Contextual Intelligence Panel: Removing Digital Distance
Organizations are overwhelmed with data and need an organized & personalized way to receive it. At our latest Innovate conference, Symphony customers and partners discussed how the industry is using technology to read news and content rapidly, and integrating systems to collaborate and make decisions more efficiently. Read more in the transcript below.
Jonathan Christensen: I'm not here to talk, I'm here to hopefully facilitate and make everybody behave. I'd like to start with just having you guys give like a three-minute tour of who you are and your motivation and how it relates to this topic, and we can just start on the end.
Danny Watkins: Sure. Hi, there. I'm Danny Watkins. I'm the CEO at EarlyBird. EarlyBird is, we call it, the safe, smart Twitter app for finance. We set about solving the problems that really held Twitter back in finance to date, and those are first of all a whole lot of regulatory concern about obviously making sure people can't tweet out, but also making sure that Twitter can't be used as a message passing service. You can't leak strategy and that kind of thing through that. We really focused on, first of all, solving the compliance issues with a read-only and fully recorded ability to go and get Twitter. Then having done that, we were then able to build a full-service Twitter application, so that's what it's all about. A Twitter application and then with a bunch of bells and whistles and smart searches and so on, which are specifically designed for the finance industry.
Alexandra Villagra: Okay. Hi, everyone. My name's Alexandra Villagra. I work on the Citi Velocity team. I run North American sales, and I've been very involved in product development globally because I do product as well. For those of you who may not be familiar, Citi Velocity is Citi market's digital platform for all of our intellectual property within the market's business. That means we're delivering research, market commentary, data, and proprietary analytical tools to Citi's institutional client base across the globe. We have over 83,000 unique clients and upwards of 10 million hits a year.
I'm passionate about this topic because I became intensely passionate about the digitalization of the markets business as a line salesperson and understanding really how our clients' needs were not being met in the kind of traditional brick-and-mortar way that we were doing business. And understanding that by digitizing both our intellectual property and the assets that we have within the market's business, we could deliver more efficiently to our clients and solve their problems. That passion and interest eventually grew into Citi Velocity as a platform. I think more than anything, I'm excited about how we can do a better job for our clients through the digitization of the market's business.
Ryan Terpstra: My name is Ryan Terpstra. I'm the founder and CEO of Selerity. If I stumble my words, I have three children in diapers at home right now.
A. Villagra: I feel your pain.
R. Terpstra: I have not gotten a ton of sleep. Some background on Selerity. We're a workflow automation company that really specializes in unstructured data analytics to either increase our customers' productivity or find them new revenue. If you look at the segments that we play in financial services, it ranges from the largest financial institutions, many of whom are here today, to digital wealth platforms like Yahoo Finance, and also large buy-side firms.
One of the compelling use cases that we're demoing today is the ability to intelligently process your private data. Selerity really started by looking at public data, whether that was news, Twitter, or filings. What we found recently, and I think in the past year-and-a-half is, helping our customers harness the power of their private data really has some compelling use cases.
The use case that we're demoing today is the ability to mine digital communications between the sell side and buy side and uncover new fixed income trading opportunities. Cause if you look at a conversation between buy side and sell side that does not result in a trade, very little of that is recorded, so we're helping our customers better understand their clients and find new trading opportunities.
I think my passion for unstructured data really started in my background building things like Thomson Financial News. If you look at this concept of digital distance that David Gurlé talks a lot about, I think one of the biggest culprits of digital distance is unstructured information. If you look at some of the studies, over the next two years enterprises will generate eight times more content that they are today, and a large majority of that is unstructured information. I think in our opinion and our experience, bringing structure to unstructured information empowers a lot of powerful use cases that are relevant to people in the audience.
Matthew Poland: Hi, everyone. My name is Matt Poland. I run research technology at Fidelity Asset Management. Basically in charge of all the application that our analysts use to publish and also consume research. In addition to that, really responsible for making sure that we're bringing in the right type of data, external data in particular that is necessary for our investment professionals to do their jobs.
This is a fascinating topic that we're going to share with you today. When I think of a place like Fidelity, we're very fortunate to have the resources to bring in vast amounts of information, and sometimes that can be more of a negative than a positive thing, believe it or not. Because there's so much out there that if you're an analyst, someone who's coming in relatively new, and you're trying to really ramp up on a certain industry or company, it's hard to even know where to look first. There's just so much information. Certainly looking forward to the discussion, and it should be interesting.
Bijon Mehta: Good afternoon. My name is Bijon Mehta. I look after the financial services vertical for Box. For those of you who are not familiar with Box, we are a cloud content management company. Our focus is helping organizations revolutionize the way they work. We specifically secure content for people, for information, and for applications.
Our world is unstructured content, and so for us, we've been noticing over the last several years an increasing focus within financial services in particular around how do we get more actionable insights from all of the text files, all of the audio files, all of the video files that we have as an organization. With the content sitting in Box, we applied machine learning to that to then allow organizations to better manage the information and then also draw insights from it that either can be used internally or for their customers.
For me, being a financial markets person I'm really excited about this topic. Obviously Wall Street's focused a lot on unstructured data. Obviously they have lots of structured data. What we're seeing now is organizations focusing on a lot of that content that's sitting in files, archived files, and how to draw more insight from that.
JC: Cool. Elon Musk recently said that AI is going to be able to do things better than humans. All things better than humans by 2030. In 12 short years, we don't have to come to work anymore. What do you think about the state of AI and machine learning, and has it held up to its promise? Are you as bullish as Elon? I'll let you guys squabble over the mic.
R. Terpstra: I think there's probably a use case that's pretty interesting that demonstrates the power of AI. In 2015, Google came out and said, "Our AI now filters out 99.9% of spam emails." That means for every email that you receive that's spam, there's 1,000 emails behind it that they've intelligently filtered out.
I think that's a good use case of deep learning in that particular example of eliminating spam, but from our perspective in financial services is much harder especially if you look at algorithmically curating content for a user. Because they're used to the scrolling headlines and getting everything that a particular source will send at them. There's a real fear of missing out in terms of actually being able to get a piece of content.
I would argue that there is so much content that you're already missing. You can never manage that, so you have to move to an algorithmically curated feed but give the user the ability to drill down. I think the Gmail example is an interesting use case of deep learning being applied to spam. I certainly cannot do my job with 1,000 initial emails every day.
A. Villagra: We have a tremendously complex vernacular in the markets business that I think is not quite ready for that level of elegance that Elon Musk is describing. I think that there's some low-hanging fruit in the form of financial objects that can be capitalized on before we get into the full elegance of AI applied to our content.
JC: Say something more about financial objects.
A. Villagra: They need to be defined centrally, better. It's a topic that Ryan and I have spent quite a bit of time talking about. The need as an industry as we digitized to have a more common vernacular so that our bots, our applications, our APIs can make sense of the complicated language that we speak in the market's business. I think there's a need. Obviously, Symphony has been a leader in this space in the establishment of hashtags, but the question becomes adoption obviously. There's a tremendous depth of things that need to be defined and standardized. How can we all be part of that challenge, and what is the best solution for our clients?
M. Poland: I think you look at investment professionals, I think they tend to be skeptics to begin with, especially they're being trained to be good analysts. You look at, again, that issue with, imagine I was just talking to an analyst the other day, and he was trying to explain his biggest fear is missing out on information. It was like walking up to a buffet and the buffet table is the size of a football field. You're looking, and while you see a few things that look like they'd be good to eat, but you don't know what's way down there.
It's just, how can you at least filter? Let's start with the basics. I feel like we're talking about AI. Yeah, we're making some strides in that space, but there's just a lot to go, and I think at the moment it's just about harnessing and really building out some of the basics in terms of filtering down data to a point where you're also gaining that trust of the investment professional. I think that's hard too. You can't just serve something up and say, "This is what's important to you, and just go do your thing." You need to be able to build up almost that trust that the information is credible.
B. Mehta: I think we're in early days in terms of how organizations are really applying, and I hesitate to use the work AI because I think a lot of what we're doing right now is really just applying algorithms, and we're just looking for patterns. To your point about the deluge of information that's coming in, what we're seeing our clients saying to us, we have that problem, but then how do we interrogate that against our historical information that we have?
A lot of work we're doing and exploring with organizations is around things like syndicated lending businesses, general lending businesses, where you want to look at where different triggers were in various contractual addendums that's already sitting in some archived file somewhere, and you've got a deluge of business that may be coming in from a certain sector of an industry. You just want to compare and contrast different types of terms.
We're seeing it applied in capital markets from a research perspective, and again, creating APIs off some of the information that's sitting in documents. Right now I still think it's early days because no one's creating it in an industrial manner. It's a lot of experiments going on.
R. Terpstra: I think the financial objects discussion is really important when you talk about intelligently filtering information. I'll give you an example, #COP, what does that mean?
A. Villagra: Colombian peso.
R. Terpstra: Does that mean Colombian peso? Coro Mining? ConocoPhillips?
A. Villagra: Yeah.
R. Terpstra: It could be a range of things. That is a flat tag. That is not a concept-based tag, and I think financial objects has the opportunity to divide things at the concept level. One of the digital wealth platforms that we're integrated with before integrating our service, you'd go to the quotes page in Apple for their US listing, and then you'd see robust content. You'd go to the London Stock Exchange listing, and there's nothing because they modeled all that content at the instrument level. But when Apple releases the iPhone, that happens at Apple at the concept level, not an instrument level. Financial objects, I think has the ability to get those things right. It is required to be able to intelligently filter content.
D. Watkins: The other thing, back on the AI that's slightly concerning is the danger that too much AI, too much automated curating, too much of that takes away the individual trader or analyst's ability to do his job in his way with his unique spin on the information that we're getting. There's one thing that traders hate more than anything else, and it's being spoon-fed exactly the same information as the guy next door and being told what to follow.
Our first big customer said to us, "We've got the best analysts in the world. We want them to decide what information they have, not some machine." In the context of the conversation we're having about contextual intelligence, the context of who that end user is, is important as anything else. That two guys doing the same job are getting that same information will interpret it very differently and want a very different kind of information or see something as important compared to somebody else. Too much of that AI, too much of that will tell you what you're going to see is not necessarily great for people's ability to have a competitive advantage.
JC: Is personalization a big part of all of your strategies?
M. Poland: Yeah.
D. Watkins: Very much. Very much for us. Yeah, it's all about allowing that user to tune his feed. Whether in it's our case specifically Twitter accounts that he wants to bring in and the ability to follow absolutely anybody, not a curated list, or in the terms of the smart searches to be able to tune those and say that a particular tweet or whatever that we may have scored in a particular way is not what he wants to see, so he's going mark this one down. Very much that personalization. Then being able to share that with the rest of his team, so you build up a collective intelligence as well.
M. Poland: I think the key is then taking it to that next level. You have the personalization. There's definitely a need for that to be able to customize how someone wants to see the information, and is there a way really to then be able to say, okay, based on that and based on the usage patterns that that person's employing, can we start to maybe start feeding recommendations? Recommended content base on that, and see how successful it is.
I like the idea where you can have the actual user put a thumbs up thumbs down on that recommendation to feed more information into the process. I think it's helpful, and it allows them to participate. There's definitely potential there.
A. Villagra: I think a practical example of how we're trying to make that come to life through Symphony and through Velocity is a market commentary bot. It allows you to receive your own custom personalized feed in a Symphony chat room of the market commentary that we have available on Velocity. It removes the challenge of the vast amount of content that you have in front of you and gives you a very specific personalized feed. I think there's a hundred examples of how with digital assets you can personalize content to the end users' needs.
B. Mehta: I think this feeds into this broader narrative around how are people leveraging technology to do their job more efficiently. Having spent the early part of my career on a trading floor, there's just so much information you've got to digest and then turn it into something that you can then go up back out to clients with. How are you consuming all of that? How can you more efficiently leverage your time?
What I see now with various tools is we're almost automating the decision stack of people. You come in in the morning. You've got Bloomberg. You've got internal commentary. You've got your research site. You've got Symphony. You've got a variety of other mechanisms through which information is coming from. You want to curate. You want to have it centralized, and then over time you want that information to then feed to you, like some of the panels we're talking about earlier this morning, things that you should then act on.
M. Poland: I think the reality too, and we haven't really talked about it, is this all assumes that the data is easily accessible, and it's accurate, and it's all there. From my perspective, we spend 90% of our time just trying to get the data and making sure that the data quality is there and you can access it and you can apply whatever type of algorithms you want on top of that, but if you don't have that, if you don't have it sourced correctly in the right place. If you don't have a solid, I think this morning someone mentioned having a good data science platform to be able to work on and to be able to throw data into, it becomes very difficult to succeed.
R. Terpstra: I think what we found with personalization in this industry is that you have to give the user some control of the parameters that you're using. Our platform looks at source credibility. It looks at the degree of relevance of a piece of content to a specific financial concept. It also looks at virality, so what's actually trending right now to give us an idea of what's important.
I think when we first came out with Selerity Context, we had sent those parameters in a way that we saw fit, but then you would speak to traders and say, "Where's the Reuters story?" "It's the same story as the Business Insider story." "No, I need to see the Reuters story."
Then you end up giving them the parameters to adjust that, where they're tolerant of more duplication, and we give those parameters. But on the other extreme, you have LinkedIn and Facebook, where you have little flexibility or control of those underlying parameters. There's definitely two schools of thought, but this industry is certainly not ready for the full, I think, Facebook or LinkedIn model. You need to give that user some control.
B. Mehta: I think you touch up on like self-service. The ability to modulate how you want to consume what you want to consume, and I think that's becoming. We see that in our personal lives when we interact with different technologies, and I think that that bar and that expectation exists now in the banking world and in the financial markets world.
JC: There's got to be a generational aspect to this as well, right? I'm a piler, and some people are filers. I expect the system to go find the things that I need for me. David, my boss, is a filer, extreme filer. Nested folders, dates, and subjects, and so on and so forth, and he likes to work that way. But I think generationally, like with my kids for example, who are coming into the workforce right now, they're all pilers too, and they expect the tools to bring them insights.
R. Terpstra: I think the number one requested feature of our product sets is personalized push notifications. Don't make me go and have to find something important. Push me something that's important. Technically, that is very hard. Interrupting someone's workflow, especially in this industry is a very dangerous thing. You better be very confident at what you're pushing at them is important, but I would say that is the number one requested feature, but technically it's very hard. Companies that have raised hundreds of millions of dollars have still had to hire human beings to do that final layer of curation.
B. Mehta: Yeah, I think we solve several problems as a company, but I think if there's one common problem that people have is when you're trying to find something, you can spend an inordinate amount of time doing that. Whether it's nested in some email as an attachment, or it's sitting in some shared drive. Then you've got to figure out, am I looking at the right version of that? Is it the latest version?
What we fundamentally do as a company is we remove email attachments, so everything is now in a way that you can identify it, and it's one single source of truth to that information. That helps that process out.
M. Poland: I think just talking about everybody wants things pushed to them, but in those situations where they're looking for data, you need a good quality search capability. I think people assume search is ubiquitous and a good search is always available. That's not the case. It takes a lot of work to be able to search really well and make sure all your data's indexed properly, and when you do that you can save. I was just talking to someone about how we had a quant that was looking to search for a particular topic, and they were literally working for like a week or something like that trying to search through all this data. We were building a more intelligent search capability, and we have a POC right now. We were able to pull up that data in like three seconds.
Everybody knows what it's like when you have a search capability that doesn't work well, and what's the first thing you do? You stop using it. That's really important.
B. Mehta: It's amazing how many PhDs come into Wall Street only to find that they spend most of their time cleaning data. As opposed to interrogating it for clues and signals.
R. Terpstra: That's actually a very good point. When people talk about artificial intelligence, they get very excited about deep learning and neural networks and adaptive learning and all these new technologies. But no one really talks about the data required to train these algorithms. For our product that analyzes digital communications, ie, chat, we went to our customers and said, "Hey, give us your chat. We want to train our machine learning and identify bonds," and we were almost laughed at. You can't do that.
We said, "This isn't good." We tried to buy defunct hedge funds. That didn't work, and we had to hire former bond salespeople to actually write synthetic chat. It's very challenging in this industry. It's not like the consumer world, where you have access to massive amounts of information or millions and millions of users giving you regular feedback. I think that's one of the big challenges when you're building AI and NLP in this industry that many times gets overlooked.
A. Villagra: Yeah, I think you have to have both. Based on hundreds of conversations with clients over the years, they without question have information anxiety. You can't go into them and tell them that you've curated the content for them. You can't. They want to have both the option of everything and the ability to choose, and the ability to deliver that choice really effectively, be it by notification or really outstanding search, I think is a critical challenge that we're all looking at.
B. Mehta: Are customers asking you for API feeds?
A. Villagra: Yes.
JC: Yeah, as soon as you get started they want analytics. They want raw data, and they want to be able to manipulate it themselves. Of course with Symphony, we're blind anyway, so we don't see anything. A search has been a long time issue with Symphony. We're finally moving to client side so we can do indexes locally to provide better search, but when you're blind to the data stream, it's a problem. Can you guys share maybe a success story? Or if you want, something that went horribly wrong?
A. Villagra: Okay, you guys do the horribly wrong. I'll do the success story.
D. Watkins: Something that went horribly wrong, but which I'd use as an example for why you need, in our case obviously we're focused on Twitter, some very strong reasons why you need to have control over that. Everybody knows that you can't go around tweeting and that kind of thing, but certain people have got that capability. There's a chap in the San Francisco Fed who had access to their main account, and it was his job to send stuff out.
But when you're on a mobile on Twitter, you can be logged into several accounts at once. He thought he was on his personal account. He managed to send an incredibly derogatory tweet about Trump in the run-up to the election on the San Francisco Fed account. In terms of making sure that there are controls of that kind of usage in our particular vertical, there's some astonishing examples of misuse in that case.
JC: I still want to know what “covfefe” is.
A. Villagra: We all do. I think a recent example that came up that I really like was a client of ours who has an incoming class of juniors, of analysts that they weren't prepared to roll out Bloomberg to. They were looking to deliver live market monitors quickly and efficiently to a large group of incoming juniors and came to us and said, "Hey, Velocity, you guys have tons of live market monitors, live market data. We've got to get this product out to a large group of analysts as quickly as possible. Can you partner with us?"
Obviously having Symphony in the middle of that and instant distribution, combined with the many years of building up live market monitors and OTC products are difficult to come by was a really wonderful example of how we could work collaboratively to deliver a quick, effective solution to our clients.
R. Terpstra: I think when we started building what we call PCE, our Private Context Engine, there was real fear on my part in the beginning because we were very proud of our first version. We went to a trading desk and a trader had put some things in, just “GM.” “Good morning. General Motors.” He's like, "This is terrible."
We had to really iterate, but every time we got more data, the accuracy really improved. It was just last week we had demoed to a prospect that we engaged with six months ago, and he took some of his chat. Obviously changed the names and changed the instruments, and we were able to identify not only the instrument, but it was that client making an inquiry, and what did I quote them? And did the trade get done?
Being able to pull out not only an instrument, but the actual trade workflow, he was incredibly impressed. We've grown leaps and bounds from that first instance of “GM,” which is not “General Motors.” It's “good morning.”
M. Poland: I think as a technologist you go to an AI conference, and you start thinking, especially if you're in the financial services industry, and you're like, "You know what? This is great. I'm going to figure out some way to run some algorithms." Basically, you're going to mimic what a portfolio manager does, and we'll all make a ton of money.
I think it's easy to, and maybe it's something that's aspirational, but it's something that you really can get a little carried away focusing that far down the line. What we found that has been hugely successful is really focusing on some of the productivity savings that you get from just being able to bring together the data in one place. Be able to search accurately, be able to use tools like Symphony to not only collaborate, but also pull up the latest information quickly.
That type of stuff resonates incredibly well with the business, and that's what it's all about. Investment professionals are tired. They're working hard. They're overwhelmed with data, like how can we make their life easier? And if we stick to that principle, things tend to succeed.
B. Mehta: I'll share a story about a success in the making. We've been getting a lot of inquiries from the asset management community about how can we take some of the research reports and then create API feeds to then go into our models? The idea was let's put all the research content in Box. It's in one spot. We've created algorithms with a partner that can then mine the content, create APIs, and then deliver that for models with various analysts on the buy side. That's one.
The other one I'll just quickly highlight is our integration with Symphony. Our hope is that what that does is eliminates all the email attachments that otherwise would have to be sent around. They'll be just Box links, and you'll benefit from our security posture and the work that we've been doing in that space.
JC: Right, super complementary in terms of that, the security posture in particular. Cool.
D. Watkins: Having thrown out a failure store, not my failure story, but the San Francisco, I'll try a more positive one. One of the features that we introduced very recently, I alluded to just now, is this idea that the collective intelligence of the team of people is greater than the individual intelligence.
We introduced a feature where one person can, our system recognized Chinese...the team within a Chinese wall can say somebody can build up and customize a set of searches and so on, and then publish it and make it available to the other people. When we introduced this, we thought what would happen as we went into a particular organization was they would find the young intern or something that was into Twitter and say, "Can you customize a feed for us, and we'll give it to everybody as a starting point?"
That didn't happen. What actually happened is a Global Head of Strategy for oil or whatever it was, he was very influential on Twitter, a very key user. Then you find that what would happen is he would set up a really interesting feed, and then as we go to the next person, they would say, "Whatever Harry's got, that's what we want." We were staggered at how we thought that you would imagine Twitter being an extremely generational thing, where the young people are into it, and the older are not, but actually some of these very senior people are very active and can construct very interest feeds that are delivering real value across the organization.
JC: Music services have leveraged this, the taste-makers. People will follow them, and you know that they're going to be consistent. It's interesting because, I don't know if you remember early on, Netflix had a social network built into it, and it was supposed to predict and give you recommendations for movies. It was a horrible failure, so the model's super important. The people you know and the movies they watch. Cool.
JC: Why don't we let the audience ask some questions? Yeah?
Audience Member: You heard a bunch about Google, and Gmail. I was thinking of using, if it's possible, the question is, to use the same approach that Google uses for searching the internet, using and deploying to apply to people and the analogy is the following: twenty percent of people produce eight percent of content. We use this model to somehow run people on the organization on topics and knows how people would find interesting for their work. Instead they have access to content, they'll have the potential access to people who are knowledgeable about particular subject, trading or compliance, legal, scientific methods.
R. Terpstra: Yeah, we've actually thought a lot about that, which is how do you automatically build an interest graph for users in the enterprise? That could be based on what they search, based on what they write, based on what they told you they're interested in. What's really important is understanding things, again, at a concept level.
When you dig into oil, there's lots of different types of oil, and you want to know who's a subject matter expert maybe for a particular petroleum. But this concept of using user activity to build interest factors is something we've spent a lot of time on. It all goes back to access to data, quite frankly. Because if you look at neural networks and deep learning today, they do require a lot of data.
Now, there's other technologies coming down the pipe, and our CTO has a PhD in machine learning. More than happy to make that introduction to talk about those technologies, but absolutely. Being able to use people's behavior to understand their interests and then infer if they're a potential knowledge expert is something we've actually thought about.
Audience Member: That is being done, or?
R. Terpstra: Slack is trying to it. I know that's a dirty word. I apologize. Slack is trying to understand, and that's one of the issues that people in financial services had. We talked a lot about encryption and some of the things that a platform like Slack does not have, but yes, they're trying to understand if you have a question, who's the subject matter experts in your organization. But what makes that really tricky from our industry's perspective is you need to then read everyone's communications. That could get very tricky from an information security perspective.
D. Watkins: We do that, exactly that specifically on the Twittersphere in that we'll take each tweet. Part of that analysis of is this financially relevant is, not only is the tweet content financially relevant, but the user behind it, what does he typically tweet about? Does he typically tweet about finance and politics? The people following him, what do they tweet about? Or do they follow other people in that interest?
We do part of that. The big interest, the big reputational draft, if you like, of Twitter users, and that's very much one of the key factors that's used to say this tweet is probably more relevant than this one because although it says the same as that one, this guy knows his onions.
JC: I can't mention who, but we do have a customer who's building a bot that is a reputation. Who's the best person, kind of bot for a topic. It's all natural language. You just type in a question, and it comes back with some candidates and links directly to them so that you can connect to them and talk to them right away. Pretty interesting application. Other questions? Right here?
A. Villagra: No. The FINOS, the open source foundation for FinTech companies, has a working group on financial objects, whose task is precisely to define standards. But I think it's being thought of in many different ways, and that's one of the problems. Is a financial object a tag? Is a financial object an RFQ? So many people have so many different definitions of what a financial object is that it has been difficult to find consensus, but I think we have the right forum to do it now.
JC: It feels like its really core problem, at the core of the inefficiency, counterparty interactions. You guys have got to sort this out.
R. Terpstra: The biggest issue, quite frankly, is one around commercials. The people that have solved this are some of the big market data providers. They know what they have, and they like to charge because they have thousands of people helping them curate and manage that. I think that is one of the issues because it takes an immense amount of work.
We cover a small section of financial services. Yes, we have 500,000 instruments, all the central banks, all the currencies, 30,000 public companies, but that's a team of 12 people working four years on nothing but that. I think that's one of the issues is maintaining.
A. Villagra: Yeah, I think the industry needs to make some decisions around prioritization of what objects they want to pursue in what order. I think it's also an issue of a marketplace discoverability. There are plenty of organizations, to Ryan's point, who have already defined objects and standards. They're just not very discoverable, so I think that's one of the issues that FINOS will help solve.
JC: Any others? We've got four minutes. Yeah?
Audience Member: In addition to the financial objects problem, is the next step which is people use all kinds of nomenclature to describe this. It's not just ... It's sterling, cable, pound. Which of these different things, terms, means those things? It's not just even once you've decided this is the standard, do people want to use the standard hashtag, or are they just going to type whatever they think?
R. Terpstra: People are going to type whatever they think.
A. Villagra: Yeah.
R. Terpstra: Changing people's workflows in this industry is incredibly hard. What we've tried to do with our products is just integrate with their existing workflow and understand the lingo specific to rates or corporate bonds because trying to get users to adopt a standard and enforce that is a loser's game. That's very difficult.
D. Watkins: It's bad enough internally, but of course if as we both are pulling in also sorts of disparate information from external sources they're just not going to play nice. If the Guardian's reporting a scoop about Apple doing a tax deal with the Italian authorities, they're not going to helpfully write $AAPL in the text. They're going to be talking about Apple, and that's the case in any of these things. It's our job unfortunately to try to extract what it is that this person is talking about.
JC: Cool. Elon made an audacious prediction. Any audacious predictions or last insights before we wrap it up?
R. Terpstra: Google buys Tesla.
JC: Google buys Tesla? No disclaimers.
R. Terpstra: Elon needs an Eric Schmidt right now, I think.
JC: All right. Thank you for joining us today, being here with us.