Hi everyone. Welcome to this edition of HFS Unfiltered. I am Elena Christopher, the Chief Research Officer and Head of Financial Services Research at HFS. And we have a fantastic topic for today. We're going to be talking about hype busting Gen AI, you know, generative AI, aka chat GPT, in financial services. I think we can all agree that Gen AI is gigantic rise. It's captivated consumers and businesses alike with its seeming quick win benefits and maybe its still untapped potential. So we're going to talk about this today, about how Gen AI can be safely and effectively leveraged in the financial services context for value and maybe even differentiation. I am joined today by a couple of fantastic folks. First, I've got Gil Perez, who is the Chief Innovation Officer at Deutsche Bank. Hello, Gil.
Hi there.
We also have Sean O'Donnell, who is the CTO for Financial Services International at Publisys Sapient. Sean also has responsibilities for a range of emerging technologies such as cloud and AI. But welcome, Sean.
Great to be here. Thanks, Elena.
Before we go deep on Gen AI, I'd like to level set a little bit with Deutsche Bank's innovation history, partially because you guys have been, and I love when firms do this, where you are very public about some of the things that you're trying to do in the way of modernization. Like, for example, you've been open about some of your investments in innovation and cloud in an effort to impact your cost to income ratio. Could you maybe, and I think this will set us up for that NBT, as you've been working on with generative AI, to help us understand a little bit about where Deutsche Bank has been with some of your innovation to set the stage with where you are and where you're going at this point?
When I joined, I joined Deutsche Bank in 2019. And at the time, we did not use the public cloud. We had what we called private cloud, which in essence was running our own data centers, but outsourcing them. And what has happened is over the years before that, up to that point, a lot of the technology and the innovation of firms around the world, of the industry, has gone into the cloud, while not as much as has been done on what we call the on-premise world. And also talent. If you look at schools and all the new technologies, capabilities, everything is now shifting more to the cloud. So, in essence, the first thing that we had to do was to create a platform, a basis for innovation. And really, the cloud is that. And if you think of even generative AI, if we didn't have the compute power of the cloud, if we weren't able to tap in and out consumption into compute, it would be almost impossible to leverage and to use AI and generative AI in a way that we're doing right now. So, really, we had a couple of years of actually deciding that we're going on this cloud journey, creating the right frameworks and controls to use it responsibly, prudently, within the constraints and the guidelines of regulatory. And you can imagine, as I mentioned before, we're working with 46 different countries, 46 different regulators. There's a lot of stuff that needs to be done, an explanation and approvals that need to be done before we could have had our first production workload on the cloud. And that basically took about a year. So, throughout 2020, sorry, 2021, we really laid those foundations and then we started moving the workloads on it. And in parallel, we started working on generative AI as the next thing.
I think, look, it's a very rapidly evolving space. I think what we've seen, and we've obviously seen this gilled together both in terms of Deutsche Bank, but obviously looking outside, is that I think it's gone through mini waves even in terms of Gen AI, where the initial wave was a little bit of discovery in terms of, OK, well, really, what can this technology do and how can it be applied? And again, I think that's kind of a general pattern that we've seen with other clients through to kind of both getting the business on board, but also kind of unpacking use cases, right? Because I think people have seen the power of what they can do. And that's where great things like the access to technology like ChatGPT has really opened that up to an audience that we couldn't have dreamed of, I think, beforehand. But now, I think we've gone past that. Now it's a case in terms of productionizing. So for us, what we're seeing broadly across the board is, yeah, we get all the use cases. We have many organizations helping us and we kind of know that ourselves. What can you do in terms of saying to actually put the right guardrails in place so that from a CRO perspective or a legal perspective or an audit or compliance, and frankly, even from a regulator, that we can kind of evident that we're pushing the envelope in terms of technology, but we're also kind of playing safe. Because, you know, at the end of the day, you know, lots of use cases, yes, can be used in back office and they can be used in terms of replacements for RPA, streamlining kind of, you know, the flow of work. But ultimately, the real power comes if you can put it in colleague or customer's hands. And that's really then where you, you know, you rub up against the ethics questions, the, you know, is this safe? Are people doing what they should be doing with my data?
What I'd love to do briefly, Gil, is because we've been talking broadly about Gen AI, I'd love to, if you're game for it, talk a bit more specifically about how Deutsche Bank is leveraging Gen AI. Because one thing that was mentioned both in Sean's comments and in your comments, Gil, there's a lot of use cases out there. And with any new exciting tool, you run the risk of it's that hammer looking for a nail type of scenario. Oh, I have this cool new thing. What the heck can I do with it? Now, mind you, you have to go through that cycle to figure out what it is, what it's good for, what the risks are, what guardrails you need. But you've been at this for coming up on a couple of years now. What's really passing muster? What, when you sort of go from use case and a good idea to what's really yielding value? Do you have anything you can share with us?
Yeah, sure. I think I mentioned it, but I'll give three examples. The first one is just software development. Software development is going to be to transform. It's not so much as the generative aspect of creating new code. It's actually documentation also. It's also understanding old code. We still have, and we're not the only ones that have a lot of legacy code, for example, COBOL, a language that was used in the past. You have new people coming in having no idea how to fix or to deal with that code or maybe replicate it. Generative AI tool can easily go through the code and actually explain what it's doing and even suggest a way to migrate it or rewrite it in a different. There is generative AI capability, but there's also a very close human in the loop in that entire process. Documentation, it's not something that people like to do. It's required. With a single click, you can get a documentation also. It's important for regulatory reasons. It definitely will improve the efficiency of our developers. All of that space is really, really exciting. The second thing, I would call it the chatbots, the various chatbots. That could be from interactions where we're seeing right now. This entire conversation could be transcribed, summarized with action items. There's going to be a lot of interesting capabilities around chatbots, around just transcription, translation into multiple languages, but not only into multiple languages, but include the different nuances of every business and their specific additional terms and elements that they used in their conversation. Being able to then take a body of documents and being able to inquire it, to summarize it quickly, again, with a human in the loop is going to be extremely important, and we're seeing a huge amount of that, which falls into the banking area and what we call the research area. The research area is not just researching a market or a specific stock. It is even a salesperson or somebody, anybody that is researching the company will use the same kind of tools. I think that large language models and the body of work and the chatbot will be initially being used as kind of an advisor, somebody that you could use in order to accelerate your, refine and accelerate your output. That, over time, will get more and more comfortable with that. The last but not least, I would say that we're seeing also the use of large language models with anti-money laundering and trying to figure out how do we also use all of the data that we have in order to improve the resiliency and the compliance of the bank. I just think that that requires the regulatory on board, and that will take a bit more time, but those would be my three topics. Again, software development, the chatbot interactions, and obviously regulatory.
And then the other piece of it is, and this is again, is more in the kind of revenue generating areas of streams, right? So rather than pure cost outplay, so now what can be done in terms of, you know, intelligence around pricing and product and service offering? And frankly, how can you use that technology to be more, even more insightful and more personalized around, you know, dynamic generation of that? We're, you know, kind of, you know, used to the kind of physical world, the lock-in in terms of the price at the point in time is what it is, or the product at the point in time is what it is. And frankly, you know, financial services has been guiltyついに。