View Full Transcript

Carl James   00:00:00

Hello, and welcome to the Capital Markets Conversation podcast with Chris Fenske and myself, Carl James.

Chris Fenske   00:00:12

Thank you, Carl. As always, looking forward to today's conversation.

Carl James   00:00:15

Today, what we want to talk about is what we see as a structural shift in capital markets. My own view is having been in capital markets for 40 years, we're starting to see the tectonic plates shifting at their most highest speed at their greatest velocity for the past 40 years. Chris, from your perspective, what is it that you're sort of thinking about in capital markets and the acceleration of these changes that are happening?

Chris Fenske   00:00:44

Over the past 5 years, you've seen the pipeline start build. So pipelines be the interconnectivity between markets, electronic trading platform connectivity. This is creating the infrastructure that's setting us for the next stage. So AI, I mean everybody is talking about AI, how it's going to advance the space, improve efficiency, potentially shorten some investment bankers Workday in the best case scenario. There's a lot of moving pieces to get this to work. And it's also not going to completely change the world in a day or a week or a month or a year. It's going to be an evolution, not a revolution despite what is being assumed in the market where this is going to completely change things in the short term, which I don't believe that's the case per se.

Carl James   00:01:28

And what do you think the duration of that change is? I mean just to frame it for myself, I actually think this will -- we will see the greatest change in capital markets that I've ever seen. And I do think it's on the scale of probably maybe 2 to 5 years where capital markets could become unrecognizable to how they have been, say, over the past 5 years.

Chris Fenske   00:01:51

It will be a significant change. Unrecognizable might be difficult. I mean it's such a highly regulated space that regulators can't necessarily move as fast as the technology. And there's a certain way you have to do a deal, you have to issue with an equity or a debt, which you can't necessarily change that overnight. I mean, legislation has to be much more nimble. And this can't just -- it's not just the United States. There's Europe and different parts of the world have to be as nimble to make it as fast.

Carl James   00:02:19

But if we look back, if we think about the emergence of electronic trading and specifically algorithmic trading, MiFID I, which was a European piece of legislation, was really a retrospective piece of legislation because algorithms were already being used by the market and MiFID I was that retrospective, whereas MiFID II was trying to be a little bit more forward-looking post the financial crash. So you had the financial crash in '07, '08, but it was 10 years later that MiFID II came in and was really -- what was it, 1 million paragraphs of legislation that came in. And I just wonder if actually regulators, legislators, despite when they bring that regulation in, I actually think technology will just have its onward march and the adoption of it, if people can see greater alpha or workflow efficiencies, this will be adopted. Do you not see that?

Chris Fenske   00:03:18

I see that, but I think the starting point is the workflow efficiencies. I think basically taking someone's day-to-day job that in some case, investment banker might be a 12- to 15-hour day, maybe a 24-hour day in some cases and compressing that, basically giving them the same ability to do that amount of work in less amount of time and then focus on other things, focusing on strategy, focusing on pitching to more potential issuers. There's a lot of things you could do with the workflow. And then probably most importantly, those junior banker analysts, now you actually freed up a lot of their day to actually join meetings with clients that they wouldn't ordinarily have time to do because they were too busy working on crunching numbers and working on models, updating models. So there's a lot of spaces in the short term where that's going to -- you'll see that right away, the efficiency gains.

Carl James   00:04:06

I really like that because I think sometimes we -- when we think about the take on technology, we completely forget about the day-to-day engagement with people.

Chris Fenske   00:04:17

It's a human -- I mean, this is a human business. The markets will always involve humans. I mean computers are going to accelerate a human's ability to cover more markets to put together a more accurate bid or offer for a bond or an equity, but it's not going to replace humans. Humans are a necessary COG in the whole machinery.

Carl James   00:04:36

I think if we look back, it provides a template to maybe what we'll see as in going forward. As a trader on the floor, I could probably look after anywhere between 6 to 8 trades because I just had a book and a pencil essentially. electronic trading came in and you were able to cover a larger number of trades because you had algorithms playing and you can just look at is that doing what I wanted to do? Yes, move on to a different trade. So then exactly to your point, there was the ability to expand that -- how many trades I could look at. So instead of it being 6 to 8, it probably became 15 to 20. I think what you're saying is that with the adoption of AI, that will happen. And I do like the idea of what you're saying is that you can -- instead of sitting there for 4 hours crunching numbers, you can do that much more quickly, maybe better, and you end up being in front of someone and being able to articulate a particular narrative of why you should issue and how you should issue or whatever it is you want to be doing for that client.

Chris Fenske   00:05:42

Exactly. And all of this collectively improves liquidity. The improved price transparency, the ability for more traders to focus on different markets. It enables funds of debt to be raised, equities to be issued. And at the end of the day, still price transparency. Price transparency, I still believe is one of the strongest drivers of liquidity.

Carl James   00:06:06

I agree, and I think that's a very interesting point. And when I think about the utilization of different trading methodologies, which ultimately is leading to price discovery, whether it's dark pools, whether it's algorithms, whether it's portfolio trades, there's multiple ways of finding that. And technology and the use of AI will make that even more accurate and more focused. And I agree with you, actually, that price discovery will help liquidity. And just as a sidebar, if you think there are a number of exchanges that are starting to talk about 24-hour trading because the computer technology is there and the computing power is there, but also the price discovery can be done anywhere in the world. So you don't have to anchor yourselves to whether it's Singapore or London or New York, it becomes 24 hours.

Chris Fenske   00:07:01

You mentioned also portfolio trading. So in the future, and I'm sure there's some shops that are already doing this. I imagine an agent that focuses on the credit aspects of the portfolio you're trying to trade, another agent that's taking pricing from all the different sources and culminating a price and these all working together to price this portfolio, give some sort of reasonable mark-to-market on what you're bidding collectively at $0.5 billion at a time, maybe potentially $1 billion at a time in an accurate way where you're not going to lose a significant amount of money based on an error or oversight or something like that.

Carl James   00:07:38

So we talked about the workflow efficiency, which if you think electronic trading over the last 5 years has, I think, changed massively. And then we look forward and that workflow will come more and more efficient. I just wondered about how do you think AI will change capital markets software because that's a slightly different dimension.

Chris Fenske   00:07:59

That's a fear that came -- I say almost nowhere late January of this year, where it just started that, okay, so AI could start getting start-up software company to have given a lot more capabilities to solve problems that long established software companies took years to build up enterprise-level type software. And in some cases, there's some truth to that, like creating a proof of concept or working AI model that could do something very complex that exists today. And it's a relatively low cost compared to what it was 3 or 4 years ago. But the truth is a real enterprise platform has a client service group. It has very strong information security. It has a legal team. It has a lot of things, and these are expensive functions that a start-up can't just put into the software. So you're not going to unseat an incumbent large-scale software provider because that's an industrial strength platform. But on the other end of the spectrum, these are creating tools, more of the picks and the shovels for different types of software platforms that could create some disruption in certain spaces.

Carl James   00:09:12

Yes. I do like what you said actually about focusing on the trading element or creating alpha. But you're right, those ancillary service like legal, they will also be utilizing AI as well. And you'll start to see the compression of their work as well or they will start utilizing AI to compress that. So I wonder how that looks across the whole piece on a horizontal level rather than a vertical.

Chris Fenske   00:09:39

And to that point, too, from even the industrial strength software providers, you're actually using AI to drive down some of their costs by improving efficiency to your point. So you're actually making customer service much more efficient, much more effective, legal review of documentation and contracts and commercials and things like that, much more efficient. So it's actually -- it is benefiting even those large-scale software companies as well.

Carl James   00:10:03

Do you see it as the death of UI?

Chris Fenske   00:10:07

No, not in any way, not in any way. But I do like the fact that a large language model is getting to the point where it could almost customize the UI based on the user. So you don't have to create -- do a release to add 20 features of functionality for a given piece of software for the UI. -- the AI could potentially one day do a lot of that functionality without a release and customize it for yourself. I mean a UI is only the opinion of the person that created that software of what they think their users need. A lot of that opinion comes from a consensus view. You met with a lot of people, got a lot of feedback. You cannot make everybody happy. We're all in the software business to some extent, but you can make the majority happy. And then as time goes on, you could add features and functionality that makes more and more users want to use the software, improve their workflow. But AI has got the potential to actually customize that even further.

Carl James   00:11:02

I'd agree with that. I do think what it does, it places the control in a different place actually. So the use of an AI agent, it will -- for me, I will tailor-make it to exactly what I want it to look like. yours will look slightly different or very different, and it doesn't matter the ability for the AI agent to be able to tell me, you normally look here or you normally use this piece of data. And actually, it customizes. The piece I think that could be interesting as well is it may tell me something that I didn't realize I was doing. And you can either accept and go, oh, yes, and lean into it or you might go, I didn't know I was doing that, and I think that's a complete waste of time. So it actually allows you again to assess and analyze your own, I guess, biases and behavior. So I would agree with you, I don't think it is the death of UI. I think it's just a slightly different one. I do want to ask you something else though, about -- and it's a bit more forward-looking. And it's a broad way of saying, AI is sort of being chattered about lots at the moment. And I just wonder, do you have a particular view of how you think AI where it's placed within capital markets over the next 2, 3, 5 years, that sort of duration and whether it's 3 or 5, pick your poison.

Chris Fenske   00:12:24

So I mean, out of the gate, it seems like a lot of the investment from an investment perspective is going actually into investment banking. Obviously, highest revenue-generating portion of the bank, typically, the longest workday of anybody that works in the bank typically. So a lot of the investments going there as far as automating the generation of models, automating the generation of pitch books. Many firms have already figured this out. They've already created -- they're using AI to do a draft pitch book and update models. So that's where a lot of it is going right now. That's where the focus is. on that, but make it more efficient to my point earlier, enabling the junior analysts now free up their time so they could actually join more meetings and be mentored more from the senior investment bankers. It's also, of course, being used across sales and trading and research, other areas as well. But this is probably the most exciting part of the workflow improvements at the bank right now.

Carl James   00:13:18

That's the tip of the spear.

Chris Fenske   00:13:20

That's the tip of the spear. Once that gets improved, the potential revenue acceleration is almost limitless in some ways and the efficiency of creating capital will improve significantly.

Carl James   00:13:32

Now you and I have lived through many evolutionary processes of capital markets. And as I said at the beginning, this for me is the fastest pace of adoption of any particular piece of technology, specifically AI. What do you see is -- are there any current frictional points to adoption right now? Do you see anything particular that's people why they wouldn't take on AI?

Chris Fenske   00:13:58

So data is not data. Not all data is created the same, and LLMs don't handle data extremely well. They need it on almost the silver platter in an LLM-friendly format that it gets delivered to the LM model. It's very easy to read and access through the LLM model. And that's a major friction point right now. Converting legacy data sets that have been around sometimes 10, 20, 30 years into an LLM-friendly format that makes it very easy to feed into an LLM model.

Carl James   00:14:31

Is there an element as well of training that's required. People have been used to doing things in a particular way. Do you think there's something of training required or pushing people to utilize AI? Is that one of the hurdles or the friction points?

Chris Fenske   00:14:49

Training is essential with all this training. And I think that the first wave, it's been more of a carrot. We're basically, hey, we've given you these great tools. We're teaching how to use them. Let's see how you can use them. Let's see how you can improve efficiency. But these are big investments. These are big investments, putting a big AI infrastructure within a firm and providing these tools. Everything is done its utility. So everything is being paid for tokens. Every time you use it, you pay some portion of a token. So at some point, the rubber is going to hit the road where they have to see efficiency gains in some of this technology across all the different industries that are leveraging it.

Carl James   00:15:26

Do you see any parallel with what happened in the TMT space 25 years ago, I think it's just over 25 years ago to today, as in the TMT space was very much lots of use cases for the Internet. And then there was a bit of a crash and burn and then it started again and the -- whoever emerged from the ashes as it were, went on to become hugely successful. Do you think there's any parallel with that?

Chris Fenske   00:15:58

A lot of parallels. First of all, a lot of overhype, a lot of hype, a lot of ultra-high expectations, unrealistic expectations. A lot of -- back then, it was -- if it's tagged to the Internet, if it had a dot-com next to it, it got a super high valuation and a lot of crowding and investment just because it was part of the Internet, not really knowing what the Internet could do. But fast forward to today, I mean, it was such a -- the foundation of the Internet, changed commerce forever, changed everything we're doing today from rideshares to retail, to everything, electronic trading, it was revolutionary. But even back then, nobody really fully knew where it was going to go. They just knew it was going to go somewhere, and it was going to go somewhere in the right direction, but nobody could have ever predicted that it would have the type of impact it is today. And the other thing with AI, Carl, it's an important point. AI is currently improving itself. So the more you use it, the better it gets. And it's almost like a -- it's not linear anymore, the actual efficiency gains. I think the Internet was almost a linear game. This is almost like a hyperbolic game where it's -- the better it gets, the more people use it, the better it gets, the more -- so you're training it and it's improving itself over time, which is improving efficiencies at the fastest pace I've ever seen. And I think that's to your point as well.

Carl James   00:17:16

Yes, I completely agree with that. I think to sum it up, I think what we're saying is AI will change capital markets at an ever-increasing speed that we've seen previously with other technologies. Let's be careful, though, to shift between, I suppose, truth and hype. Let's be careful. Let's be careful as well about what we think it can deliver and what it actually will deliver. I've heard this from a lot of people, and Chris, you being one of them as well, is that lean into it, Train, train, ask, get mentored because AI will definitely help you be more interested in your role and in your job, whatever it is. So I think the final message we want to give to people is, absolutely, you should be utilizing AI and find that way that is most applicable to yourself.

Chris Fenske   00:18:16

I've said this before and many people have said this as well. AI is not going to replace people. people that can leverage AI will replace people that cannot. And then the next level, people that now have that free time and that are actually using that free time for something even more productive are going to replace the other people that were just using AI just to free up time for other things. So it's the use of that free time, which is going to be the key to success for everybody, for society as a whole.

Carl James   00:18:41

I've heard you say that before, and I think it's very well put. Chris, thank you so much. I think that was very interesting in your thoughts on AI and where it's going to take us. Look forward to next time.

Chris Fenske   00:18:50

Thank you as well, Carl. Always good to chat. Looking forward today as well.