize, and then it's going to be like this for quite a while in most people's opinions. So it's a good time to invest now, get your systems in place, get the data connected. You need connected when this starts, it's going to be a tsunami at some point of issuance and things to return to normal again.

William Kim (50:39):

I want to echo what Chris said earlier, which was these technologies save money, save time, and help you make money. So you really can't afford to be left behind and not adopt these technologies because ultimately it's helping your bottom line.

Lynne Funk (50:55):

Anybody else? Alright, I'm going to open up to the floor to some questions if anybody has in the audience. Impressive.

Audience Member 1 (51:09):

The Financial Data Transparency Act is on the horizon, so some days soon there are going to be tens of thousands of state and local government financial statements that are machine readable and then add AI to that mix. When that day comes, what do you see there?

Abhishek Lodha (51:36):

It's a topic near and dear to me, so I'll kick it off again. I think it just lowers cost of data for everyone across the board. I think that's something that, so what we do is we do essentially the same thing, but we're doing it internally is we develop our own credit taxonomies for different sectors, for financial statements. We work with a lot of analysts to do that, and then we essentially scrape the PDF documents that are available today as audits and then normalize that data. So if issuer A calls it cash issuer B calls it investments, issuer C calls it cash and investments, we'll normalize a lot of that and then we create XML files or tables or whatever you want, however you want the data to be presented. It's a huge burden on us to collect that data. So if issue, and I mean I'd love to provide that solution to the issuer as well, right?

(52:28)

Saying if you can systemize that, it helps across the board and the benefit. So outside of reducing the cost, and I think that's why we're all in the business, connecting communities and capital and bringing that closer. And if we can do that at a cheaper way, that's great. What it also does is scale, right? So a good example is Mac, Texas, they've been providing structured data for years now at this point, and that has helped that transparency, which we were talking about earlier, has helped a lot of small utility districts tap into the muni market more efficiently. So I think just that idea of more structured data available would be easier. Where AI I think would come in, and this is I the, I call it the building Ironman suits for the analysts. You call it something else, but you'd be able to do more.

(53:20)

On average, the largest asset managers in the country, they probably follow two to 3000 credits approved just because they can't do more. The small issuers are not even looked at if they don't have a hundred million or so on debt. So there's the whole scale where an analyst can now start looking at a thousand, 2000 credits because the data is available now. It's about contextualizing and not collecting, and there's huge benefits out there for FDTA. Now, obviously there's a question on how much does it cost to the issuers and that's It's free. It's free, yeah. The investors pay for it, right? But keeping that aside, I think there's huge benefits for structured data in general.

Greg Beanstock (54:02):

I'll just add to that. I think you're spot on. I mean, there's two things, right? You look at Texas Mac, right? They have a lot of people who are helping extract all that data, so being able to do that electronically is going to be a massive change for our entire market as opposed to state Texas. I think the other piece is once you have a proper taxonomy and once you have all of that data ingested, I just think of, and you just made the point, I'll say it slightly different ways, being able to, for an analyst, being able to assess relative value in an instant based on data and information is I think going to be a game changer. It's going to really make those in our market that much smarter, that much more enabled.

Lynne Funk (54:47):

Anybody else? Alright, well I'll ask all of you to give us some final thoughts here on what you see in the next, let's say five years. How fast are we going to move with technology?

Abhishek Lodha (55:05):

I think that's how long a sales cycle takes in Muni was.

Lynne Funk (55:09):

Well, it's true. We do the Muni market lags. We all know it. You don't have to predict anything, don't take out crystal balls, but give some final thoughts please. You've guys been great.

Gil Schulman (55:20):

I think we will see more trade automation, so I think traders will have stronger tools that we see maybe today a little bit more in the equities market and I think we will see similar trends. I think Goldman have automatic trading tools for corporate bonds and I think we will start and see this in Munis as well Yeah, go ahead. Sorry.

Dan Silva (55:51):

I'm excited about the primary issuance process. I mean, having been an MA before this life and seeing how long it takes deals to get to completion and the weekly working group calls and all the hours that go into preparing plans of finance, doing iterations, preparing documents, and just seeing the kind of, well, the antiquated way some of that gets done. I think over the next few years you're going to see opportunities to automate a lot of that process, to streamline it to shorten time from when you're starting the ideation to when you're actually putting a deal out there to offer to investors, and that's just going to take all those working group calls with everyone having an attorney and everyone on the hours that were compounded on those calls and cut that in half. I think that's a realistic outcome within the not too distant future. I'm excited to see that come into play.

Greg Beanstock (56:52):

I would just say that I look forward to the acceleration of the application of technology in sleepy Muni land. I go back 13 and a half years ago when we started Lumis, and I remember going to a conference just like this and calling my partner and saying, you won't believe it, but there's no technology, man, this is crazy. I was part of an organization that worked in the industry, but to really come to a conference and see, and to sit here on this panel and we're fighting for speaking time, there are six of us here for Christ's sake. I think it's incredible to see how far we've come and to hear the ideas of my fellow panel members, what they're working on and others in the room who I know working on other technologies and other things. Matt, over there. I kind of look out there and say, there is so much opportunity in our space and it's really great to see some great minds going at it for the benefit of our marketplace, and I am looking forward to seeing how that plays out.

Lynne Funk (57:51):

Great. Who's next? Chris?

Chris Fenske (57:53):

I have to agree with Dan actually. I think one of the most inefficient processes, a lot of the financial markets is document creation. As far as deal documents, I come from securitized products originally and the drafting of documents, you're taking a document, you're doing fine to replace for the new series, AI could actually normalize documents, normalize the language, maybe standardize the language. You don't have to put as much legal hours into creating these documents. Could be more error-free, customized. So I think there's a big improvement of efficiency there. Going forward, we should get deals closer to the market. When you cut back into legal hours, sorry, if there's any lawyers in the room,

Lynne Funk (58:30):

Will?

William Kim (58:30):

I think ideally everyone adopts the technology. Everyone's on it and that gives, whether it's a banker, whether it's an issuer or a trader, more opportunities to differentiate themselves and show how their platform is different how their ideas are different and bring more to the table as opposed to be stuck in the weeds of doing the bread and butter of the business. So I do think that that'll open up a lot of opportunities. Obviously there's a lot of concern about rates and compression of spreads, compression of firms in the market. Again, that cost pressure should, the silver lining would be maybe there's more adoption of technology and that speeds that along.

Lynne Funk (59:14):

Alright, Thank you to our panelists. They were excellent. Thank you to the audience.