1
00:00:00,000 --> 00:02:39,279
SPEAKER_01: Hello and welcome to the Data Democratization Podcast. I'm Alexandra Ebert, your host and mostly AI's Chief Trust Officer. Today's our second episode of our brand new season. And similar to the first one, it was recorded live at Money2020 US approximately two weeks ago. And my guest for today's show is Daragh Morrissey, Microsoft's Director of AI for Worldwide Financial Services. And given Daragh's role, he has this unique insights into how financial services institutions across the globe are proceeding on the AI journey, the learnings they get from that and what makes out success factors. So we talked a lot about the patterns that he observes there. And I was, for example, surprised to learn why particularly Canada and Australia are two regions worth watching when it comes to rapid AI adoption. Of course, we also talked about Gen AI in financial services, and Daragh had lots of actionable advice on how to make this work and how to succeed with it. So for example, it's important to not over strategize and to focus on the outcome rather than on perfection. We also dove into the different AI use cases that see the most traction and that Daragh considers to have the highest potential for the financial services institutions Microsoft works with. And of course, we also talked about the importance of responsible AI and why it's essential to have it codified to make sure that this feeds into any AI system that organizations deploy and develop. Next up, we talked about the challenges that financial services organizations have to overcome when it comes to them getting AI ready and moving towards the cloud. And Daragh also shares his excitement for the future, not only in terms of AI innovation, but also regarding other emerging technologies like quantum computing. And Daragh is a fantastic storyteller and is full of great anecdotes. So I'm sure you will enjoy today's episode and will take a lot away from it. So without further ado, let's dive right in. Welcome to the Data Democratization Podcast, Daragh. It's a pleasure to have you here and record an episode live at Money2020. And we have a lot to talk about today. But before we dive in, could I ask you to briefly introduce yourself and maybe also share with our listeners what makes you so passionate about the work that you do?
2
00:02:39,279 --> 00:02:46,360
SPEAKER_02: Sure thing, Alexandra. It's fantastic to be here. And I have to warn you that I love talking and I love podcasts. I listen to them every day.
3
00:02:46,360 --> 00:02:47,639
SPEAKER_01: You sound like the perfect guest.
4
00:02:47,639 --> 00:03:59,639
SPEAKER_02: We might run over. I hope not. I think they will kick us out. They'll kick us out. So, yeah, I work in Microsoft Worldwide Financial Services. The really cool thing about my job is I get to see what all of my customers are doing all across the world. I envy you. I'm very, very lucky. And, you know, I work across banking and insurance capital markets. I don't know how I ended up with this job either. You know, I'm eight years with Microsoft in Redmond, Seattle. Before that, I worked in the banks in Ireland. And I was very lucky. I was very lucky to get this role. And the great thing is I do executive briefings too in Redmond. So, customers fly in to meet me every week. So, I get to present to them, learn from them. And I think in this, what excites me so much at the moment is not the power of this technology. It's the things it's doing to bring people together. I actually kind of said it on the mention on the panel yesterday. It's bringing business and technology together. It's bringing all kinds of disciplines together. It's just an incredibly amazing time. So, I'm very lucky to be right at the center of all this stuff in financial services.
5
00:03:59,639 --> 00:04:42,640
SPEAKER_01: I can totally get it. And I'm wondering if Microsoft is considering opening up a second position to support you a little bit. No worries. But fantastic that you get all of these insights. And this is also the reason why I invited you to the podcast because our listeners are always curious how they can succeed with AI. What the lessons learned are. And somebody at your position obviously has insights that many, many people crave. So, let's maybe start off right with that. You work in a global capacity. You see so many different financial service providers. What are the patterns that you observe that really separate those succeeding with AI, getting a benefit, getting value out of it, versus those who are more in the, okay, we're trying and playing around with the technology, but we fail to connect it to real business outcomes? Sure.
6
00:04:42,640 --> 00:06:37,640
SPEAKER_02: I think the first thing I'd say is there's really two types of AI. So, predictive AI, machine learning AI has been in the industry for years. And I think banks are doing a great job with that. It's extremely efficient at what it does. There's actually a lot of money still being spent on it despite everything you hear about generative AI. And I was working on that too. I started working in the predictive AI space in around 2018. I moved from working on blockchain and things like that to AI. And honestly, with machine learning AI, I felt I was pushing a rock uphill a bit. The benefits of those use cases and patterns was very specific to different parts of the bank. So, what I love about generative AI is there's general purpose use cases that run across the whole bank. So, going back to the patterns of what's working, I think the key thing is there's a lot of things that are making this easier to get into the hands of the end user. The really successful organizations, what they've done is look across three dimensions, the technology, people, and then thinking about how it really moves the needle for them. There's a lot of companies who are kind of over-strategizing and over-testing as well, which I don't think is the right thing to do. I think I've one customer at the moment and they said we've a 96% accuracy in the answers that we're getting out of gen AI. And they weren't happy with that. My kid got 96% in an exam. I would buy her a new iPad or something. So, the thing is the technology is moving so quickly that even anything you measure today in a month, that measurement won't be any value. So, instead you should be focusing on the outcome and not perfection. You know, you could spend an awful lot of money to get to 100%.
7
00:06:37,640 --> 00:06:59,640
SPEAKER_01: We also know about this 80-20 principle. So, I think this also highlights the purpose. Okay, so kind of not waiting too long, not getting everything to 100% is one of the advices that you would give our listeners. Is it also that it should be more an experimental nature and that you learn by doing versus thinking too long, strategizing too long and not getting moving?
8
00:06:59,640 --> 00:08:20,640
SPEAKER_02: Yeah, I think banks were incredibly smart about the use cases that they picked initially. You know, they all started internally. They picked use cases that didn't have much regulatory baggage. Or, you know, if something goes wrong, they won't be on the news. So, they all started internally. You know, a lot of people started building knowledge bases. They started adding some to the contact center, you know, very basic stuff like transcription. And, you know, what's great about these use cases is that they were able to learn, you know, build trust in the technology. There's a lot that we do in Microsoft and our platform to do that too. So, I think that those are great steps. The problem with some of these use cases, though, the safer it is, the less impact it might have from an ROI perspective. So, building an HR bot or an IT bot, it might not make that much difference. So, one of the things we did at Microsoft is we put Gen AI in the hands of our sellers. They were the first to get it. It went off as co-pilot. We're also building more tools there. And, you know, your sellers are kind of, or your advisors are your relationship roles in a bank. They're the ones that are, you know, have the relationship with the customer. If you can start selling or upselling or cross-selling more instead of doing low, not low-value stuff, but basic servicing, that's where we're seeing impact in the bottom line.
9
00:08:20,640 --> 00:08:46,640
SPEAKER_01: Makes sense. Makes sense. Would you also say that now that we have Gen AI tools that it changed the picture of when is it too early for an organization to start embracing AI? Because what I've observed over the past few years was organization getting more and more interested, sometimes also pressured to start with AI. But if it was an organization that in their data maturity, data quality, data governance just wasn't there yet, it didn't really make sense. Does this change the picture now that we have Gen AI tools?
10
00:08:46,640 --> 00:09:38,640
SPEAKER_02: I think so. You do need great data to do this. But I wouldn't, again, try not strive for perfection. Don't wait for your data to be perfect. I think that's a bit of a unicorn, if I'm honest. I think what you should be doing is the moment that you're not actually doing anything, you're not learning anything. It's a bit like learning to drive. My dad took me out to the supermarket, brought me up and down very slowly. And then when I taught my kid to drive, then after about two weeks, she wants to go on the highway. And she did a great job. But you can't learn to drive in a book. And it's the same with this. Your organization is going to have to build up a muscle around this. And the more that you can't wait for other people to do it perfectly and then copy, you need to build up the muscle inside the organization.
11
00:09:38,640 --> 00:10:07,640
SPEAKER_01: Exactly. I'm also wondering, since you mentioned all the executives that come to visit you and learn from the best practices that you observe within other companies, we know that there is this paradigm shift between traditional software development and AI projects, also in terms of the experimental nature that it's hard to tell how long will it exactly take to get to a specific result. What is your observation how executives are faring with this paradigm shift and that it's challenging or it's a different way of approaching development, approaching new projects?
12
00:10:07,640 --> 00:11:38,640
SPEAKER_02: Yeah, I think the way a lot of my banks do this successfully is bringing data to their executives on the impact of a use case. So with developers, for example, we had an Australian bank who did a case study. They actually published the results of their testing. So they had two sets of developers, and they gave one a set of projects to do, and then they gave them Gen AI to the developers, GitHub to the developers. And they measured the productivity. They measured the quality of the code coming out. It was all better. And the developer satisfaction in the developers was higher. Even we see in the adoption patterns, we see kind of unusual things. We see a U-shape sometimes with that use case. What's funny is older developers, it sort of starts high, comes down a bit, and then it goes back up again. And older developers are the ones that really see the value. Maybe they're doing sort of bigger projects, and they want to really be more strategic with their time than a junior developer, but that use case is evolving as well all the time. Now you can do Gen AI to do unit testing. It's great at looking at old code as well. I think one of the magical things it can do is look at a really bad piece of COBOL and document it. You don't even need to change it. It will just tell you what the code is doing.
13
00:11:38,640 --> 00:11:41,640
SPEAKER_01: I can imagine that this is something financial services organizations appreciate.
14
00:11:41,640 --> 00:12:14,640
SPEAKER_02: I know, because there's a lot of old stuff there. Some of it isn't documented particularly well, and the skills are leaving the industry. Even at Microsoft, we're using it to modernize older bits of Windows code. Windows was written in C++, and it's got some security issues. We're taking that to a new language called Rust, and we're converting that at scale using this technology. There's a ton of testing going on. Developers are not going to disappear. They're just going to become more like developer managers, basically. I think that's really exciting.
15
00:12:14,640 --> 00:12:36,640
SPEAKER_01: Exactly. You still need the expertise, which also brings up this question how to get to the expertise, particularly as a junior who starts working out with these tools. What's your take on that? If you start out with AI technology helping you along the way, how will you get to that experience that current senior developers have to really scrutinize the results that they're getting?
16
00:12:36,640 --> 00:13:20,640
SPEAKER_02: I think the thing is with any of these rollouts of technology, the important thing is to train people. One of the things we do with Microsoft was when we rolled out Office Copilot, we did it in waves. One month I got Teams recording, and then the next month I got PowerPoint Copilot. That meant that it was more digestible for the audience. We could train in a scalable way. You didn't take two days out of your job to do everything because there's so much power here. If you can figure that out, that kind of plan, I think that would be great. Breaking it up and really making it digestible.
17
00:13:20,640 --> 00:13:51,640
SPEAKER_01: I think that's the important lesson also here that you need not only to think of how to get the theoretical knowledge within your company, but how to actually foster adoption and breaking it up and giving them one tool at a time as opposed to saying, okay, here's your Swiss Army knife of everything you ever wanted to do. Now figure out how to integrate it. I think it's not the right way to approach this. Coming back to the initial question of the patterns that you observe in those that succeed with AI, given that you have a global role, are there also some differences for you that stand out when you look at American banks, European banks, Asian banks, for example?
18
00:13:51,640 --> 00:15:36,640
SPEAKER_02: Yeah, there's different patterns. I think I do see incredible innovation in Asia. I see some of the first production use cases were there. We have a bank in Singapore. They had it in production in September, October 2023 in their contact center. What they did is link up their contact center to a knowledge base, so you learn from every conversation, which I think is really cool. I think there's also kind of differences in the markets. I think Canada and Australia are really interesting to me. I love my Canadian customers and my US customers as well, I should say. The thing I love about them is there's four banks in each of those markets competing with each other, and that drives this constructive tension to innovate. It's not quite there in the US. What I see, though, in the big US banks is incredible technology capability. Some of the banks have more developers than Microsoft, and they love building on-prem. But as you get down into tier two, a lot of the banks don't have the same capability. But the great thing about Gen AI is it's much more accessible and easy to build. I think you don't need to do what you had to do with predictive AI, which was get the data into shape, engineer it, feature it, and all that stuff. You have a model that's already pre-canned, and it's just about bringing your data to it in the right way and checking the outputs of it. I think it's much more. Even in the early days of deploying Azure OpenAI, we had sales teams building demos and POCs with customers, and we never had that with predictive AI.
19
00:15:36,640 --> 00:15:44,640
SPEAKER_01: Yeah, the hurdle to entry is definitely much, much lower. This also begs the question, where are we going with that? What's next, and what will we see in the AI space?
20
00:15:44,640 --> 00:15:51,640
SPEAKER_02: I think what I'm really excited about is we announced the capability last week at the AI tour in London.
21
00:15:51,640 --> 00:15:54,640
SPEAKER_01: Can you give a date? Because we will publish this a little bit later.
22
00:15:54,640 --> 00:17:55,640
SPEAKER_02: This would have been mid-October. At Microsoft, what we do is we do a set of AI tours. We go to every country. Satya was in London and France and Italy maybe last week, so the week before Money 2020. So yeah, mid-October. There was a presentation that we did on a new autonomous agent capability that we're building. One of the things we've built is a low-code platform to build out your use cases, so you can actually describe what you want to do, and it will build. Fascinating. It's amazing, and you don't need to build a front end to this. You can just use Teams. You can push your custom co-pilot, and it can be very simple. It could have a HR document, and you just wrap a co-pilot around it in five minutes. One of the demos I do sometimes with banks is just put it in front of their website, and it's a bot that's kind of better than the one they have currently in five minutes. It's just incredible. With this capability now, what we added was these autonomous agents, and what they can do is they can work in a process that you design, and if they can work by themselves, and if they need human intervention, or human approval, or rerouting, or if something goes wrong, they can reach back up to a human for help. In the demo, what we saw was an email coming in to McKinsey, and it was asking something about a client engagement, and what it did was five autonomous agents kind of broke out the email. One of them went off and gathered the data on the customer. They found an advisor for the email, and then it wrote it back in the style of this advisor. It was doing about five or six things, and we have a whole load of inbox autonomous agents that do financial reconciliation and a whole load of common tasks, so you don't have to build these yourself, and you can actually build them with natural language as well, which is so fascinating.
23
00:17:55,640 --> 00:18:33,640
SPEAKER_01: It is fascinating. I always have to think about my granddad who is approaching his 90th birthday in a few months, and whenever he runs with my grandma on the street, they're taking a walk, and she loves to sing old Viennese songs, and she forgets a word. He just takes his phone, and with his kind of Viennese dialect traffic noises in the background, the phone gets him to YouTube to listen to the old song, and they proceed singing that song. So it's just to see what becomes possible for people who never had an education in that sense, who challenge or are challenged with even using a keyboard or something like that, is just a glimpse into what we will see in so many walks of life and obviously also in the business side of things.
24
00:18:34,640 --> 00:18:57,640
SPEAKER_02: I do think it's hard to predict. It was a question on the panel. I asked what's going to happen if we're on this panel in two years, and I really don't know. I'd love to know. I'd probably be very rich. But I do think a lot is going to happen, though. So if you're only starting now, you're sort of already behind. So just start. Don't wait.
25
00:18:57,640 --> 00:19:38,640
SPEAKER_01: Exactly. We had somebody on stage yesterday also mentioning, hey, there are two big mistakes. One is running too fast. The other one is not running or not walking in that direction. So his recommendation was crawl, work, and definitely get going. But usually when I ask guests on the podcast is, okay, what's next? What's happening in the next three to five years? I either get some shy laughs and, okay, I don't have my crystal ball with me, or people who say, okay, now I need to make a really blunt and bold, over-exaggerated thing, and then looking back in five years, I will laugh about myself because it was way too little what I predicted. I once even got invited to TV to say, okay, how are we going to do in 50 years? And I was like, sorry. Not sure what I can tell you about that. So it's really, really challenging.
26
00:19:38,640 --> 00:20:12,640
SPEAKER_02: I do like thinking about it, though. I think it's fascinating because there's a lot of things. Like I started working in this industry. I was always working in technology, but like 10 years ago, I opened up this deck from 2010 when I was back in sales. I opened it by accident, and it was all the same stuff. Customer experience, modernizing car, digitization. So a lot of things haven't changed. I think what's changing, though, is the technology is pushing it now. It's kind of pushing it to modernize. I think banks have to catch up now with the technology.
27
00:20:12,640 --> 00:20:28,640
SPEAKER_01: What else could be an incentive for banks to do that? You mentioned that in Canada, Australia, there is this competitive environment which is fostering the readiness and openness to embrace technology. What else do you think will facilitate other organizations on their journey towards embracing AI?
28
00:20:28,640 --> 00:21:41,640
SPEAKER_02: I think in some ways there's no incentive needed. When I look at the patterns of this, in the early days of GPT, banks had to block it. And CEOs got very excited about it. So for banks, really the thing is there's people in the middle that don't believe in it, which is okay. Some people need to be convinced by it or will try and resist. But you sort of have to embrace it. One of the things I end my keynotes on is my youngest kid. She got herself into trouble with me. She hacked her screen time. I've said this on another podcast, but again, she gave us an AI apology. And it was really long. My wife taught. It was lovely. The point is she is using it every day. She has ADHD as well. Sometimes she uses it as a tutor to just get things explained to her in different ways. But I think in two years or five years maybe, it will be like if you're a knowledge worker, it will be like we don't have AI in our bank. It will be like you're not going to give me AI. It's like not giving you a laptop.
29
00:21:41,640 --> 00:21:57,640
SPEAKER_01: Exactly. So that's definitely a commodity that we're going to expect. I was also curious since we talked about the success factors of organizations embracing AI. What are the most promising use cases that you see banks currently exploring?
30
00:21:57,640 --> 00:23:25,640
SPEAKER_02: I think there's really the kind of the one to start, I think, was code modernization. And I would continue to evolve on that. That's getting even better. I think contact center, each of these use cases has a sort of crawl walk run within them. So contact center was just transcription of a call. Now we have real-time agent assist, which is sort of coaching you all the way through. I think where you get this knowledge base that's learning from that conversation. And what I'm seeing that use case evolved to is hooking your complaint system into it as well. Sometimes these complaint systems are very off to the side. There's just somebody at the end of an email firing them around. Again, you could be getting the same complaint 10 times and the bank has never fixed the underlying issue. So I think that one is one that's really exciting. The one that's most exciting to me is how we might help the humans in banks create new connections with their customers or advisory roles. All the banks are trying to give them more time back, double the number of clients they have. Because these advisors are great. People do want to talk to a human at the end of the day. I don't think they're going to go away. I think the bank of the future might be more advisors than they have now. Is it? It could be more as you go up a segment or two.
31
00:23:25,640 --> 00:24:21,640
SPEAKER_01: I'm thinking a bit of different things I've read here. One study looked into what's the age gap or what's the age bracket where people really want to talk to an advisor versus the younger generation. Hey, if I have to do a phone call, I'd rather not do it. Or if I have to go somewhere to talk to a human being and they would appreciate a more in or non-human involved experience. So that's one element. Then I can't cite the sources now, but I know that there were some researchers looking into this. Which type of information would you like to get communicated by a human being versus a bot? And I believe it was something about positive news. I would love to get broken by a human versus something negative rather by a machine. So I think it would be curious to figure out the balance where we want to have more humans in the loop, where we want to have something that's happening in the app. Maybe not that apparent or in your face as other interventions. So I'm super curious.
32
00:24:21,640 --> 00:25:44,800
SPEAKER_02: It could be that the contact center will be our advisory channel as well. I think that's where it's going. And one of the great things we're seeing with that use case is that role is a very unappreciated role. Like you're getting people on the phone who've been waiting half an hour if you can't answer that question. And what we saw pre-deployment to these use cases was like 30% churn or 50% churn. We're even using it in our own support functions. Like every support call into Microsoft, we have an agent that's running it all the way through. And what happens when you call into Microsoft, if your Xbox isn't working, it goes around a different tree of things that can happen. We track everything along the way. The whole conversation is recorded. Nobody comes in cold to the problem. And what it's doing is it's freeing up really expensive support people. I mean, we're actually making them build new features and engineering things. So there is still people that don't want to talk to a human. I think what I would love to see in an advisor is one that can really do tax and advisory. Like I think every time I go to my financial advisor, it's like, here's what you can do with your 401k. But they can never tell me what to do with the tax, you know, or let somebody else. And I think that could be something.
33
00:25:44,800 --> 00:26:54,039
SPEAKER_01: Yeah. I'm also thinking, we see the shift also in the healthcare space where today's healthcare still is something that's very reactive. If you're ill, this is the pill, take it and hopefully you get better. But we all know that lifestyle influences your health. And there's so much that we all can do from optimizing our sleep, nutrition, etc., which will contribute to keeping us healthy. And I think a similar possibility is in the financial health of people that we say, okay, everybody wants to have money when they retire. Everybody wants to have certain saving goals, investment goals. And of course, our financial behavior is also something that has certain implications, but we might not be aware of it until we realize, hey, there's something I should really buy. There's a kind of thing I need to get fixed on my house and now money's getting tight or something like that. And I think there's a lot of potential with behavioral data analytics to have a coach built in your banking app telling you, hey, you should really do better on that. And then maybe the role of the human advisor could be similar to a fitness coach, the capability to interpret all of these day-by-day metrics and I don't know, have a yearly or a six-monthly check in telling you, hey, to achieve your financial goals, this is where you really should be better. And this is the big picture if you look at it. So I think that would be something that's super valuable.
34
00:26:54,039 --> 00:27:32,960
SPEAKER_02: Yeah. I did see the concept of self-driving financial management and well-planning. I think that's really interesting to me. I'm terrible with money and the bank I do have, they give me a pie chart and sometimes they slap me if I have a Starbucks and it's like, that's not very helpful. So I would love to see my own copilot just for me, not kind of even doing this best action, but just be super intelligent. I think all the data is there. I think banks have to be a bit careful. They can't do what Amazon do or retail do.
35
00:27:32,960 --> 00:28:43,880
SPEAKER_01: I think it's an interesting picture of, we all know what we need to do to get healthy, to improve our financial health. The question is, will we do it? And I think what moves the needle and what incentivizes people is different from individual to individual. Doctors also always complain about that, to tell the patients over and over again, hey, don't eat as much fatty food, et cetera. So I'm thinking of an interview that I recently read of a mom that described how she's doing her financial management and she says her most important thing is quality time with her kids. So whenever she's deciding, okay, should I buy this expensive pair of shoes or will I get that car or that car, she's calculating the extra amount in holiday, vacation days with her family and then kind of relativizing, should I really do this investment or would I much rather save the money to go on vacation with my kids? So this might be the incentive for one type of customers to tell them, hey, you actually want to have more nice vacation days, do you really now want to get at Starbucks versus with others, it's the slap on the wrist and hey, you want to do the extra coffees or something like that. So I think there's so much potential in having more behavioral data analytics and helping organizations help their customers to do much better and achieve their financial objectives.
36
00:28:43,880 --> 00:29:41,399
SPEAKER_02: Yeah. And there's some, some of the banks are doing some really nice things there. There's a lot of gen AI customer facing things now and they're all doing financial education. They're not giving you, they're not telling you to buy shares at a certain time, but I think that's a really good safe use case, you know, and the things to do are the same for everyone, like get rid of your debt, you know, get rid of, start investing as early as you can. And, and, you know, one of the banks did one really targeted campaign and they did a true Instagram and they just connected it up to their wealth products and I think that's really, really smart. Yeah. And again, low risk, you know, you're not, you're not going to do anything silly. And one of the great things we have in Microsoft is we have a whole lot of responsible AI filters and how that we can put in the conversation. So it keeps it safe and lots of content safety wrapped around it because you don't want, you don't want a bank to go on the news to say, this bank gave me a recipe for a bomb or something like that. You know, I think it's those, those are the things that I think CEOs really worry about.
37
00:29:41,399 --> 00:29:42,399
SPEAKER_01: Yeah.
38
00:29:42,399 --> 00:29:50,040
SPEAKER_02: That's what my CEO said to me, one of the banks said to me back in Ireland, just keep me off the news. Number one objective.
39
00:29:50,040 --> 00:30:11,200
SPEAKER_01: How far can organizations like Microsoft and providers of all these different building blocks and components that more and more organizations use to build their AI systems take over that work of responsible AI versus where needs the expertise of the business chime in to kind of avoid landing on the news and something bad happening? Yeah.
40
00:30:11,200 --> 00:31:25,040
SPEAKER_02: I think we, we've been on a journey of Microsoft, we've been doing responsible AI for years and I think we started off with like a set of principles, like a Bible concept and it was all around protecting the individual, avoiding denial of service, safety, all those things and transparency, but what we've done in the last couple of years is we've codified it now so you can actually build something and build it right into, into the models. You know, we've, we've, we've tools that are looking at bias and data and they will actually tell you there's a bias in this data. It's all predictive learning, predictive machine learning based, but we open sourced it all as well. So you know, anyone can use it on other platforms. So I think I do think there's, I do worry like sometimes about this though, I think with Gen AI, what we're seeing is more AI in the corporate functions of banks that might not be as well kind of regulated if you know what I mean. So I do, I do see, I've seen some sort of bad use cases where people, you know, banks are looking to scan it, scan CVs and things like that and I don't, I don't like that, you know.
41
00:31:25,040 --> 00:32:18,480
SPEAKER_01: Yeah. I think you always need, I mean, I have a focus on AI ethics in my work and I think it always needs this combination because on the one hand we don't have the expertise around the world. We know that AI talent is scarce and responsible AI talent is even scarcer. So I think it needs organizations like Microsoft, yesterday I talked with NVIDIA to provide as many tools as possible to make it easier and lower the hurdle to really assess is there bias in my system? Is there something that I shouldn't do? But I think we also need this collective knowledge and expertise to ask the questions, okay, only because we can do it, should we do it or is there any other way how bias could enter into the system? Unwanted discrimination could be there. So I'm super curious where this will evolve in, but I'm optimistic that in the midterm we will find a good way to make it happen and maybe even improve what we're currently seeing in terms of systemic discrimination between human individuals. So I'm a tech optimist when it comes to that question.
42
00:32:18,480 --> 00:33:28,080
SPEAKER_02: And a lot of the, I mean, there's a lot of good use cases here, you know, like when I came to the US, I didn't have a credit score. Now I'm not complaining about access to credit now or anything, but I'm just saying it was weird for me and I couldn't get a, couldn't buy a toaster on a credit card, but they will give me a mortgage. So you have to ask yourself, what's that about? So it's, I think, I think with some of this tech, you know, we've some, we've some fintechs and they're building financial inclusion, you know, people that didn't have access to credit, they're, they're able to get access to it in the right way. And I think, you know, the, the signals that are currently used, I think there's still, if you bring in more data to that, you'll actually see that, you know, maybe single parents will work twice as hard as anyone else. They're not, you know, I'm just getting, I'm using that as an example. I'm not saying anyone's doing it, but there's more, more data actually would, would have sort of triggered that, you know, made that easier for me to do, you know, even my credit score was very low for a couple of, couple of years, you know, so, but I'm not complaining. I'm just, that's a real force for a problem for me, but I, I, I can see it. Absolutely.
43
00:33:28,080 --> 00:33:49,640
SPEAKER_01: I think this potential also to target and equip the community that's currently underbanked underserved with better offers to credit, et cetera, I think is a possibility because the more data you have also the risk assessment from the bank side works out better. So I think that's another area where I'm excited about the data analytics and use of data will
44
00:33:49,640 --> 00:34:39,880
SPEAKER_02: help us. And we saw some cool stuff in COVID. You know, we have one, we have one partner here and they have, they have a tool that does that self-driving financial management, but what it did was it looked at signals for people that might be experiencing financial difficulty or it could be small businesses. And instead of kind of next best offer, it was sort of, you know, it looks like you might be, you might be in trouble here or, you know, are you, are you okay? We will give you three months on the holiday on the loan here, if that helps. And here's some financial aid from the government. And I think that's really powerful and, and, you know, banks don't, I think, I don't know, banks and customers sometimes don't communicate to each other. I think people go into a state of denial and then, you know, if you can even do it without having to talk to a human, I think that's even better.
45
00:34:39,880 --> 00:35:09,520
SPEAKER_01: I can imagine that this would be one of the areas where many people might feel ashamed or something like that and don't want to reach out to human being. But if the app would offer it, they would gladly take it. So as mentioned, I think we still need to figure out where and how to present it so that it doesn't freak out people that still have a human in the loop. But I think there's much potential for what we could do better. But I also wanted to discuss with you. We talked about what's next in terms of AI, but there are obviously many other new technologies that will impact the financial services sector. So which of those are you most excited about?
46
00:35:09,520 --> 00:37:01,840
SPEAKER_02: Yeah, I'm very excited about quantum. You know, I think that is a lot of people have forgotten about it. And well, I think it's, I did a couple of white papers around the use cases there. I think there is a Y2K moment coming up when quantum encryption will break encryption and the good news is that there's an army of people looking at it. You know, we're working with our other competitors, regulatory bodies and governments to get ready for that. But I think one of the things that strikes me is more protection, you know, more, I think, better security. It'll be more secure. Yeah. Just last week, like my wife went to pay our tax, our motor tax, you know, and the worst thing about using a search engine in 2024 is you get a series of links that are just fake sites, you know, and you have to kind of go through them. So she went to one of these sites anyway and put 20 bucks into it. And I kind of said, OK, that's not the DMV. We need to change the credit card numbers here, you know. So I think those kind of things, I would hope they would sort of disappear and better. You'll have better identity protection, better protection on the payment, on your wallet. You know, account takeovers will be much more difficult. But again, banks need to be ready. You know, they need to be in the cloud, actually, and well in the cloud, not dipping the toe or doing tests or SaaS even. They need to be, they will need to be able to get access to the capability. Now, the good news is we're going to democratize that or make it as accessible as possible. You know, we've even seen it in the last year and a half. The barriers to entry to Gen AI are coming down all the time. Our cost is coming down. The accuracy is coming down all the time. But it might be like that for quantum in the early days.
47
00:37:01,840 --> 00:37:12,359
SPEAKER_01: Interesting. Maybe since you mentioned the barriers to entry, what are still today challenges that organizations have when it comes to moving to the cloud, when it comes to embracing AI and then in the later years, quantum?
48
00:37:12,359 --> 00:37:56,080
SPEAKER_02: I think it's coming down. I think a lot of them, you know, in the early days of cloud, it was cost, you know, and then it was like the regulator won't let me and those things have gone away. You know, I still think banks, banks in the US love being on-prem, they love their on-premise data centers. But again, I think we're going to be in a multi-cloud world as well. So it's always going to be hybrid. I don't see, I don't see a vision where everyone is going to take everything there. So it's going to be a journey. But there are certain things now with AI, if you want to do really high-end stuff, you know, you won't be able to, it won't be cost effective for you to buy a rack of GPUs, you know, but you can use them in our cloud, you know, as you need to and then spin them down and run on Azure, run on our platform.
49
00:37:56,080 --> 00:38:50,799
SPEAKER_01: Those type of machine learning services in the cloud are actually one of the top use cases for our financial services customers to embrace synthetic data because they then train on the synthetic replica of the data and bring it then back to their on-prem solutions and the production data. So that's definitely also one use case that we still see over and over again because there are internal policies and concerns of data security. So curious how long it will take for them to be more comfortable with the protection that they actually enjoy in the cloud. I also wanted to talk with you briefly about this people component, because what we see obviously with AI and with all the new technologies is that the acceleration, the speed is something that society simply wasn't used to with all the technological advancements that we had in the past. So do we have any top tips, secrets for our business executives listening when they're embracing AI, how they should make sure that the people side of things doesn't get left behind and people really feel comfortable embracing AI wholeheartedly?
50
00:38:50,799 --> 00:40:16,440
SPEAKER_02: Yeah, I think a couple of things. There's training that you need as a user, you know, so we've done a lot of that. We've built tons of training around how to do prompts better. I think in the data side actually is where there's huge opportunities for people to reskill. You know, if you think about the value chain around data, you know, an AI, a predictive AI, you've got, you know, data engineers, you have people that build the models, then you have people that sort of do the testing on the models and then they push it into an application. And then you have people telling the story around the data. And I think a couple of years ago, like people had data scientists do all of that, you know, the really good ones could do it and they got paid tons of money. But it's not sustainable for a bank to do that. You know, what you should do is break that chain into five or six different people and become really data driven, try to build a data driven culture. That's what a lot of my CEOs want to see. You know, I was on stage with one financial institution and the CEO said, I want everyone, my vision for this bank is everyone is going to be part of this value chain. If you don't want to, your talents might be better somewhere else. But it was very, this was already a very data driven organization. But it was, you know, he said, we're going to give you a ton of tools to do this on training. And, you know, I think the tech is getting easier. Absolutely. Thinking of my granddad again.
51
00:40:16,440 --> 00:40:45,039
SPEAKER_01: So I think the entry barriers are what we see with how Gen AI can help people who don't have any data background, any education to simply get answers to questions about their customer base, about some analytics where in the past, it took them weeks until somebody from the BI team, analytics, data science team finally had time to tackle their requests. So I think once they get the hook of that and figure out how easy it is to talk with your data, powered by Gen AI, that they will also start embracing that more.
52
00:40:45,039 --> 00:41:14,840
SPEAKER_02: Yeah. And I think, you know, even, I think about Microsoft as the last mile for data. You know, we've had Excel, we've got a great Power BI tool, you know, in teams as well. And I think what we're, what I, what I see, what I saw with Excel was, you know, organizations built tons of macros. It was the first time we put data tools in the hands of the business. Yeah. And I think that too is going to, I think Excel in 10 years, you'll be able to do any kind of machine learning, like with a couple of clicks, you know.
53
00:41:14,840 --> 00:41:17,359
SPEAKER_01: So I think. If you even still have a device to click.
54
00:41:17,359 --> 00:41:19,559
SPEAKER_02: If you still have a device, but it'll be linked to your brain.
55
00:41:19,559 --> 00:41:22,919
SPEAKER_01: Blink of an eye, the brain, whatever it's going to be.
56
00:41:22,960 --> 00:41:35,679
SPEAKER_02: But yeah, I, and again, no, I don't, don't, don't, people try not to quote me on that. That's just my own opinion, but I don't, I don't, I don't speak for the Excel product. But, you know, but that's kind of what my own two euro cents on them.
57
00:41:35,679 --> 00:41:44,880
SPEAKER_01: Makes sense. I have a few more questions for you, Daryl, before we go into that direction. Is there anything else we haven't yet discussed, which you think would be valuable to share with our listeners?
58
00:41:44,880 --> 00:43:19,199
SPEAKER_02: No, I don't think so. I think, I think the thing is on the skills thing. The other thing I'd say is we're all learning this together. You know, I'm not an expert at all. Like I just learned from, I'm very curious about what other people are doing. And I spend a lot of time looking at the patterns too. And, you know, one of the things I'd say too, if you're, if you're coming up at 300 use cases, one of the things we've done across all our industries is we know they're not really 300 separate technical stacks. If you know what I mean, there's actually six or seven common technical patterns. You're chatting with data. You're looking at a conversation where you're generating some content basically. So what you should think about is it's not building a stack every time for each of these, you know, build. Yeah. You know, we have accelerators for all these patterns as well. Just start with these partially assembled Lego sets and you'd be able to crank out, you know, a hundred in no time or, you know, the top 10 that you have, you know, so, excuse me. So, yeah, so don't be, don't get, don't get, don't overthink it as well. Like, you know, a lot of consulting companies come in and spend six months trying to find the perfect silver bullet use case. And, you know, business cases as well. I've, I've, I've been burned by them. You know, if you, if you do a business case and then, you know, you put it in a drawer, you do the project. You're lucky if nobody opens that drawer. Sometimes it doesn't look a little bit different. It does. Or sometimes it over, over delivers as well, you know, and, and, but they're like sharks, you know, they just come on, come back and bite you.
59
00:43:19,199 --> 00:43:50,479
SPEAKER_01: Pandora's straw. Don't look at it. You actually provided me with a good segment because I wanted to go towards the skills question. And you mentioned that you landed in your role more or less per accident, but I was curious about your take on what are the elements that make you successful in your role? And what are also other elements where you say, okay, with the banking executives embracing AI, what are the soft skills, the qualities which you think are much, much needed in today's time, times to really succeed with AI and new technologies and digital transformation?
60
00:43:50,479 --> 00:44:14,680
SPEAKER_02: Yeah, I think, well, the first thing is, yeah, we are learning all together. I think people have, I'm amazed by humans' ability to change what they do. You know, I started out in technology. I worked on a help desk and I'd worked in, and then I started working on infrastructure stuff and then I did some architecture stuff. And then I did some, I came to Microsoft to not travel anymore, which is ironic because then I came over here.
61
00:44:14,680 --> 00:44:17,680
SPEAKER_01: You just talked about all your business trips before we started the recording.
62
00:44:17,680 --> 00:45:17,239
SPEAKER_02: So, so yeah, so I, then I fell into sales and then I ended up. So I think what I, what I try to do myself is be a translator between the tech and, and the impact of the tech and, you know, what you need to do with the technology. It's very, I don't, I don't try to, but I don't try to think that I can do all of it. You know, I think the thing to do is, you know, as many stories as you can, you know, what is the bank doing, what's a similar bank doing to you, you know, what you can learn. And, you know, there's people making mistakes here and they're not really, I don't think of them as mistakes. They're just experiments that went wrong, you know, at one bank and they, they built their own LLM with like 90 engineers and it was pretty awful. Another bank who, they took a CRM system and fine-tuned it into a large language model and then asked, what do I do next with this customer? And it didn't know what to say, you know, just, I don't know what you're asking.
63
00:45:17,239 --> 00:45:18,479
SPEAKER_01: What do I ask you to do?
64
00:45:18,479 --> 00:45:51,560
SPEAKER_02: So I find those things very interesting because nobody knows, you know, they could have been brilliant, like they could have been, it could have been fantastic, but it wasn't. So I try to sort of share those things with my customers, like don't try and build all this yourself. Don't, don't try, banks love building stuff. I'm always saying that, but, you know, take some, build, you know, use services from AirCloud or another one. Don't just, you know, don't focus on what is, what's unique to your customers, you know, and what unique value you have. You know your customers and your products better than anyone else. We don't know them.
65
00:45:51,599 --> 00:45:56,079
SPEAKER_01: Yeah, absolutely makes sense. Anything else?
66
00:45:56,079 --> 00:46:06,800
SPEAKER_02: No, listen, just thanks for having me. You know, I didn't, I didn't expect to be doing a podcast of Money 2020. So it's a first. So you'll be very, very easy to chat to, so.
67
00:46:06,800 --> 00:46:16,719
SPEAKER_01: It was a pleasure having you in the show. There were many takeaways for our listeners. So I think they will thoroughly enjoy the episode once we're going to publish it. Awesome. But thank you so much for your time, Daron. Thanks very much.
68
00:46:22,560 --> 00:46:25,040
unknown: Now we can all see why Daron likes to be on podcasts.
69
00:46:25,040 --> 00:46:42,160
SPEAKER_01: He's simply a fantastic conversation partner and I learned a lot from him today. As always, if you have questions or feedback for us, you can reach us via LinkedIn or write us an email to podcast at mostly.ai. Thank you for listening today. The next episode will already air next Thursday, so stay tuned for that.
