Financial Markets with Kevin O'Connor

From same-day settlement (T+0) and 24-hour trading to token management, governance, and the future of regulated and emerging markets, Kevin explains how deep industry expertise is becoming the competitive advantage in an AI-driven world.

Whether you’re in finance, technology, or simply curious about where global markets are headed, this conversation offers practical insights into the next generation of capital markets.

Michael Larsen (INTRO):

Welcome to ThoughtHive. The most valuable insights rarely make headlines. They emerge through experience and the people driving change inside the world’s most influential organizations. In each episode, we sit down with leaders, innovators, and practitioners to explore the ideas, challenges, and lessons shaping the future of business and technology. This is ThoughtHive, where insight meets action.

Matt Heusser (00:38):

Hi, I’m Matt.

Michael Larsen (00:39):

And I am Michael.

Matt Heusser (00:41):

And this week we have Kevin O’Connor, who’s the VP of Capital Markets at NuSummit joining us. Welcome, Kevin.

Kevin O’Connor (00:49):

Hi, thanks for having me.

Matt Heusser (00:50):

It’s our pleasure. So it’s right there in the title, Capital Markets. Let’s start off there. How does NuSummit see financial markets and how does NuSummit serve them?

Kevin O’Connor (01:01):

Yes, NuSummit was born by being an internal IT department for a global exchange. When that exchange went public, NuSummit was spun off as an independent company. Leveraging 25 years of experience serving direct market access participants and all the workflows surrounding that, we feel like we can be impactful in solving problems for anybody touching any market or any exchange marketplace in the world.

Matt Heusser (01:34):

So let me ask you, how does that view of financial markets differ from everybody else’s? How’s that different than the standard industry consensus?

Kevin O’Connor (01:45):

Sure. There are three areas where we can help. We’ve got data and analytics, and digital transformation, which includes cloud services and cybersecurity. But having supported a global exchange, we have domain expertise which we can leverage. So it’s really a consultative approach with technology solutions, including AI, which is interwoven into each of our solutions. While you can go to any AI startup and engage with them, we will really understand the workflows and what your goals are and how to achieve them, taking into account governance alongside the technology challenges. That all becomes relevant in the discussion.

Michael Larsen (02:35):

Very interesting. So I have a question on this from the fact that AI is now the topic du jour, maybe the topic du year. AI is consuming all the oxygen in the room, let’s face it. And I guess the question I would ask here is, “How has AI changed the game for you?” Are people really using AI? Or what are they using it for? Realistically, not just, “Oh, it’s got this promise or this potential pledge.” What are people actually using it for right now?

Kevin O’Connor (03:07):

Yeah, good question. That promise and potential pledge is what’s driving valuation of the individual stocks associated with AI. But machine learning is not new. It’s existed in capital markets since the 1980s. A whole bunch of the quantitative firms began leveraging it. The only difference with AI is that now some of these tools are commercially available, and that’s happened quickly, and the conversation is changing quickly. If you talk to some of the asset managers and exchange marketplaces a year ago, a lot of the institutions’ pitches were, “We’d like to do more with AI.” And then that evolved into, “Well, what would you like to do with AI?” To your point, what we’ve realized, what I think everybody’s realized, is that AI can be very effective in processing. All of the queries that you would do previously can now be done in seconds. Those queries used to take minutes or hours, and queries that used to take days or weeks can now be done within an hour.

(04:17): So there’s a great impact on all of the workflows. We’re seeing a lot of traction on the operations side when we help exchange marketplaces advance from 17-hour trading sessions to 23-hour trading sessions. Some of those exchanges are going to 23/5 from 17/5 with an eye on 24/7, because that’s where it’s ultimately all going to wind up, and they have to get this done pretty quickly. So we’ve been helping people fill gaps in their workflows there. But a lot of that batch processing that used to be an entire overnight process can now be done much quicker, which allows them to have extended trading sessions. We’ve taken some exchanges from what used to be T+3 to now T+0, T+1 being their most recent settlement and clearing. It’s the same type of contribution from AI.

(05:18): It’s not one thing, though. It’s not just AI that can get you from 17-hour sessions to 23 or 24-hour sessions, or from T+1 to T+0 clearing and settlement, because you still need that domain expertise to know what regulatory headwinds are coming at you, not just technology. You really have to understand the workflows and the other elements around the technology in order to pull anything like that off.

Matt Heusser (05:48):

So we’re getting into some of my interests on the retail investor side. I’ve always found it very frustrating. First, I have to send the money to the brokerage and wait two days for fund settlement, then I buy or sell whatever I’m buying or selling, and then I have to wait two days for fund settlement again. And if it’s a weekend, it could look like four, could look like five days, if I buy something on the 3rd of July for a three-day weekend. So when you say T+2 to T+0, you’re saying moving to same-day settlement. The money’s gone, I have the stock, we’re all good. Everything’s good, all in the same day.

Kevin O’Connor (06:31):

Yeah, the retail audience, they love it. Their buying power remains constant, it doesn’t dissolve, and it allows them to day trade effectively. It can be in, out, back in, back out the same day, as opposed to previously, where for margin reasons they had to wait for everything to settle out in order to continue their activity. So yeah, we are seeing certain exchange marketplaces moving toward that, some more aggressively than others.

Matt Heusser (06:59):

Yeah, I haven’t seen that. I mean, I guess since I started trading online 20 years ago, I’ve seen the rise of no-fee trades. It does feel like settlement is faster, but I really haven’t seen anyone advertising or talking about same-day settlement in the markets that I’m in. That seems to me to be kind of a big deal.

Kevin O’Connor (07:23):

Yeah, and the workflows are the same, whether you’re talking about going to extended trading sessions or T+0. They’re very complementary, but the idea is kind of process-as-you-go, as opposed to waiting overnight to do that batch processing.

Matt Heusser (07:41):

I would think that you’d be breaking it up into smaller and smaller batches to eventually get to processing one transaction at a time. Or is that not how it works?

Kevin O’Connor (07:50):

Yeah, so it’s not real-time, but it’s almost real-time.

Matt Heusser (07:54):

Same day. Yeah. The other thing you mentioned briefly, and I’d love to drill down on for a minute, is you said something like queries moving from a day to an hour. What kind of queries are we talking about?

Kevin O’Connor (08:06):

Any quantitative-driven model is dependent upon large data queries. If you’re doing these queries and trying to refine your trading model and your systems, this can all be done quite quickly compared to what used to take quite a long time. The other side of that is, what does it cost to do quickly? Because as we know, compute is expensive and getting more expensive. That’s another way we can help, in the sense that we’ve identified this through discussions with our customers, and we’ve created an entire token management solution. It’s not completely dissimilar to a smart order router, just applied to token usage: we identify by user or by business unit where you’re using more tokens or compute, and then we’re able to make suggestions as to which LLMs you should be pointing at to get those jobs done, depending on how quickly they need to be done and what kind of priority they are.

(09:15): Some of these things are not latency-sensitive, some of them are, and latency means different things to different people. Ultimately, the end product would dynamically take those decisions away from the end user, automate them, and provide an oversight situation for the business owners to be able to limit consumption of those tokens.

Michael Larsen (09:40):

So you just touched on something that I’m very interested in, in the sense that, again, like Matt, most of my exposure to investing and financial markets is on a personal level. I definitely feel the situation of, “Oh my gosh, so many things take so incredibly long to settle, to be able to get in place.” And you made a point there about the ability of people to just get in, get out, get in, get out at a retail level. This opens up my thoughts to international markets, because a lot of these are international markets. I shouldn’t even say “a market,” there are lots of exchanges where people can interact with a whole bunch of different things. What levels of oversight are being done in this regard? What challenges do you run into when it comes to regulating or having proper oversight of these things, and helping give confidence that things are working the way they’re supposed to?

(10:39): And I guess, what’s coming up in regard to that? Boy, that’s a lot of questions. I’m sorry.

Kevin O’Connor (10:45):

Yeah, it’s kind of a fragmented environment right now, in the sense that you have highly regulated exchanges, equity exchanges, futures exchanges, options exchanges, et cetera. Every asset class, which has been around for a hundred years or more, and the SEC, the OCC, FINRA, they’re always monitoring and auditing, and all these exchanges have regulations they have to deal with across all of these elements. And these are all things we’re intimately familiar with. On the flip side, you have prediction markets, which are brand new and largely unregulated right now. So there’s a bit of a race going on between those prediction markets and traditional marketplaces, where, maybe because of some of the pressure on digital assets over the past year or so, the prediction markets have just decided, we’re going to tokenize everything. And now it’s kind of the Wild West over there.

(11:49): The regulators really won’t have time to keep up with them. They just don’t have that kind of bandwidth. And it’s resulted in litigation very recently between a large exchange in the US and the CFTC, the regulator. We all know how it will end, in the sense that those prediction markets will ultimately be regulated, but how long does that take, and what kind of damage is done in the meantime? There are significant risks to the entire exchange, and when I say exchange, I mean the entire ecosystem: all of the market participants, all of the benchmarks, and everything else being introduced by these unregulated entities that have just swept in and treat everything as if it were a digital asset, without a real eye on regulation. Publicly, they say they want to be regulated, at all of their political action committee meetings and things like that, but they know that’ll take forever.

(12:53): So a little bit of a double-edged sword there.

Matt Heusser (12:56):

So the prediction markets, Kalshi, Polymarket, those sorts of things, we talked about this a little bit earlier in the green room. They look like financial markets. They say things like, “I’m going to place a bet that the S&P 500 is going to be above or below this number by this date,” which looks very similar to an options play. And they offer you real-time liquidity, you can cash in at any time. “It’s become more likely that I’m going to win, so I’m going to cash out now.” But it seems to me that they’re trying to escape the kind of rigid oversight that the SEC offers. And with less oversight, it’s not really a fair competition, because they don’t have to play by the same rules.

Kevin O’Connor (13:41):

The regulation is in place for a reason. It’s taken decades to evolve and be refined. How this is going to play out, I’m not a lawyer, so I’m not going to provide legal insight. But like I said, ultimately it will end up with the decision of the judges and the courts, how it should end. But what the exchanges are not going to do is allow prediction markets to come in and begin listing all of the instruments they’ve been listing for years in an unregulated fashion, when the exchanges themselves are, by and large, regulated in everything they do. To your point, for it to be a fair fight, there has to be some kind of balance introduced there.

Matt Heusser (14:25):

I just want to make sure I got you correctly. It used to be that you’d have to hire a quant who would code in Python, who would get access to data sets, who would then run queries in machine language. You’d have to manage your own servers and try to get answers fast enough to keep up with the market. You’re saying now exchanges are using AI: they can get the data set, throw it into an LLM, say, “process this,” ask natural language questions to get their answers, and get them fast. The problem is, token usage is unpredictable, the right LLM is unpredictable, so if you’re not careful, you might be like Uber and spend your whole year’s LLM budget in a couple of months. There needs to be some expertise around how and where to send what queries to get the best results.

(15:22): That’s expertise that can be reused instead of relearned every time.

Kevin O’Connor (15:27):

Well, first, taking a step back, it’s not necessarily exchanges, it’s exchange participants, the people who are sending orders, whether institutional or retail or whatnot. Are your strategies going to work by running them through an LLM? I’m pretty confident that the firms that have been doing this for 35 years are going to probably be a little better at it than whatever one of these LLMs will spit out at you. There’s a little bit more to it, and you still need that quantitative mind and process to be long-term successful. With that said, all of this can happen faster. Traditionally, HFTs, for example, don’t use machine learning for every single process. It’s leveraged in certain low-latency strategies, and it’s leveraged in certain less latency-sensitive strategies. If you want to look back and find moving averages over a number of years on a minute-by-minute basis, that’s a heck of a lot of data.

(16:29): Or if you simply want to get a single instrument, we’ll call it Apple, and all the associated options, or one of the associated options, that’s a heck of a query. You’re asking a database to be parsed for a lot of information to get some amount of input. That’s where AI is really effective; you can get those queries done much quicker now than you could in the past. There are definitely benefits, and token usage is only out of control if it’s unmitigated. If you don’t have a token management service or solution, it can get out of control very quickly, and the end result of that can be very predictable. You’ve seen a lot of the larger hyperscalers, for instance, pulling back on their token usage very recently.

Michael Larsen (17:22):

So if I could follow up here, I tend, very often when we do these conversations, to act as the semi-uninformed everyperson, because a lot of the time I am. And in this case, I’ll confess, outside of individual markets, this is an area that’s like, “Wow, okay, this is all really cool and interesting.” So to just ground this for me, and maybe for some others who might be curious, what’s your ideal client? If somebody’s looking to say, “Hey, we want NuSummit to help us out with something,” that’s what I’m trying to think about. Who uses this? I get that there are markets and exchanges and customers that do this, but what is your ideal client, and what would you specifically be able to do for them?

Kevin O’Connor (18:09):

So yeah, exchange marketplaces are the most obvious. If you look at where we have experience across the entire exchange marketplace value chain, we can help with anything pre-trade, at trade, or post-trade. Asset managers, people who interact with exchanges, we can help with anything from data and analytics. All the exchanges provide that; asset managers consume it and then provide it. Those exchange groups have become data providers, so it’s kind of a bidirectional service: asset managers consuming it, exchanges providing it, and we can help with anything in between. We can build you a data lake, for instance, for your data inventory. We can provide insights as to which parts of your data inventory have value and which parts might be ready to be retired. Digital transformation, cloud services, we took an exchange from a co-located environment to the cloud, and actually back again, once they realized some of the regulatory headwinds of hosting an exchange in the cloud. Kind of an ironic end to the journey for them.

(19:24): But we have expertise in doing that. So yeah, clearing firms, we can help all of them. Again, anybody who sits in that trade lifecycle ecosystem, we can help. And then there’s a layer of cybersecurity that’s relevant to each and every market participant as well, and that’s all AI-driven. You’re getting real-time alerts much faster than you did before we started leveraging agents in that area.

Matt Heusser (19:52):

So when I think cybersecurity, I think of audits, I think of policies, I think of checking the status of the patches for all the operating systems, the status of the configuration, but also penetration testing, where you can now do continuous penetration testing, with software that maps the entire attack surface of the organization, so you can find problems before the bad guys do.

Kevin O’Connor (20:22):

That’s the idea, right? Because those threats are now AI-driven. AI works both ways. The bad guys are leveraging AI to attack, and if you’re not using AI in threat detection, well, then you’re just that much more vulnerable. Those attacks will become more frequent and more targeted, and you need to know where you’re exposed and have that addressed. Those are areas where we can help.

Matt Heusser (20:50):

There’s also scanning the logs to look for login combinations that are different from what people were doing last month. All of a sudden someone is trying to log in the minimum number of times they can attempt without getting locked out, then trying again exactly one hour later when the timer resets, that sort of thing. Are we also checking logs?

Kevin O’Connor (21:19):

Yeah, we’re scanning for all types of threats. We actually have a cybersecurity assessment that our customers fill out, which gives us a good picture of where their exposure lies. Then we work with them to build up defenses in their most exposed areas first, working toward complete insulation from those types of attacks. But yeah, to your point, you’re seeing more attacks, more targeted, more often. You have to keep up and be advancing as quickly as the attackers are.

Matt Heusser (21:57):

So the two things that worry me right now on the AI side, and they’re kind of related: one is the ever-changing price of tokens. Right now I see Anthropic and OpenAI subsidizing token costs. At some point, they’re going to want to be real companies and charge a profit margin on their services, so we’re going to see an escalation in token costs, which can be mitigated by open source models. The other problem, and I’m curious if you see this as a problem too: there’s Mythos, and there’s the other one, the sort of hobbled model that Anthropic put out recently. A federal directive said, no, you can’t let customers use Fable. It took a couple of weeks to sort that out. If I were using that for cybersecurity, or to do my analysis on what to invest in, I’d have a problem if that got pulled.

Kevin O’Connor (22:54):

Yes and no. With any institution in financial services, whether it’s an exchange, an asset manager, or a brokerage house, it doesn’t really matter, they all have primary, backup, and tertiary systems. They should all have the ability to switch from one to the other before anything goes into production. No different than market data: what happens if my market data goes down? I’ve got a problem. Well, yeah, you switch to a backup system, whether that’s something you built or something a vendor provides. I think everybody’s learned the lessons along those lines, to be prepared for these things. The risks are no different than any of your tools or vendors suddenly becoming insolvent and you can’t use them anymore. Nobody is painting themselves into a corner with a single-vendor approach, because they’ve already learned those lessons.

(23:51): It’s a pretty mature audience that you’re dealing with in this sense. There have been lessons, and they have not been free, so people have learned how to mitigate them along the way. And these are areas where we can help institutions that are starting up as a new business, for instance. We’re seeing compute futures being listed on particular platforms, and there are just things you don’t know that you don’t know. We have a lot of experience with the way governance can, and will, eventually turn out, and the way the market will react to these things. These are just areas where we have a level of expertise and can help.

Matt Heusser (24:33):

Yeah, I tend to see a different side of the IT world, and I’m glad to hear that. I suspected it would be kind of like virtualization, where eventually we got Docker and Kubernetes, and the idea that I could have these artifacts and run them in IBM’s cloud or Google’s cloud or Amazon’s cloud, wherever, and if something goes wrong or the price isn’t right, I can interchange them. I don’t see that much maturity yet in the LLM world, especially because the results tend to be less predictable unless you configure extremely well, and they’re much less predictable between models. But it’s encouraging to hear that the people safeguarding our financial transactions have thought about that ahead of time.

Kevin O’Connor (25:19):

Yeah, absolutely.

Michael Larsen (25:20):

This dovetails into something I was just thinking about as Matt was mentioning the whole idea of virtualization. It reminds me, because 26 years ago, that was the world I was involved in. I was part of Connectix, making Virtual PC, and that was like, “Ooh, a big thing, you can run a PC inside of a PC,” and that was wild. And now, of course, we’ve broken it out into things like Docker and Kubernetes, and all that. Virtualization is now the basis of the cloud. So with a lot of what we’re talking about here, what plays do you see coming down the pike? And given the nature of AI and how it’s developed just in the past couple of years, what do you see in the next few years, or maybe even the next few months, becoming mature and more solidified?

(26:12): And what avenues do you think are on the horizon that we’re not paying attention to yet?

Kevin O’Connor (26:19):

Yeah, well, I think you’re already seeing some of it. Firms have become very conscious of things like hardware footprint over the last 10 years or so, and there’s been a cost-reduction effort around that. You’re going to see the same thing around token usage and compute usage. That’s no different than when the cloud was a new idea and all the large asset managers looked you straight in the eye and said, “We will absolutely never put any of our data in the cloud.” At which point, sitting right next to them, they already had an RFP out to every major cloud provider and knew exactly what it would cost and how long it would take to get there. That all happened. So the idea of a new frontier, when you’re talking about AI and all of the pitfalls and mistakes you can make, they’ve got people on staff who hopefully know most of those pitfalls.

(27:17): With that, AI is moving so quickly that they won’t have a whole lot of time to think about it. And what I think you’ll see is some of the more bureaucratic, large organizations will begin to outsource a lot of this process management to external vendors like NuSummit, because they won’t be able to react quickly enough on their own. The environment’s changing so quickly that it would take a pretty sharp tack to keep up with it. We keep up with it pretty well. I think the greatest minds out there in financial services, the largest hedge funds with unlimited funds, keep up with it incredibly well, but it’s not necessarily a level you can expect to compete on if you’re a very large organization, especially if you’re fragmented. So yeah, there’s going to be a lot that changes, but they’re all changes we’ve seen before.

(28:14): It’s just going to happen faster. There’s no reason to think that’s ever going to slow down. In fact, it’ll continue to accelerate.

Michael Larsen (28:21):

Wild times ahead. That’s very exciting. So for anybody out there who wants to, if this conversation has piqued their interest and they want to get to know more, understand a little bit more, or just pick your brain beyond this conversation, what’s the best way for them to do that?

Kevin O’Connor (28:39):

Yeah, well, our website, nusummit.com. We’re also on LinkedIn, you can find out more about us there. I’m kevin.oconnor (no apostrophe) at nusummit.com. I’m happy to engage on any of these subjects at any time with anybody.

Michael Larsen (28:57):

Fantastic.

Matt Heusser (28:58):

This was fascinating for me. It’s interesting, because in IT we tend to think of financial markets as slow to respond and maybe a little fuddy-duddy, but they’ve clearly put real thought into their maturity and their ability to act. Even same-day settlement is particularly impressive. I worked for an insurance company that came up two or three decades after Blue Cross Blue Shield. We just had all of our data on disk. We didn’t have data backed up to storage somewhere that would need to be loaded from tape, and we didn’t run COBOL, because we were newer. That really gave us an ability to move, and I’m hearing similar things in the financial markets, which I didn’t know was the case.

Kevin O’Connor (29:47):

Yeah, well, the level of sophistication in capital markets is really unprecedented, the things people are able to achieve. These institutions are generally well-funded, so they can operate as needed. There’s a lot of thought put into everything they do, which keeps it interesting. There are some really smart minds in the pool of people.

Matt Heusser (30:07):

And in the trader-broker space, is there a little bit more nimble cowboyism?

Kevin O’Connor (30:13):

Oh, no, not at all. They’re all in that same space. The people who’ve been doing this for a long time have seen things they’ve learned from, and all of those lessons just add to your portfolio of experience.

Michael Larsen (30:28):

Fantastic.

Matt Heusser (30:30):

The people that don’t go to jail.

Michael Larsen (30:33):

That’s one way to…

Kevin O’Connor (30:35):

…learn a lesson.

Michael Larsen (30:36):

And with that, I think that’s a good place to put a pin in it. We want to say thank you very much, Kevin, for joining us today for the NuSummit ThoughtHive Podcast. And to those listening, thank you for giving us your time and attention. We’ll be coming back to you very soon with new topics, new explorations, and new discussions with people who are making waves in these spaces. So thank you so much for joining us, and we look forward to talking to you soon. Take care, everybody.

Kevin O’Connor (31:04):

Thanks, guys.

Matt Heusser (31:06):

Thank you. Bye.

Michael Larsen (OUTRO):

You’ve been listening to ThoughtHive, a platform built for those who lead. For more conversations, expert perspectives, and practical insights on enterprise transformation, visit ThoughtHive by NuSummit. Thanks for listening. We’ll see you in the next conversation.