Tyler Pathe: Hello everybody and welcome to FinTech Futures AI Insights event here in the heart of London. My name is Tyler Payne. I'm a reporter here at FinTech Futures and today among our many guests who have joined us here up in level 19 of our office is Matthew Chan. Matthew, how are you? Very well, thank you.

Tyler Pathe: Thank

Matthew Cheung: you. What do you make of our events? That's been fantastic. Everyone's very curious about

Tyler Pathe: AI. That's it. Exactly. We're going to be getting a little bit more into into the technology during our conversation here. But why don't you just give our, give our audience a little bit of snapshot of who you are?

Matthew Cheung: Yep. Sure. Yeah. So I'm, I'm Matthew Chan. I'm CEO, one of the co founders of Lightbushball.

Tyler Pathe: Fantastic. Well, I think the, the, the leading question from, from that, you know, is, is what is iPushCall mainly, mainly what is its core services and who are they for?

Matthew Cheung: So we're a data sharing and workflow automation platform.

Matthew Cheung: We focus very much on capital markets. So our customers are banks, buy sides, brokers, exchanges, trading venues, data companies, anyone who has a problem. In capturing unstructured data and being able to share. So typically customers will work with, they'll use our platform, which is a no code platform to be able to get real time data in from sources such as a spreadsheet or in a chat platform, or maybe sort of third party platform via an API, get that into our service.

Matthew Cheung: You're able to map and enrich it and you can share it in real time to one of your clients, for example. You may want to consume it in a spreadsheet or chat or an API. So we sit in the middle and doing all the heavy lifting in terms of the integrations and then providing what we call this omni channel approach.

Matthew Cheung: So if I'm a customer and I have some data, I can connect it into us. We can deliver it into, you know, the right application at the right time and the right place.

Tyler Pathe: That's really, really important. We'll get along onto the topic of data in just a sec, but in terms of like the, the workflows that are based on Gen AI that you've built for the capital markets industry, I mean, can you give us an example of one of those?

Tyler Pathe: I mean,

Matthew Cheung: we're a technology company, so it keeps out on on AI for a long time, but I think a big, uh, big change in adoption and perspective. Some people wanting to learn more about AI obviously happened this time last year when Chanting PT came out. And that has been fantastic for our company because we've always been building chatbots.

Matthew Cheung: We've been building chatbots for about five years now. And customers are now looking at how can they utilize chatGPT, which they probably played around with in their spare time. How can they utilize it in their business? And if their business is like financial markets and capital markets and sales and trade, it's not always that accessible to go and use chatGPT on a regulated trading desk.

Matthew Cheung: So where we come in is, you know, we're able to then through the previous kind of eight years of our company, we've spent a long time building this data integration platform. We've built 20 odd integrations into, I've already mentioned kind of Excel, chat and API, but also things like a blotter on a trader's desk.

Matthew Cheung: We've got a web apps, mobile apps, desktop containers, like open fin and fin sample, a whole variety of places where people in the front office kind of live and breathe. So the data integration platform. We've got the actual end user integrations into these different applications. We've got our own bot framework.

Matthew Cheung: So in our chatbot framework, you can configure a bot once and it will work across all these different applications. And now you add a large language model or AI onto it, and all of a sudden the entire kind of stack that we've built becomes, you know, kind of 10x more than what it was, which has been fascinating for us.

Matthew Cheung: So customers We've been talking to customers about generative AI probably since about January, February of this year. And obviously ChatGPT came out end of November last year. So we, we very much like I said, kind of just geeked out on, on AI and, and, and already had expertise kind of in the house.

Matthew Cheung: So we then started to bring some of the tools that's available for, for more of the consumer, you know, like ChatGPT. into our platform. Every single conversation we had with customers that kind of in springtime, every single one was touching on Jenna, Jenna, which I, which is no, no doubt. You know why you've had this conference today.

Matthew Cheung: Just so much curiosity around for it. And we, we then started embarking on a few different projects. Yeah, we've, we've done a project with AWS where They have their own machine learning kind of component bolt on called bedrock within bedrock. You can then select different large language models. So we've got a project using anthropic board, which is a large language model, very similar to GPT for, and, you know, we're using that in production.

Matthew Cheung: We've got customer using it where they're using it to capture prices that are seen in chat and receiving from their clients, taking that adding structure onto it. They're using it. Generate more data and insights and automation as well. That's kind of one use case. There's another use case with a major European exchange where that's more based on kind of NLP and natural language where we can connect to large data sets like reference data and index data.

Matthew Cheung: We can make that data available via a chatbot. We've been doing that for many years already, but now you use AI and large language models so you can have a natural language input. Rather than more of a command line kind of input to it. And then we've got some asset managers that are using it who are consuming again, prices and so on, coming from different counterparts.

Matthew Cheung: We get to take that, aggregate it, so it makes their workflow more efficient. And for us, in every application in general in CIV AI that we're seeing here, we we, our customers are using it to make the process more efficient. And that's ultimately, I think, was mentioned in the conference. Where it's taken the drudgery away from like the manual processes that people are doing and that exists universally whichever job you're in Yeah, and then in sales and trading and kind of front office part of the financial markets anything that's traded OTC Not anything, but most of the things that's traded OTC is still done manually copy paste double click double keying things Jumping into different applications So we were able to stick to all of that together.

Matthew Cheung: We've always been able to do that, but then AI has been it has accelerated the technology kind of roadmap that we have, and has accelerated people's perspective on using technology like this, because people are willing to take some risks and look at, okay, how can I use AI in my business? But when I say take the risk, it means that people are quite happy to create the OC or projects.

Matthew Cheung: And if it fails, they're not, they don't really care. They just want to try the technology. Whereas we've been using it now for quite a long time. So we're kind of beyond this experimental phase. And it's already, okay, exactly. If you've got this problem, we can solve it. If you've got this problem, we can solve it.

Matthew Cheung: If you've got this problem, we could help you solve it. And we could work collaboratively, collaboratively together on it.

Tyler Pathe: That's really, really interesting. I mean, you know, we. We one of the major takeaways that I personally had from this event is people have been like, how can we use AI? Where can we use AI and it's companies like yours that are really sort of delivering those solutions and and really thinking Sort of quite outside the box really in terms of what it's what is sort of immediate Uses are you think well actually this can work for something else as well where there are other popular Of course, a major theme today and throughout this financial year at large has been AI.

Tyler Pathe: This event coincides almost, unfortunately, I think one week short of the one year anniversary of the launch of ChatGPT, right? Of other language models. One other thing that we also talked a lot about today, maybe not as not as intentionally is, is data and the use of data. Now you know, as well as in your work and in my work, I see a lot of, of use of data of harnessing it, of like harvesting it as well, you know, where to find, The best information.

Tyler Pathe: And for the financial services industry, it's become sort of like the water of life. Why do you think

Matthew Cheung: that is? Well, what's interesting at the moment is everyone is now thinking about, you know, what's our AI strategy? And you can't have an AI strategy without a data strategy. And what's happened in financial institutions is in the last 15 years or so, people have been having data lakes and data houses and data warehouses and all, and all these different.

Matthew Cheung: Types of way of capturing siloed data and try and get it into one area. So all of that really is to get you to a point where you can start extracting useful insights and analytics from that data to help you predict things that might happen in the future. So there's no AI strategy, there's no data strategy.

Matthew Cheung: And with having data, the other piece that you need as well is integrations into that data. And, and I think that's where The market and where we're positioned at the moment is particularly interesting because we're getting to this point now where if you do have data that sits inside Snowflake, for example, there's certain ways you can connect to it, you know, using SQL like language to query it.

Matthew Cheung: Which is not that usable to your average person who might be working in a back office role. And now you can have chat as an interface to connecting to that data, which is, which is really, really interesting. The other pieces around having If you want to be able to get more value from data that's not in this data warehouse, where does all that data sit?

Matthew Cheung: It's still being manually shared around in a conversation, maybe face to face, that's probably quite difficult to capture, but definitely on a voice conversation. You know, a lot of the world's markets are still traded by voice, you know, people talking on the phone, and if it's not voice, it's in chat, or it's people typing into chat platforms.

Matthew Cheung: We go back and forth, ask if we agree something. They're not going to a different system to double key it. That type of workflow means that the data is only captured after someone books it in the system. And downstream from there, it's all captured. But what happens before, and this is the bit where we kind of live and breathe and we're operating, is we can capture all of this data that's never been captured before.

Matthew Cheung: So then there's a lot of, there's a lot of value in that. So, there's different ways you can capture it as well. You can be capturing it for to, to, to gather prices for that, for the, for the user. You can be doing it to gather analytics and insights. We're doing it to help like trade surveillance.

Matthew Cheung: There's many different use cases using the same technology to be able to capture the information. So I think data is, there's many different levels of data. Where we live is, is on the kind of end user generated bits of data where it may be, you know, Maybe one price or maybe a list of prices that you can see with your eyes on a spreadsheet that type of data So far has fallen outside of any of these big data projects.

Matthew Cheung: So I think it's super interesting so that's now going to be added into the mix as well and Again, AI is an enabler for that because it enables you to capture that data Structure it using it. You know, if it was an image, you use computer vision on it. If it's, if it's a conversation, you could use a large language model on it, whichever, which, whichever AI kind of tool or approach you take, it makes it easier to take the unstructured information, put a structure on it.

Matthew Cheung: And again, automate it or gather some analytics or insight.

Tyler Pathe: You know, and then those that will be turning to iBridge pool will be able to Squeeze every drop out of their data in a way and and really put it to best use, you know in such a competitive landscape, I think is really is necessity is a necessity for firms to really be tapping all of their resources to the max.

Tyler Pathe: And it sounds like that that's, you know, your company is in right in the center of that ability. In terms of mixing something like AI with sort of. Traditional like financial mapping and parsing tools. How does iPushPull allow that to happen?

Matthew Cheung: So our, our service does a number of things where we're connecting to data that may be sitting in these other applications, comes into our service, we can map it, transform it, enrich it, and then fire it off into another application.

Matthew Cheung: The bit that's, the bit that we've kind of got to now is, again, is I think everyone has this idea of the speed and the pace of how stuff's happening. It's quite overwhelming sometimes, because it's moving so, I mean, even this week, you know, you know, a lot of things have happened. And where we kind of sit, and what we've learned in this fast market of AI, is actually there's a number of approaches you can take.

Matthew Cheung: You've got on one end, there's large language models, which are pre trained. But some of them actually don't work that well in very specific financial market workflows. And actually for some of those financial market workflows, you could use something as simple as mapping. Because if I'm trading a particular product, there's only four ways I may ever ask for a price in that particular product.

Matthew Cheung: And there might be only three ways or two ways you respond to it. So then it's very simple just to map those together. And you may kind of put some things in there to to cope with if something was spelled wrong, or you, you know, it was typed a different way around. You can cater for that in like mapping, you know, in mapping tools, which we have already.

Matthew Cheung: So you're mapping on one end, you've got large language models on the other, and in between where we've kind of got to now is, you've then got these narrow models or very process specific model, and then, and then you've got parsing as well. Now what's been quite interesting for us is, as a, as a company, We're not, we're not an AI company.

Matthew Cheung: We're a capital markets kind of technology company, but now AI is so easy to use. It's kind of have, it's got everyone's focus on how you can utilize these tools. Now, if you look at a mapping we were doing anyway, but if you look at passing and you look at narrow models and specific machine learning models, we've now hired an AI engineer that's building some specific models.

Matthew Cheung: For specific use cases, but these two things have been around for like 10, 20 years. We could have actually done that 10 years ago, but the only reason we didn't is we always thought it's going to be a little bit more to, to bite off than we can chew. Having like AI and NLP, it's always something that's been on our roadmap by two, three years out.

Matthew Cheung: Now what's changed now, rather than being two or three years out, you can do it right now today because it's so easy, but these things off the shelf. But then what we've discovered is the things off the shelf. I'm not always fit for some of the use cases that we want. So we've gone back to more traditional methods.

Matthew Cheung: So, but anyway, I think that just to kind of recap that is, it doesn't really matter what's going on under the hood. If I'm a customer, I've got a problem and all my pain is, and I'll understand the value in it if it can be solved. And we're a software vendor. So we're about solving problems and giving value to our customers.

Matthew Cheung: What we have now with all the kind of technology piece is super interesting and intellectually stimulating. But as a customer, they don't really care what's happening underneath the hood. You don't care what's happening on your cloud service when, you know, what's happening on AWS and what databases I provide.

Matthew Cheung: I just care about something that needs to pop up on my screen and that's it. Similar to the customer, they want to solve a problem. So they don't really care about the tech stack. So again, where we sit in the middle is... We can take care of the heavy lifting, finding what's the most appropriate approach, or model, for doing a particular use case.

Matthew Cheung: And over time, that will just expand and expand. And then it will just kind of disappear and, again, it's just a component that no one really talks about.

Tyler Pathe: One of the major points of discussion at today's event has been the integration of AI technology into chat platforms. How do you think such an integration revolutionizes trade workflows?

Tyler Pathe: In the capital markets industry.

Matthew Cheung: Chat has been prevalent in capital markets for a long, long time. You've had Bloomberg. Bloomberg is probably one of the oldest fintechs. And yeah, they've had iBChat. That's been around for a long time. Before that, there was a FX trading platform called FXALL. And it had this very simplified messaging ability to talk to one another.

Matthew Cheung: So chat's been around in financial markets for a long time. Because of the need for you to always communicate to your counterparties to talk about trades and so on. So it's nothing new having chat, but what's really interesting now is then having chat as an interface into AI. Because, and that's what chat GPT is, right?

Matthew Cheung: You know, you can just ask it anything and it returns a very great answer. And now people have have started to use that. They start to think, okay, well what's my chat strategy? How can I use AI across the chat platforms or across the the different applications I have internally? So what, what we're seeing is people embracing chat in a bit of a different way now because of what's happened with AI.

Matthew Cheung: It's people are like, okay, we've got these chat platforms and all these desktops. How can we use these tools? So where we come into it is, you know, we have our own chatbot framework. You can put a chatbot into one of these particular chats, and then you can plug in, you know, different large language models or other models like we were discussing.

Matthew Cheung: The interesting thing in capital markets and financial markets is obviously it's a regulated industry. And because it's regulated, it means, I think one of the topics of discussion today and is always around, how can you trust an AI model if it's giving some sort of decision on something. So where we stand is, You still retain and keep the human in the loop in any decisions.

Matthew Cheung: And what you're doing is you're enabling that, like I said earlier, the drudgery, you know, we can, we can automate all of the painful manual processes that people do. And if you, if you take a, you know, there might be something on a particular trading desk where if I happen to book something post trade, it might take them five, 10 minutes, sometimes 20 minutes to book something.

Matthew Cheung: Instead of doing that. What you can do instead is if I've got a chatbot that's sitting in there, I can take that information, I can structure it, I can turn it into a, you know, a ticket or fire it off directly into an API, and then all of a sudden I've saved 5 to 20 minutes work down to a few seconds. Now, what you do is you keep a human in the loop there, because when it extracts that information, it can display it to the human, you know, the trader or the salesperson, they check it.

Matthew Cheung: And then they can fire it off into a different system. So that's

Tyler Pathe: where the cross sharpening sort of

Matthew Cheung: comes in. So yeah, so then the humans, the human, and they say co pilot, but the human and AI are co pilots to each other. You know, they're helping each other. Like in the last time I was in a plane, right?

Matthew Cheung: You have the two pilots helping each other, and then you have the automatic co pilot. It's the same concept. But the key difference is, is the human is a regulated individual working in a regulated company. Right? So, therefore... You're, you're helping that, that particular user, because they're reducing their time to do something into seconds when it was taking 20 minutes each time.

Matthew Cheung: You're keeping it in a regulated environment, and you've still got the human as the ultimate decision maker. And more what'll happen in time is, you'll be more like a cockpit, but the human's just overseeing various processing. And you have processes and they're just watching a dashboard of things and the humor will intervene when they need to.

Matthew Cheung: That is the general direction of travel that we're seeing. What

Tyler Pathe: can we expect I push forward to do

Matthew Cheung: next. So we've had a a, a great year. Know 2023. 2020 three's been fantastic for our company. You know, with the advent of chat, GPT has kind of put this focus on into chat and AI and data, which is kinda what we do.

Matthew Cheung: And we also had a funding round which we closed six months ago, so we had you know, a major financial institution, you know, led a strategic investment into the company. And those two things have just enabled us to go even faster with what we're doing. So I think next year is all about productizing a lot of the Solutions that we've been building for customers.

Matthew Cheung: So there's totally blood and play so people can move quicker. Because I think someone made the comment earlier from another company where they took something into production in, you know, from idea to two months. We can do that from an idea to like, you know, in a few hours. Really? Because of the nature of all of the data integration we've got, in the application integrations and in the bot framework.

Matthew Cheung: So very much for us next year is expanding out the range of different touch points that we have with integrations, be it more chat platforms, the other industry kind of platforms like OMSs and EMSs and trading platforms and so on as well as adding more predictive analytics or data enrichment and just having, just doing a lot more in chat because going back a few years ago during COVID, that was another really interesting year for us because everyone was working at home.

Matthew Cheung: The only means of communication was to use chat platforms. So all of a sudden, you know, Microsoft Teams users went from I think there were like three or five or maybe even 10 million users, from there to 300 million users during COVID. Because everyone uses Microsoft in, you know, most, most big companies use Microsoft.

Matthew Cheung: It's prevalent everywhere. So, so something like that, again, plays, plays nicely into kind of where we are. Now, I think now that people are using something like Teams, for example, but any chat platform really. But if you're using it for a video call or music, just for chatting to someone, what more can I do?

Matthew Cheung: And then I'm now I'm aware of chat, you can see what you can do there with agents and things like that. How can you start bringing these things together into your business? But importantly, how can you add value into what you're doing? How can you be more efficient? How can I automate things? And that's the, that's always the target for our customers.

Matthew Cheung: It's either delivering a better client experience for their clients using our tools. Or it's making their own processes more efficient and ultimately so they can automate a lot of these manual processes as well.

Tyler Pathe: That's really, really fascinating and you know, you've got a good year ahead, well you're coming off as a good year as you said, and you've got a good year ahead of you, just you've got a fresh capital in your sales.

Tyler Pathe: You've got the whole industry really watching what you're doing and waiting on tenterhooks. To see what you're going to do next in terms of, you know, the spirit of collaboration within the industry, are you, are you going to be looking at, you know, not confirming or denying anything at the moment, but are you going to maybe be looking at working with other companies or, or, or, or engaging in collaboration?

Matthew Cheung: We've been a very collaborative company since, since the start. And that's the nature of, you know, we're based in London and London has a fantastic capital markets, FinTech ecosystem. And the nature of the ecosystem means that we all personally kind of know one another really well. So, so pretty much everything we do, we make sure it interoperates with other companies.

Matthew Cheung: There's various standards that have been formed in different facets of capital markets that we all adhere to. So it means that when you're when companies are looking at building their own solutions using best of breed products, it just works out of the box. And this. Yeah, interoperability has been a key theme for the last year or two, but I think we'll continue to move forward because there's, there's interoperability on the desktop, there's interoperability of data, there will be interoperability of Machine learning models, which is already kind of happening where you've got these kind of multi chain models and so on multimodal models, you know, having the ability for even having AI to select which is the best model to go to, you know, so all of that I think is, is collaboration on technology level.

Matthew Cheung: And on the company and industry level. But yeah, we're highly collaborative.

Tyler Pathe: Fantastic. Well, it's been an absolute pleasure to talk to you today. Matthew Chung, CEO of iPushPull. I could talk to you all day, but thank you so much for, for sitting down with me for this conversation. And again, for, for attending our event.

Tyler Pathe: I hope to see you again very soon. Thank you very

Matthew Cheung: much.