Matthew Cheung: Okay, this is Matthew Chung, CEO of ipushpull. And in the podcast, the Capital Markets Forum podcast today, we're delighted to be joined by Eren, who's Head of Compliance Technology. Thanks for joining us, Eren.

Eren Erman: Thanks, Matt. Good to be here.

Matthew Cheung: So, can we do a deep dive into, well, a limited deep dive into compliance technology and some of the trends that we're seeing, some of the innovations that you're going to see, and how that's applied, particularly, into a large inter dealer broker like TP ICAP.

Matthew Cheung: Thanks, Matt.

Eren Erman: Sure. Compliance technology, right? Tech, as we all know, is is is growing space. A lot of vendors are entering it to support and provide solutions to govern communication. But that's not the only use case. We could spend an hour on all the use cases, but I'll focus. Probably on record keeping.

Eren Erman: So voice and e coms as a focus for this conversation. That's all right. So no surprise channels are growing. People are communicating using the standard classic voice channels and communication channels written. So things like Bloomberg chat, ice chat or voice. Turrets cloud nine. But there is evolving channels, social media and social applications like what's up again.

Eren Erman: People are communicating. We've seen a lot of fines behind that specifically in the U. S. But that doesn't mean that it should preclude you from being able to use it as long as it's governed. So I'd say in a nutshell, the first trend is how do you ensure that you can evolve your communication channels?

Eren Erman: Govern it and make sure it's recorded, archived correctly. And you train your, your your traders or brokers to be able to use those channels within policy. I'd say another thing, too, is just on the voice. We're seeing transcription as an evolving trend. It's technology is becoming more prevalent and available.

Eren Erman: I'd say transcription is becoming even more commoditized. Vendors like OpenAI offering large language models as open source can allow most vendors to tap into it pre charge or limited charge, but that's The ability to transcribe voices is is now available. I think regulators globally are are looking at that as a use case, and I think the expectation is is is there already definitely in the financial services space.

Eren Erman: I think it's evolving in the in other sectors like energy and commodities, but that that's evolving and how you use transcription to proactively transcribe voice and just set into your your your communication. Monitoring system and then monitor that proactively as well

Matthew Cheung: is that transcription technology particularly accurate because over time, you know, in the last few years has been an area where it's not been fantastic, but presumably large language models are helping the accuracy of that data.

Matthew Cheung: But then how do you find the different syntax and nomenclature that different desks have and the accuracy of those transcriptions? Okay.

Eren Erman: So let's answer your question. The accuracy depends on the language. I think English has a better starting point. As you get another language, there's no surprise.

Eren Erman: It gets more complicated. That being said, though, there's two parts of the transcription journey. There's the first mile transcription, which is what you're talking about. Word error rates, how accurate can it transcribe the actual voice. But, yeah. That's the first mile, and I'd say it's more commoditized, e.

Eren Erman: g. it's more available and more vendors can use it and produce it and make it available to their customers. But the last mile is where the magic happens in terms of the policies within the surveillance platforms to identify patterns. You can use basic lexicons, which are binary words to identify and match to voice transcription, AI and policies that are, they're still scaling across the vendor landscape, aligns to more looking at combinations of words or patterns of words and finding behaviors that might not have been found on just monitoring individual words or combinations of words.

Eren Erman: So it depends on the vendor who's built the policies, how the customer's built their policies on top in terms of what they're monitoring for and the technology they're using, e. g. AI, that can find more patterns that the needle in the haystack classic example.

Matthew Cheung: And is compliance generally t plus one or do you do any real time monitoring as well?

Eren Erman: Or In general? Oh well, in e coms, yeah, we're real time monitoring and it flags alerts as conversations are happening. So, yeah.

Matthew Cheung: And what do you define as e coms? Because that might not be a general term for people outside of the environment. Sure, e communications,

Eren Erman: electronic communications, it's anything written, so email or anything that's written on in written format.

Eren Erman: So, could be an SMS text message, could be a Teams chat message. Anything that's written where you type into a keyboard and produce a A record electronic record and a word, whereas voiceness is when you talk so you can be on whatsapp voice. You can be on cloud 9 BT turret.

Matthew Cheung: So you mentioned about banks finding or banks getting fined rather for poor record keeping or using platforms to kind of shadow it essentially and not being monitored.

Matthew Cheung: In some industries, they may not be as regulated as, say, like the financial industry, like energy or crypto, for example. What approach does someone like TPICAP take when you're having to cover different industries with different levels of regulation? Do you see them all with the highest level of record keeping and information that you're gathering, regardless of what the regulator sees it as?

Matthew Cheung: How do you, how do you kind of approach that?

Eren Erman: Well, I guess the start where we monitor our employees, but we're not monitoring employees of other That's their responsibility from a regulation perspective. So I can't really comment on how how how our customers are monitoring. I would expect that they're doing at least the the minimums that they're required.

Eren Erman: I think if you're regulated like we are, you'd expect that they would have the same same same same. Governance over monitoring communications, but for for those that aren't regulated, and you mentioned energy and commodities, that's that's an example. It's a decision how you how they decide to monitor because they are speaking to regulated brokers like TPI cap and we are recording everything.

Eren Erman: They may not. But. Their policies, it starts their policies, what you can and can't do, and then they decide if they want to be proactive in terms of recording or not. I think there's a trend evolving and the energy and commodity space as an example that they are looking at recording more proactively, for example, WhatsApp or, Twitter.

Eren Erman: Looking at voice transcription as another use case, I'd probably say the same for asset managers, hedge funds you know, with the SEC and CTC finds their, their, their targets on their head. So everyone's looking at this. And when you talk to the vendors, there are definitely those, those verticals all looking at how they can.

Eren Erman: Better govern their communication channels, if they're regulated or not.

Matthew Cheung: So do you, as a compliance person, still have the same level of comfort that you do with, say, WhatsApp compared to, say, a Bloomberg chat?

Eren Erman: So I represent IT but I can, I can speak to that view. Comfort levels on how to record WhatsApp versus Bloomberg chat, I think they're equal.

Eren Erman: There's vendors out there that provide the capability to govern and monitor. Both channels. What's up? Obviously, you have what's up by the business API or what's up by the personal application. Certain vendors go down different routes. There's limitations on both. But ultimately, you're only you're recording everything that's available on the application.

Eren Erman: So in terms of governance and any residual risk of not being able to record something. No, that's that's not an issue.

Matthew Cheung: So one of the things The compliance technology has been doing for a very long time is obviously capturing and recording all of this information that's happening on a broking or trading floor for record keeping.

Matthew Cheung: And 30, 40 years ago when TPI caps started. iCAP, whichever part of it, was originally born as separate companies, when they started to record stuff for regulatory purposes, they probably never had in mind the state of technology and AI where we are today, where you could use some of that data to feed into fine tuned models and, and there's a big trend of people trying to capture as much data as possible.

Matthew Cheung: Is that an area where you're seeing as a kind of an additional benefit of data that you're sitting on from a compliance perspective?

Eren Erman: Yeah, I mean, and I won't necessarily talk about what we're doing here at TPI Cat, but I think if you talk to generally anybody in the space, there's a lot of opportunity to take data into other use cases.

Eren Erman: So it could be a risk use case to better monitor your enterprise risk. It could be front office use case to Identify trends and what's happening. The markets could be for compliance to do, for example, profiling of traders or brokers. More more specifically, you can start to see communication omni channels across individuals inside the organization or externally.

Eren Erman: So, yeah, that that that's That data is there to be used for other use cases and those that have the capacity, the resources and the budget to to start playing the playground and building out other use cases. It's absolutely an opportunity. And, and, you know, if you talk to anyone in the street, there's plenty of use cases being developed right now

Matthew Cheung: and to finish off.

Matthew Cheung: You mentioned there's a lot of, you know, in terms of trends, there's a lot of things happening around transcription at the moment. Where do you see do you see any other trends kind of forming over the next kind of one to two years, which are kind of hot areas that you as a person from compliance technology are excited about?

Eren Erman: Yeah, I think not to be cliche here, but I think the AI use cases are still evolving in terms of how you use AI to support, Any of the current reg tech or reg tech or compliance workflows. I was just on a call with a vendor who was pitching using AI to help with a chat bot for policies, for example.

Eren Erman: So I think that's evolving. Other things. Being more prescript on targeted surveillance as an example. So rather than just having generic trade surveillance or com surveillance platforms that look at generic policies, it's how do you target specific individuals or desks on how they behave? I think that's that's using not just.

Eren Erman: Standard data like communication and trade and order data, but anything from HR data to policies, training expense reports, things like that. So I think that's evolving as well of how you can start to be more targeted with surveillance. AML is hot. Everyone's looking at that, the payment scheme and looking at patterns and how you dip into broader databases to be able to identify the darker gray areas that you probably couldn't get out of the box from some of the existing databases that are available.

Eren Erman: Biometrics is another one. So, being able to associate people's voices to a call, that's not Easily done with just transcription, but speaker diarization is important. That also helps with enhanced monitoring and association of who's talking. I'll probably I'll stop there. There's probably 10 more things I could say, but in the interest of time,

Matthew Cheung: very much for your time today.

Eren Erman: Hey, thanks for having me. Always a pleasure.