Neil Weatherall: Hello and welcome to another episode of the ipushpull podcast. What is your chat strategy? Joining me today is Roy Kirby, head of core products at SIX. So Roy has significant experience and expertise in the market data arena. And we first met a year ago as SIX were looking to innovate in the corporate action space.

Neil Weatherall: Our conversation comes hot on the back of your corporate action calendar launch, and November saw the launch of SIX BOT. This symphony chatbot provides the latest information on more than 70 corporate action event types, and is available to over 600, 000 financial market professionals within the symphony community.

Neil Weatherall: This was built by SIX and ipushpull. In response to increasing industry demand for greater automation of the corporate action processes, and it was built with the hope of freeing up time for front and back office personnel to focus on higher value tasks. My name is Neil Weatherall, and I'm head of technical sales at ipushpull.

Neil Weatherall: I've worked heavily with Roy and team at six on this project, and without giving too much away. The work is continuing on new initiatives. So thanks for joining us, Roy. Could you introduce yourself, please, and maybe tell us a little bit about your role, what your role involves?

Roy Kirby: Yeah, sure. Thanks, Neil.

Roy Kirby: And thanks to ipushpull for this opportunity to talk on your podcast. So, yeah, my name's Roy Kirby. I am, as you said, head of core products at Six Financial Information. What does my job entail? It really entails Ensuring that our customers and our prospective customers can access quality data be that corporate actions data, price data, or reference data in as many ways as possible to help them to improve their processes, to make efficient decision making decisions.

Roy Kirby: And yeah, it's. It's really to extend our customers usage of our quality data globally.

Neil Weatherall: Thanks. So I guess that there's a number of things we can talk about. I mean, Specifically on sort of chat and chatbots. When did you first I guess start realizing them or using them and Wonder if this could be applicable in your sort of space

Roy Kirby: Yeah, so I think in particularly in chatbots and we should we should probably look at internal versus external a big company like ourselves which processes An awful lot of data and has customers that it serves uses chatbots internally.

Roy Kirby: Okay, so we use chatbots internally and we use chatbots. alSo externally for things like customer support. So, when you have customer tickets, when you're talking to a customer, you sometimes use chatbots, and we certainly use chatbots to speed up the process and to magnify the process and deal with lots of customer interaction.

Roy Kirby: So we've always used chatbots internally. And we've seen it as a good tool, a good communication tool. So then we started to think well if we're using this internally and we're communicating with customers about some quite complex topics when you're discussing data sources and data fields, when you're talking to a customer you can use a chat bot to really solve some quite intricate Workflow type issues that they might have.

Roy Kirby: We thought, how can we take that into our products and services? And we then said, right in, in the world of corporate actions at the moment, there is an awful lot of manual work that's still done. There's a lot of email that flies around. There's a lots of discussions that I had. Could we use something like a chat bot?

Roy Kirby: In the corporate actions world to simplify those workflows to make people, people's decision making criteria easier, slicker through a chat mechanism. And that's where we then said, yes, we think this is a really good idea. We think this can work. We had discussions with one or two customers about that.

Roy Kirby: And that's when we. Came to yourselves and said, look, we have not got much chat bot experience internally within six, how can I push pull, help us to go to market quickly. And, and that's what we did. So it was really identifying a problem, which was the complexity and the manual processes around corporate actions, looking at something that we knew worked.

Roy Kirby: The chat mechanism chat bots that we knew worked from our support organization, putting that together with yourselves to fix a customer problem. I think you wanted to know what, what was our strategy around chat? That's not really a strategy around trap chat in particular, or a chat bot in particular.

Roy Kirby: It's a strategy around making the life of our customers. easier and the chatbot technology that you have and that you built for us allowed us to do that and allowed us to do it quickly and with a real sort of target market involved.

Neil Weatherall: And I guess out of interest, what, what chat platforms do you use externally?

Neil Weatherall: And are they the same chat platforms that you have been told that your clients would like to use with you when they want to communicate with you?

Roy Kirby: Yeah, so the two platforms that we, we typically use internally six includes an exchange business and the six group itself as as you can sort of understand is.

Roy Kirby: is quite heavily regulated. So we don't, we don't internally have a mass of chat tools. What we have is one chat mechanism that we use internally amongst ourselves and that is based on Microsoft Teams. Historically it was on Skype but now we all communicate internally on Microsoft Teams. But we know that externally our customers Do not rely exclusively on Microsoft Teams.

Roy Kirby: We looked at our customer base and we found that the one of the biggest chat platforms open chat platforms. I should say out there was the symphony platform, so we decided to link our chat bot into symphony. Day one because as you stated in your your opening, discussion Symphony already had six hundred thousand end customers.

Roy Kirby: It has license agreements with lots of financial institutions So we found that as our best first method of going to market with the corporate actions bot

Neil Weatherall: great so I guess one of the things that I noticed that came out of covid was perhaps a Huge uptake in the use of chat that at times could be overwhelming.

Neil Weatherall: I mean, I, I came from a front office trading background where we were on IB all day, every day, but I definitely noticed a, an increase in the use of chat rooms to try and make up for the fact that you weren't sat next to people. I'm always interested to hear whether other people had the same or whether they thought it was more of a, a natural progression.

Roy Kirby: I and again, just, just talking internally, we had, we had exactly the same, you know, and the, the rollout internally of Microsoft Teams actually coincided with COVID. So it gave us a natural chat capability out of the box on the corporate laptop, which of course, uh, created lots of communication. And I think.

Roy Kirby: To your point, there's a little bit of over communicating as well. Yeah, there's, there's too many chat rooms. There's too much going on. And I think, I think that's where things like bots can help, because actually as long as you can ask something, a simple question and get a response with some quality data in the background you can take away a lot of the noise of chat as well.

Roy Kirby: I think The bot that we've built with yourselves is not just a, it's not just a chat mechanism, it's it's informing the end user as well. It's not just a hello, are you there bot type thing, which is sort of typical chat. It's can you give me some corporate actions? Yes, here's the corporate actions.

Roy Kirby: He's lots more background in the in the back as well. So it's more than just a chat. It's more. I see the chat bot tool as more of an information sharing. Platform.

Neil Weatherall: I mean, it's, that's a, a concept. We sort of, we try to publicize quite heavily in that we use the term self-service. So rather than having a strict timeline or process of, you know, you having to ask somebody else for some data, it goes into a queue.

Neil Weatherall: They literally retrieve the data and push it back to you. The idea of self-service data as a service is about sort of empowering people with their ability to, to get it for themselves. Often that means it's coming back much quicker because it's Essentially a database query or a chatbot query, which is measured in milliseconds, rather than relying on somebody else being sat at their desk, waiting to answer your email or your phone call, not being on their lunch or tied up in other meetings.

Neil Weatherall: Has it. So in, in terms of allowing people access to the data, have you, have you had to rethink perhaps how you permission people or whether like internal data can be shared as widely because obviously the role of that help desk person is, is not only as a sort of client contact and support, it can also be a sort of gatekeeper to the, to the right information.

Neil Weatherall: And if you're trying to open that up, then that obviously implies that you need to think about how are you going to open it up?

Roy Kirby: Yeah, you do. You do. And obviously with data, entitlements is very important. But again, I think that's the great thing with With the platform that we've built is that it can be entitled as well.

Roy Kirby: So we can entitle an organization, but within an organization we can entitle different user types as well. So, yes, it's important, but actually, as long as you understand the data, as long as you understand the entitlements, you can actually put that through a bot to get the right information to the right People at the right time.

Roy Kirby: I liked your self service quote, and I I would say that particularly in the corporate actions, but what it's allowed companies to do is they've already got this data. They're already licensed for it. It's already there in their security master or back office used by a few people. Through the extension, through the symphony bot that you've built for us, you can extend that to the middle and front office.

Roy Kirby: So you can reuse the same data across the whole value chain within a company. And that exposing of data when people need it is really important and really key and is a feature of this type of bot application.

Neil Weatherall: So one of the things I've always been curious about is My, my knowledge of corporate actions and market data is slightly limited, although it's increased significantly over the last year.

Neil Weatherall: You know, I, I would look at it as perhaps an industry where overnight large files get delivered. They then get consumed by your client. Whereas you as the data provider, how do you, how do you know exactly what they're using it for? How, how do you, how do you get the feedback that you need in order to create new products?

Neil Weatherall: And. Is the use of chatbots going to improve that for you because you're going to see it almost in real time as in what, you know, what people are doing, when are they using it, what time of the day, is it always the same securities or types of question like how, how, how do you view that aspect of it?

Roy Kirby: Yeah, I, I think that's a very good point that you made.

Roy Kirby: Historically, how we, how would we have found out? Whether we had the right data, whether it was being used was very much, and we still do this, obviously, direct content with the end customer to see what their problems, what their issues are, that will continue and needs to continue the way I see the chat bot world forming for us is yes, we will be able to take in some of those questions.

Roy Kirby: And to your point, does everybody call it a corporate action? They may not. Does somebody call it the life cycle of an equity or the life cycle of a bond? Do they always just talk about dividends when they're talking about corporate actions? When people start to chat and talk to the bot in their own language, We'll be able to adapt based on the language that they, they use.

Roy Kirby: Now we fully understand the corporate actions universe, the corporate actions terminology and language. And as we go to large language models and natural language inputs to the bot, I think we'll see some differences. Between the two languages and it will be up to us product wise to close those gaps so that somebody coming in with a term that they're used to gets the data back the way that they wish to see it.

Roy Kirby: And I think that's where the bot technology will help us because it will give us a a much larger slice of customer interaction. Customer language that we can look at to make sure that we're. Distributing the data in the right way so that people can understand it.

Neil Weatherall: Yeah, you've preempted my next question there, so thank you.

Neil Weatherall: So yeah, I mean, the rise of AI LLMs is obviously exponential. One thing I've been particularly interested in is, how do you choose the right part of it to use for your use case? Are you applying the right technology? Are you using a sledgehammer to crack a nut? And I think one of the interesting distinctions is, is around, you know Using chat GPT to write your essays and whether you're going to get plagiarized, you know, whether you're plagiarizing or not, and are you going to get caught is very different to.

Neil Weatherall: interpreting different language that is used to ask a question. I mean, my, my, my idea on this is and obviously we can disagree. If you're in financial markets or this sort of world, you're typically not writing a perfectly eloquent sentence. But you are using different terms or terminologies or dictionaries to ask the same question.

Neil Weatherall: Be it dividend or div or payment or PMT or, you know, all sorts of slang and syntax that maybe how you use it locally, maybe how you use it because you're used to using a certain platform. So I think that's where it's interesting.

Roy Kirby: It is, and if I can give you a an example, we, we also have a desktop product that we call 6ID, which is your traditional financial information desktop product.

Roy Kirby: We've been Not playing, it's more than playing, but we've embedded a chat GPT type tool within that desktop and what we've found is, to your point, somebody might say, can you give me the EPS for Nestle, okay? And using the new tools, you can take that very simple language. Give me the EPS earnings per share for Nestle.

Roy Kirby: We can go to our databases. We can look up Nestle. We can find it on the Swiss exchange. We can find today's earnings per share. We can send back the answer. The earnings per share for Nestle is. X, Y, Z. Actually, though, what you can do is you can do more than that. As long as you have access to the data, you can say the earnings per share for Nestle today is X, Y, Z.

Roy Kirby: However, historically, the earnings per share, and here's a table for the earnings per share, you can give more back. And that's where I think, again, you've got this step between the, the very simple sentence. Understanding it, you can use an off the shelf tool to understand it. You then need to, as you're training the model on your internal database, say, Okay, what value add can we put into that?

Roy Kirby: What additional information can we give back? So that the end user doesn't have to ask a second question.

Neil Weatherall: Comparative stocks, what were their areas? Yes, all sorts of things.

Roy Kirby: Exactly. Exactly. Yeah, you know, and very, very simple. What, what is the share price for Microsoft? Well, did you mean the Microsoft share traded on NASDAQ or did you mean the Microsoft share that might be traded somewhere else?

Roy Kirby: You can do the, you can do the value add as well. What's exciting to me is the language part of it is almost sorted out right from off the, off the shelf tools. It's the value add piece that people like Six and people like yourselves can bring. It's the, it's the bit, it's knowing the financial markets, knowing some background that serves up the additional data.

Neil Weatherall: And how, so is that in production at the minute or is that sort of beta testing?

Roy Kirby: It's, I would say, yeah, beta test is probably about the right description for it. Before we go into production, we just want to make sure that it's all bulletproof. So we're using it internally. But yeah, it's been a very good experience over the last six months to integrate those types of tools with our data

Neil Weatherall: sets.

Neil Weatherall: So one, this isn't a loaded question. Let me just make that clear up front. So one of the other important. Factors I think people need to consider with use of LLMs is if it's generative, there is this element of randomness. I think it's called temperature that allows what it produces not to be just the most statistically likely, but there's an element of randomness that allows it to appear more human in its responses.

Neil Weatherall: With financial markets data. I don't think you want to be that human. You want to be

Roy Kirby: accurate. No, you do, you do. And I, I, the way I see this is all of these things, chat, bots, LLM, they're all tools, yeah, that help people to be more efficient. Yeah, people to be more efficient and allow people to do a little bit more.

Roy Kirby: But the decision making is still with the individual in our world. Okay, so you, you don't want it to be too generative. You want to use the tools to help you to make those decisions. But at the end of the day, it's still your decision. The regulator is going to regulate you on your decisions as a firm or an individual.

Roy Kirby: What we can do with these tools is. Give you more information to make a more informed decision on quality data sets that you trust and we can expose those to you through the tools, but at the end of the day, it's still you that's got to go and make that decision.

Neil Weatherall: So you said an interesting word, regulator, how, how, how does that, how do they view the use of?

Neil Weatherall: LLMs in terms of the data that you're presenting back within a screen. Is that something you've had to consider?

Roy Kirby: It's no at the moment. It's not something we've had to consider because when you when you look at the the regulator particularly EU wide around, Around these tools and technologies there.

Roy Kirby: There are only certain areas that they actually legislate on particularly and it's particularly around things such as individual credit worthiness. So if you were doing a credit score, right, the regulator heavily wants to get involved in that, right? But there are lots of regulation, for example, investor protection regulation, right?

Roy Kirby: You're still bound by that regulation, whether you're, you're drawing something that you're going to post to a customer or you're using a chat bot to come up with a decision. The overlying regulation in, in the industry applies. independently of what technology you use. You still have to, you still have to make the right informed decision with the regulation in mind, be that a MIFID 2, a GDPR, a Basel 3, whatever.

Roy Kirby: You're still, those umbrella regulations apply independently of what technology you use.

Neil Weatherall: So, so you've sort of talked about the beta testing of the LLM to provide, let's say, more informed answers and help people think about What they might be asking next and give a rounded view of something moving on from that.

Neil Weatherall: What do you see as the next step? Is there anything that you're sort of thinking comes next and you're excited about? Is it, is it going to spread to other areas of your business?

Roy Kirby: To be honest, I think that if you look at the move from where we are today, right, and to getting into production with some of these tools, I think that's our first step, right? I think that's it's still quite a big step. I think it takes us into a completely different type of interaction with our end customers.

Roy Kirby: So I haven't really thought. Beyond that, at this point, I'm thinking as to how can we take some of these things that we think are really exciting internally and productize them and get them to the end. That's the thing for me that's really interested in the next three to six months. And in six months time with the way that technology is moving today, we might be talking about something different.

Roy Kirby: So get some of this out there and in production is our first

Neil Weatherall: step. And so I guess are you using in house tools for the LLM or is this with Microsoft as your, I can't remember the exact word, strategic partner

Roy Kirby: or? We we use so we're using different tools because we because we do use Microsoft a lot internally for things like this.

Roy Kirby: Things like teams and also for some of our cloud compute. Yes, we're doing a lot with Microsoft, but we also do a lot with all of the other cloud providers as well. So we have sandbox tested. Tests around Google and AWS as well. So we compare all of the technologies and we'll take the best bits from each one.

Roy Kirby: I think for us, what we've learned over the last year is the importance of your data. And your data set and not putting that out into the public cloud that it's you want that data. You want to make informed decisions on a data set that you own and trust that can be in the cloud, but it needs to be a private cloud with.

Roy Kirby: chat GPT pointed at that private cloud and not made public. I think that's a, that's a big thing for the industry at the moment. Keep control of your data because when you expose that data to an end customer, you want them to know it comes from a quality provider like SICKS that they can make decisions on.

Roy Kirby: Don't mix it with other random data that may come from a source that is not reliable.

Neil Weatherall: And one, one slightly off topic question. Does, does the use of LLMs and AI help you with translate? Is there a need for you to translate your data into other formats or languages or, I mean, are your corporate actions, I guess, cover the whole world?

Neil Weatherall: Is there already a de facto standard that everybody accepts or is there a desire for like local local language?

Roy Kirby: TO be honest we've not used it heavily for local language at the moment. And as you say Even when in finance, it's, it's mainly numbers and acronyms that we're dealing with, right?

Roy Kirby: And most people is a finance language that people understand. So yes, we may use it for a little bit of local language. Certainly for the questions that come in, they should be in a local language. People don't, you know, if we've got a Spanish customer, they should be able to type their request in Spanish.

Roy Kirby: So yes, we will use it for that, but for the data side of things, there's not an awful lot of translation that needs to be done. Although what we have used it for is particularly in, in our, our web API, which we have, which is a, a tool for financial information is in the past things were in. In database world, sort of data based in code tables.

Roy Kirby: So you might have had a code table for currencies, for example. I'm well aware of

Neil Weatherall: these after the last year.

Roy Kirby: You're well aware of that sort of thing, yeah? Actually exposing to somebody GBP or USD and a time zone is much better than transposing to them 7 comma x or something. Yeah. So, um, what we use the tools a lot for is transposing those or exposing that more more easily read.

Roy Kirby: Type of language, the financial language rather than the code tables. The presentation? Yeah, the presentation layer.

Neil Weatherall: Yeah. Okay. I guess we've, we've talked a lot about work. I mean, uh, outside of your job, is there anything that excites you around chat and LLMs and are you tinkering at home with automating your heating based on when you walk in?

Roy Kirby: ? I don't actually. No. And it's, it's interesting in. It's interesting the conversation you had earlier about the, you know, the deep fake stuff. I'm, I'm slightly concerned about some of that consumer use of chat GPT. And if, if it ever really hits my WhatsApp stream, then I will be a little bit upset.

Roy Kirby: So no, I don't play around with it a lot. I tend to we have enough enough of it at work. So eight or nine hours a day technology wise is enough for me

Neil Weatherall: And just for the listener I was talking about, using Using a platform to digitize my voice and then asking it to say things and seeing if my family could tell the difference between my real voice and my, my digitized voice.

Neil Weatherall: And I'm

Roy Kirby: very glad. Yeah. Yeah. Very glad to hear you say they could tell the difference. Yeah. Yeah. Well,

Neil Weatherall: you know, as you said, things are moving very quickly. It might not, might not take too long before, they're

Roy Kirby: indistinguishable. Exactly, and I think that moving quickly thing, and it's why it's good working with I Push Pull, we had, we had the discussion about a chat.

Roy Kirby: Oh, sorry, a box for corporate actions less than a year ago and using a company like yourself. We've got that. It's now released. It's out there in the marketplace. We're getting feedback. But last year at this time, we were not talking about chat GPT, right? It just, some people knew about it, but it really didn't exist on the market.

Roy Kirby: People weren't using it. So in one year you can see the massive difference technology wise. Yeah. Yeah. Yeah. Yeah, and it will move forward and move forward. And we just need to get some of this stuff, as I said, into production. I mean, and I

Neil Weatherall: think that the pace of change is, has just shown us that the way forward is to just make things flexible and allow, certainly for us, it's about allowing our customers a choice of, of what they can use.

Neil Weatherall: And that, that ranges from a choice of. LLM model to use a choice of cloud provider to point to data through to okay. Should you be using ai? You know other more conventional tools and yeah, could you use them sequentially? Or you know, at the same time, you know use them at the same time And compare the outputs And so, you know a little bit of an advert for ipushpull, but we've been concentrating on that Because then you're sort of future proofing Your connections and how you can do things and then as the technology arrives or new ways of doing things.

Neil Weatherall: Hopefully you can just plug them in rather than having to wait to be able to sort of harness the power. We've done a few projects, but we've done many projects over the last year. An interesting one was around voice. So translating voice to chat and then the and then the, you know, the use of chat to infer meaning.

Neil Weatherall: And reading chats to infer meaning is, is very good. The voice stuff had mixed results, but all very interesting around quality of conversation, you know, quality of recordings, accents, speed. I mean, we, we spend some time re recording clips to see if. See if the model handed them better, you know, doing different accents.

Neil Weatherall: It's, it's, it's very interesting. And then when you bring it back to a practical level about how you're going to use it on the desktop, it's also very interesting and it's great, great to play around with to see. See what

Roy Kirby: can happen to your point. You know, you asked me that question about technology outside of the workspace or workplace.

Roy Kirby: And I think, I think what we've seen, particularly in the last year, and this is something, we are trying to take advantage of is because of these technology shifts and because of the way people things do things differently in their private consumer world is having a really positive effect on the corporate world because people are starting to question if I can do this so easily over here, why can't I do it as easily At work in the office for technology departments and people internally in large corporates that's given them a a real headache but it's it really that that consumer type model that we all now have is really driving some change and that's where we want to be part of the change want to be part of the change want to see how can we change Workflows, the way that people do things to make themselves more efficient.

Roy Kirby: I think that is the key thing with these tools. I mentioned it earlier. How do we make things that you have to do day by day? How do you, you have to process corporate actions. You have to give your end users the choice as to whether to reinvest or not reinvest a dividend. How do you make that simpler and easier using these tools?

Roy Kirby: Agreed.

Neil Weatherall: That's a, that's a, sounds like a good place to end. Thank you very much for your time. Thank you for your thoughts. Thank you for your business over the last year. We're excited about what we get up to next collectively. Yeah. And thank you very much.

Roy Kirby: Thank you.