David Jones: So hello and welcome to the ipushpull podcast. I'm David Jones, the CTO and co-founder of ipushpull. We're a data-sharing and workflow automation platform, which is used by banks, brokers, asset managers and trading venues. So a wide range of organizations across the financial markets. In today's podcast, we're going to have a particular focus on one of the ways our customers interact with our platform, namely.

David Jones: Chatbots. To help me with this, I'm delighted to be joined by two guests from companies that we partner with. First of all, Rob Friend from Symfony. Hi, Rob. Hi there. And Simon Coghlan from Euromoney Trade Data. Hi, Simon. Hi, David. So let's start with you, Rob. What does Symfony do and what do you do for Symfony?

Rob Friend: Okay. So Symphony is a collaboration platform. I think we're best known for our chat product, but we also provide voice and a lot of automation products as well to help people with their daily workflows in the institutional finance space. In terms of my role. I'm a product manager here and I look after the market solutions and extensibility side of the product, so everything to do with building in vendor connections all the tools that we use to integrate to third party systems, as well as some of the features that our users want in terms of financial

David Jones: products.

David Jones: Okay, great. Thank you. And how about you, Simon?

Simon Coughlan: Yeah, so today I'm here talking about euro money trade data. I have a broader role. I look after euro monies. derivatives portfolio. But Euromoney trade data specifically provides derivatives data predominantly into tier one banks. There's lots of processes that Euromoney run for these banks that integrate different data sets for all sorts of reasons like trading, clearing, regulatory compliance.

Simon Coughlan: So yeah, we have lots of different products from web products to APIs. It's a large data feeds that are provided via SFTP.

David Jones: Okay, great. Thank you. So, first of all, in our talk about chatbots, we're going to consider the topic at a general level. So looking at how people are using them inside Symfony and examining the kind of chatbots that we at EYEPUSHPOOL have developed.

David Jones: Next we'll drill down into a particular example, which is a chatbot, chatbot called Tadar, or Tadar. This is Euromoney's chatbot, which was implemented by EYEPUSHPOOL and lives inside Symfony. anD then finally we'll try to take a bit of a long view which way is the market for chatbots.

David Jones: chatbots heading and what other innovations are going to change the way that people within the financial markets are working today. So let's kick off. I mean, if I can start off with you, Rob, here's an easy question. What is a chatbot? That's a good question.

Rob Friend: So I think probably people who are familiar with chatbots from their personal lives you deal with them When you talk to your power company to your phone company To your bank and essentially, you know, it's the same experience, but on the professional side of people's lives in finance So it's the way that a human can talk to a machine in a kind of human friendly way So a chat with a machine is the chatbot

David Jones: Okay, great and What kind of chatbots are people using inside Symfony?

Rob Friend: So we've got a wide variety of chatbots inside symphony. I think the most basic chatbot performs a function of really just notifying people when an event has occurred or when a piece of data has become available. So really just a push notification. So very simple. Interaction. We also see slightly more sophisticated chatbots where a user requests some information from a chatbot.

Rob Friend: And then I guess the third and probably the most sophisticated is where there's a whole workflow base behind that chatbot where a user can start a process, interact with a chatbot get taken down a decision tree path and eventually get to the, you know, the outcome that they were looking for. So there's a real range of chatbots

David Jones: that we see on the platform.

David Jones: The, the bots that that live on Symphony do, do you write these themselves yourselves, or do companies implement them?

Rob Friend: So, so we provide the tools to allow our, our clients and vendors like I push pull to build these bots. So, mm-Hmm. You know, there are some bots that are built by Symphony for some of our workflows both internally but also with clients.

Rob Friend: But the majority, and there are about 2000 individual bots across our, our network are built by either our clients or by vendors for our client. ,

David Jones: that's an incredible number of bots on the platform. I mean, how, how, how does that compare to the number of, of human users on Symphony?

Rob Friend: So, to give you a sense of scale, I think, you know, the network that we have today covers about a thousand firms or over a thousand firms.

Rob Friend: These are all financial institutions. Mm-Hmm. . And in terms of the number of users, we've got approximately about half a million users. And so the majority of those interactions with those bots is normally a human to a bot. It could be obviously one human to. Yeah, the bot can communicate with many humans.

Rob Friend: But we also have seen examples where bots actually communicate with other bots. So we start to see more sophisticated interactions where you can have machine to machine interactions as well as human to machine interactions.

David Jones: Okay. One of the key things about Symfony is the fact that it provides this kind of secure messaging layer between different marketing.

David Jones: Market participants. So any message that Bank A sends to Bank B, or between a broker and a bank, or between a bank and a buy side, they're all encrypted and secure. Do you find people using chatbots for... Yeah, as well as just for internal workflows inside an organization, do they use them for communicating with their clients and counterparties?

David Jones: Yeah,

Rob Friend: absolutely. I think you know, the initial use cases tend to be internal when people first start to use these sort of chatbots because they want to make sure that they they can control them and understand how they behave. But we definitely see. More and more uses for external chatbots.

Rob Friend: So whether it's to surface research to a client to help with that kind of customer service experience, so it could be decision trees. We see them on the operation sides of financial institutions where issue resolution can be speeded up with a bot interaction. It's maybe a combination of human and a bot from an institution talking to.

Rob Friend: Another human on the other side of of that interaction. So, there's a real, a real wide variety now, but definitely more and more we're seeing the external bot adoption increase.

David Jones: Okay, and, you know, why do people like using bots, do you think? What, what's, why don't they just use sort of a dedicated application to do the kind of workflows that they, that are being presented by the

bot?

Rob Friend: Yeah, there's definitely a place for both, and I think what we see is... used in parallel, so I don't think it's an either or. I think the advantage of putting a a bot over a network like Symfony is that the connections to the end user... So you don't have to go through firewalls. You don't need to deploy software.

Rob Friend: You don't need to teach someone to go to your web page and put in their credentials. All those things are kind of taken away and dealt with for you. So in terms of the friction to get to the end user, it's much quicker to implement a bot and get adoption there. What we see, you know, when you get into some of the more sophisticated, Workflows is you can then the bot can act as an entry point, maybe into that more sophisticated platform or front end that you want to take people do.

Rob Friend: So I don't think it's a one and either or it's definitely use them in parallel and use them. Yeah, really do compliment each other.

David Jones: Exactly. I mean, we've certainly from our customers. The reason why they like bots is, and this is a direct quote, is because they provide an easy on easy on ramp to an application or to a platform.

David Jones: yoU know, one of our one of our broker customers they provide lots of ways for their. For, for, for their clients to interact. They, they provide a a screen, an application that people can install on their desktop. Or they provide things like fix interfaces that they can hook up their you know, their internal platforms too.

David Jones: But both of those are problematic. You know, if, if you, you go into any kind of trading floor you can see that each end user has multiple. Monitors, you know, it could be anything, four, six, up to eight monitors on their desk at a time. And that, you know, in spite of that, the kind of screen real estate, as they call it, is really limited.

David Jones: It's really hard to get another application onto somebody's desktop. So users often just will not, they just choose not to run the application. So, so. Being able to present that kind of workflow inside something they've already got running, like a chat platform, like Symfony, that's already open on their desk.

David Jones: It really reduces the friction of onboarding that user to the

Rob Friend: workflow. The other thing that is important in that example is you can also surface that information when you need to, so the user doesn't have to have the application open all the time if they're in Symfony. And something occurs, whatever the thing is, you can then use the bot to highlight to the user that they need to take an action or do something rather than having to, like you say, get that real estate in one of those six screens and have it constantly up there and be watching it.

Rob Friend: You can actually use the you know, the symphony real

David Jones: estate for that. Totally. Yeah, people don't need to have something open all the time. Being able to sort of tap them on the shoulder, figuratively speaking, when something of interest happens is, is, is really useful. It's also

Simon Coughlan: the low cost of implementation, if you've already got the platform.

David Jones: Yes, exactly. Yeah. That's right. Because we have well. A great example of that is, is Fix integrations. You know, Fix is fantastic if you want to, if you want to hook up a platform in company A to a platform in company B. But more often than not you know, the companies. If there's not a, if there's not a really valuable business case for doing it, then they're not going to expend the time on time and effort of implementing that fixed interface and then running it, running it 24 7.

David Jones: So you know, by providing, platforms that allow you to say hook up fix into chat, which is one thing that we've done is it, you know, again, it provides another easy way of getting the getting the workflow into an application that somebody is already using without having to do an expensive implementation or or deployment project.

Simon Coughlan: That's, that's actually a use case that ETD could use as well, because we have fixed integrations coming into our cloud environment. So we could surface that via iPushPool in the Tadar bot within Symfony.

David Jones: Well, we'd be delighted to speak to you about doing that. So. I think we're talking a lot about the simplicity of servicing data or workflows using using bots.

David Jones: They obviously do need some development but Rob, I know that Symfony's implemented some, you know, some toolkits and frameworks to allow people to, to easily build bots. Can you talk a little bit about, a bit about

Rob Friend: those? Yeah, sure. So we have a number of tools that we, we provide. There's a bot developer kit, which the idea is helps developers build a bot reasonably quickly.

Rob Friend: You know, we obviously have our, our APIs, which if you really want to go down to kind of the nuts and bolts, you can, you can connect to those. But the, the bot developer kit or BDK is really there to help you. Build a bot in hopefully, you know, days or maybe a couple of weeks. And then more recently I think we've, we've released a workflow developer kit, WDK.

Rob Friend: Which really tries to go a bit further for the The less sophisticated programmers like myself out there who maybe know what they want to build, have an idea of the workflow they'd like to see and can visualize that, but maybe aren't expert Java programmers or you know, ready to go, go down into that level of detail.

Rob Friend: But it's really there to help us. Get those bots out to the people that, that want them and, and allow more people to build them. Because I think the thing that, as we're talking about, the great thing about bots is, you know, there, there's not much friction to actually getting them deployed. There's still friction.

Rob Friend: You still need someone to build them. And the more that we can provide these tools that allow users to build them swiftly without having to learn lots of new languages and lots of symphony specifics, the better it is for us and the better it is for the end user as well.

David Jones: Yeah, we've used your your bot developer kits.

David Jones: We started off using the API directly. But the BTK makes it really straightforward. We found to talk about our code, which handles that kind of integration with data and with the workflows that we implement and then kind of expose them into symphony itself. Yeah, we've used the Python. Flavor of the of the BDK.

David Jones: Yeah. We haven't tried the workflow developer kit yet. But yeah, watch this space It did it's still relatively new.

Rob Friend: So yeah, it's

David Jones: It's a really intriguing, yeah, it's a really intriguing prospect. So it's a, yeah, it's something we are going to look into. Is

Simon Coughlan: the BDK essentially an abstraction layer over the APIs?

Rob Friend: Essentially, yeah, yeah.

David Jones: So it kind of removes a lot of the complexity of hooking up directly to the... To, to, to, to the API, you know, you it simplifies things like authentication and so on. It's it abstracts that kind of thing away. yEah, so we've so I, I suppose we should drill down into a little bit about how we've we, I push Paul have implemented our bots.

David Jones: So. Well, you know, Symphony, you've got your own bot framework, which allows it makes it really straightforward for companies to hook up to Symphony chat. We've got our own bot framework as well, which allows us to hook up. Data to data into a into a chat workflow and then makes it really straightforward to create bot commands that allow people to slice and dice the data that they want to get into the get into chat.

David Jones: And this is something that we've, has been really important in, you know, in producing Tadar. The Euromoney bot that we've, we've implemented together the big challenge with, with Euromoney data is, is the size of it. There's a great deal of it out there and it updates quite you know, quite frequently.

David Jones: So Simon, do you want to you know, tell us a little bit about what what kind of data Euromoney has and something about maybe the traditional ways that you. You get out to your customers. Sure.

Simon Coughlan: So we at Euromoney Derivatives are taking data from all kinds of sources. So we're taking data from many of the big traditional data vendors on behalf of our customers.

Simon Coughlan: And those data sets are traditionally pulled via SFTP. And they tend to be the universe of data. And we'd be talking about, you know, potentially 30 million rows of data, over 150 columns. So you've got these gigantic databases. We've also got data, as I talked about earlier, FIX, where we'd be pulling specific exchanges and specific instruments into our database.

Simon Coughlan: And then we've also got various scraping operations to get free to air data and condense it down into a common model. Now, typically how we serve that out is, again, back out to customers as SFTP deliveries. That could be delimited text files, it could be XML, it could be JSON. We surface that data in a web product with lots of search functionality and so on.

Simon Coughlan: And we also have an API product that customers can consume on demand. What the bot gives us is the ability to surface all of that data. And it's updated in near time, just by very, very small amounts. So a customer, for example, could search for an identifier. They could get that related data back straight to their screen.

Simon Coughlan: And for example, fix trade breaks. And that's really a new use case for us. We see lots of potential with that bot on the Symfony platform.

David Jones: So, so what kind of data is it that you, that you have? I mean things like instrument static data, that kind of thing?

Simon Coughlan: Yes, yes, it would be. It's, yeah, futures and options data.

Simon Coughlan: All of the attributes around that. So, regulatory data such as ISINs, for example, CFI. Then all of the associated series data from the vendors. So what's trading in the market, potentially some pricing information. So generally it's static, but we can also get a flow of data in near time from our vendors that have up to date series.

Simon Coughlan: So strike prices and so on.

David Jones: So, the way that we that we've connected to your data is, well, there's various, iPushPull lets you, lets you get data into our platform and then expose it through chatbots in various different ways. At the very basic level some customers just have like an Excel sheet and they can push the data straight into us.

David Jones: But for something that's a bit more, Systematic like like we've set up with Euromoney. We use two different approaches. One of them is, is using our data loader. And that basically does a daily query against some of your databases and then pushes the data in through our API. And that's suitable for things like smaller data sets, things like daily corporate actions, which is generally a few hundred rows of data, I think, that kind of size.

David Jones: For some of the larger data sets, like instrument static, you're SFTPing that to us. And then we import that into our platform and then once it's in on a push pull we, we basically, we've worked with you very closely, I think to define the set of queries that you think your end users will be most interested in being able to be able to use.

David Jones: So you can take a piece of data and you can use our interactive tool to build. Some standard bot commands. So, for example, we have one like CA today and that gives you all of the corporate actions today. You can have corporate actions CA next, I think, which gives you which takes a parameter, so you could say give you the corporate actions for the next five days or so.

David Jones: And then there's some queries that do a sort of a much bigger search across the full data set. So that's for things like searching for static data searching for instrument definitions and so on. So this is you know, these are the kind of things that we've worked on to because we think they're the, the kind of queries that end users will.

David Jones: We'll be interested in using so, so Simon, why do you think that exposing these kind of things inside a platform like symphony in particular, I guess, is going to be particularly welcome to end users?

Simon Coughlan: Well, essentially what we're able to do is, is enable the users to get to specific pieces of information with a set of defined commands without the users having to set up the processes with setup to pump data in at scale into their environment.

Simon Coughlan: Essentially, what we can give them is a command line interface with a set of intuitive commands that enables them to surface the data. Without any implementation cost of implementing the data flows that come from all of the different data sources and databasing it and all of the maintenance that goes with that.

Simon Coughlan: Essentially, they have one common interface. They can pull the data out of on demand.

David Jones: Yeah, that's right. And I think with one of the Benefits of Symphony is that, you know, this Tadar bot simply lets you, it basically appears like another human user inside Symphony. So you can you can search through the Symphony directory and connect to it.

David Jones: And once you've connected to the bot. You can start using it immediately, so there's no installation and users don't have to do any contact switching to you know, to start using it. I mean, we've got, the bot's completely, you know, is available inside the Symfony directory. This is more a question for you, Rob, I guess.

David Jones: What's the kind of... You've mentioned the symphony has got hundreds of thousands of users. What kind of organizations do they work for? And you know, what's your, what's your kind of spread across the financial

Rob Friend: markets? Yeah, I think we, we probably touch. Every financial institution type you can think of, so from the, the big tier one sell sides through to you know, obviously the tier twos and tier three sell sides, the, the banks, we have entity, the brokers, we've talked about, you know, them earlier on the call asset managers, hedge funds so you name it, we probably have many of each flavor.

Rob Friend: And that, that network's, you know, growing constant basis. So you know, I think there's the use case for the Tadabot specifically is probably more on the sort of support operations, trade break side of things. And, and that user base, you know, is on Symfony too. So as well as the front office, we also have the middle office and, you know, and operations users.

Rob Friend: So real spread of institution types and also user types within those institutions.

David Jones: So I think it provides a really good route to market for, for organizations like Euromoney allows them to sort of reach out to, you know, maybe tier two, tier three banks asset managers and so on, who, you know, who previously might not have been able to get hold of the kind of data that the Euromoney provides.

David Jones: Yeah,

Rob Friend: I think maybe and this may be one for you and Simon, but I think the actual kind of... Cost to build versus, you know, how you'd have to develop a tool and then market it out is probably a much better value for money to kind of build these things. And then, you know, if there is adoption, really kind of invest and build on it from there.

Rob Friend: So much faster route to market, a much more agile kind of way of getting a new product out, I think.

David Jones: Yeah, totally. I mean, we provide a way of just hooking up your data configuring some commands and then off you go. Really, it can be. It can literally be as quick as that. I mean, some of the stuff that we also provide that perhaps end users don't see.

David Jones: But it's very useful for the Euromoney sales team is all the kind of audit and monitoring information that we provide in the background. So we're able to see when users sign up. We can see the kind of, you know, if there's any, if they're making consistent mistakes with, with any of the commands.

David Jones: So it can help us to sort of tune the commands. If people keep making the same kind of error it helps us to think, okay, well, actually, maybe we should change the command to make it something a little bit more, more understandable. Of course, we also have

Simon Coughlan: the analytics. So we can see by, by looking at the Ipushpool logs, for example.

Simon Coughlan: That customers are issuing certain commands to get specific data sets back that enable us to really profile what the customer is looking at and potentially tailor more solutions for that customer. around what they've already looked at. So that information is really useful to us in terms of analytics.

David Jones: Yeah, totally. That's the, I mean, we've seen that in some of the other workflows you've implemented using our, this is with our workflow framework. One of our customers uses us for a, for a pre trade workflow. And this was previously. Spread across email, chats, telephone and there was no kind of consistent way of, of, of recording you know, who's responding fastest, who's the most reliable person involved in these workflows by bringing it into a single bot framework.

David Jones: You know, it means that we can collect these kind of statistics and you can use them for kind of broader. You know, collection of metrics and so on. So if Ultimately, we want to start using this kind of information for predictive analytics, but I think that's a little bit further down the road.

David Jones: Yeah, just to talk. We mentioned fix a little bit earlier on. I think that this is something that we we are particularly excited about. Is the is what Symphony provides as a kind of market infrastructure in that it can actually sort of help to help to replace some of these quite difficult to implement you know, implement processes like fixed protocols.

David Jones: Rob, can you can you explain a little bit about the market infrastructure side

Rob Friend: of Symphony? Yeah, sure. I think one of the things that we hear. Quite frequently is the kind of last mile problem. So obviously, like you talked about, you know, big inter dealer brokers and big sales sites have fixed connections set up.

Rob Friend: And you know, that's, that's great and works well. But as soon as you start moving down the customer base getting on getting a fixed connection in place or getting onto someone's queue to build a fixed connection can take a long time. So where I think simply has a role to play is really that's a last mile connection between.

Rob Friend: One firm and another where we can actually use the infrastructure we've got, which is a secure encrypted chat room, which you could also call a data pipe to connect those two institutions. Now, you know, kind of health warning is it's, it's built for chat. So I wouldn't recommend that anyone starts putting down their level two equity market data because it probably isn't.

Rob Friend: But when you're talking about kind of pre trade or post trade workflows, there's definitely a role to play there where Symfony Connect is that kind of connector that allows you to. Quickly and easily get those, those those connections

David Jones: in place. Yeah, that's right. So what we've done in our fixed chat workflows is we've done the hook up to one of our customers fixed services and then we receive messages.

David Jones: From that customer via fix, then we inspect the message so we can see, okay, well, who's this supposed to be sent to? And what's this message about? And we use that information to route that into a symphony chat room. So our bots basically takes the fixed message, decides who it needs to be sent to, and then converts it into a human readable Message, which could be something like a trade confirmation, for example, and so that's what we're increasingly seeing is on the workflow side is customers wanting to automate these kind of, you know, these quite sort of tedious workflows where somebody's, you know, a human user is being expected to type in a trade confirmation into into a chat room, but by hooking up Processes like fix doing this kind of message routing and then mapping into human readable form and sending that message into a chat room, you know, it can really kind of streamline people's processes and make things a lot more make a lot more efficient.

David Jones: So, yeah, there's a whole level of of workflow that can be built on top of on top of chat bots. You know, we've talked in a lot of detail about the data. Workflows to request data that they're interested in or ask the bot to, you know, tell me when something changes about the data I'm interested in. But I think certainly the next exciting phase that we're moving into is, you know, is is more workflows being all being being automated and brought into the chat world, which means that people don't need to be switching out of different applications.

David Jones: They can just stay inside something that's already running on their desktop. Rob, have you got any, any thoughts about you know, what, what's happening next? Where do you think where do you think the, the, the workflows are moving?

Rob Friend: So I think the, the thing that is going to be interesting over the coming years is that hybrid world.

Rob Friend: So we, you know, we see. Voice is still going to exist in the markets. You know, people get a lot of value out of the voice. Obviously, chat has been with us for a while now. Email, unfortunately, is still with us. My inbox is out of control. And I'm sure we'd all love a much better, yeah, a much better workflow than email.

Rob Friend: And then obviously now we've got chatbots. And those, you know, those are more and more prevalent in our personal and professional lives. So, I think the kind of, the future is, Probably, probably how do those things work better together? So when voice and machine, so bot and human can work together.

Rob Friend: I, I, I'm, I'm imagining a world where, you know, those things start to blur more and the more that we can kind of build those interfaces so that You know, machine can talk in a human, understandable way to a human. So a chapel on a voice transcription, and then you start to get into all the kind of A. I. N. L.

Rob Friend: P. And all the all the information that's held in those voice conversations. Unlocking that I think is the next big challenge. And so how we bring those three kind of parts together is really going to be the challenge and gonna be exciting. And that's where a lot of efficiencies and benefits are going to come from, I think, is trying to remove all that friction and And all the rekeying and all the lost information that sits somewhere in a database that no one can actually get access to.

Rob Friend: So, there's lots

David Jones: to do. Yes, totally. How about you Simon? Well, yeah, I

Simon Coughlan: broadly agree. I mean, I think really talking about our bot, Tadar. reAlly the next stage for me is being able to talk to it directly. Rather than issuing the commands. So that's, that's where I see it going next, where there's still the use of voice, but the voice is talking to the bot, the bot's doing all the heavy lifting.

David Jones: Great. Well, we better get started. Let's get, let's get working on it. Great. Well thanks very much to both of you. It's been a really interesting session and thanks for your insights and look forward to catching up with you again soon. Thank you. Thanks a lot. Thank you both.