Digital first bank: How GenAI can reduce tech debt and self-fund digital transformation with low code

Join industry technology executives leading with GenAI solutions. We will specifically discuss an approach to reducing tech debt and how leveraging high-performance low-code platforms can accelerate your digital transformation journey. Hear leading approaches and insights on how banking is strategically adopting GenAI and solutions implemented to scale.

Transcript:

Matt Hugo (00:07):

Nice to meet you. Welcome to a lunch and learn. So you guys have the lunch and now comes the learning part. My name is Matt Hugo. I'm excited to have you guys here, guys and girls for today's talk. We're going to focus on the intersection between low code and generative AI. In many exhibit halls between those two words, you would see those words on every single vendor. So obviously they are huge buzzwords right now. We want to sort of demystify some of it for you, apply it towards your space and get you thinking about how you might be able to bring some of this knowledge back to your financial institution to hopefully take advantage of some very interesting new technology. You want to introduce yourself?

Panel Member (00:59):

Sure, thanks. Hi everybody. Am I audible? Yeah. Cool, cool, cool. So I'm solutions architect without systems. I'm here to answer, explain and help all of us learn about how out system boards well with generative AI and what kind of magic we can bring to a banking field without systems.

Matt Hugo (01:23):

Yeah, Awesome.

Matt Hugo (01:25):

Thank you. And I hate to be the guy to start one of these with a raise your hand, but just so I know what we're working with. Has everyone here heard of low code? Just generally enough nods? Yeah, we're good. Okay, cool. So I started at OutSystems five years ago. My name's Matt Hugo, Enterprise Account Executive. I was cold calling and I'd always start off with, have you heard of low code? Zero People said yes, maybe to get me off the phone, but it was not a huge thing yet. And so I've seen this transformation of companies adopting low code, it being the next thing and the type of interest that comes in and then honestly some of the obstacles and pitfalls that you need to be considering as you implement a strategy. So I want to talk a little bit about that. My focus will be a little bit more on the low-code side of things while Advait will come in and sort of apply that to the generative AI space.

(02:23)

So, low code, I guess makes sense that it's sort of exploding. Looking at developer demand, it's never been harder to maintain existing systems. The requirements have never been so complex. It's not enough to just have a mobile application that connects with some systems internally. No, you need to be differentiating, I think of like IOT, KYC, some really edge use cases that are becoming very dominant right now. You needed to have a strategy two, three years ago for that, and if you haven't, then a low-code platform is going to help you get there a little bit faster. There's also requirements for maintaining legacy systems, which is the bane of many people's existence. All while there just aren't enough developers out there, 4 million developers is the number right now. That's the expected shortage by 2025. There's also a silver tsunami. So in many spaces, banking very much included.

(03:25)

There are, it's kind of like Latin, it's like a dead language, but there are still developers needed to maintain things like cobol. So it's running into some serious issues and this is all leading to bigger and bigger application backlogs. I call them the good idea graveyard. You have idea people at your organization, they come to it, they try to trust it, and the speed at which their idea is turned into a practical solution is almost never fast enough, especially for their liking. We all know that there are reasons for that, but at the same time, those people are probably going to leave the organization. There's nothing worse than having an idea for a feature six months later your competitor launches the same thing and you haven't really acted on that strategy at all. So it leads to this IT and business tension that with a proper adoption of a low-code platform, agile methodology and some different sort of communication tactics, you can sort of overcome that.

(04:26)

Right? And this is especially important. I saw this on Twitter and I wanted to pull it in. There's this idea of wartime leadership and peace time leadership. So in wartime, obviously things are a little bit more hectic in peace time, that's what we had seen for the past, say five years, right? Everyone's flourishing, you're creating some cool features, playing around experimenting right now it's just a matter of what can you push into production? How fast and how fast can you then turn around and take on those next requirements? You need to do more with less. And even though right now they're saying that we're now officially in a bull market, again, I'd consider it more of a calf market because everything's a little bit broken. So I think it's just going to get worse and worse as far as needing to do more with less.

(05:19)

So people turn to low code and without systems specifically, we surveyed our customers, why did you use or why do you use out systems? So no surprise, 78% shows faster development. I think it was choose three or all that apply, etcetera. But something I wanted to turn your attention to is the bottom one enable citizen developers only 14%. That might be a little bit surprising. A lot of people think low code equals citizen developer. I'll talk a little bit about what citizen developer means and some things to be wary of. A citizen developer is saying, you can an app and you can develop an app and let's empower everybody. And there is definitely value in bringing BAS and a marketing director et cetera to the table to build their own app. But I'll give you a quick example of when this could go downhill very fast.

(06:15)

So there's a company that has events and their marketing team wanted the ability to configure their own event applications each time that they have one. They were about weekly, and so an overzealous marketing analyst who is allowed to free range, edit whatever they wanted, observe that at events. The biggest bottleneck was everyone was at these tables paying for merchandise. What if we just collect credit card information up front, right? Why not just say credit card number, name, zip, etcetera, right on the form to register. So this went unchecked. Marketing director didn't really have a control there to see into it. It wasn't until one month later. Again, weekly events. There were four events where this happened, where their process was a monthly review of the database and the data that they were collecting. They saw that they were in very aggressive violation of PCI. The database was not ready to be collecting that type of information. Who do you blame there? I mean people outside of it and who aren't it adjacent or security, etcetera. They don't know what PCI means. They think that they're doing themselves a favor. So you need very strict governance controls so that you can get everyone in the room together and understand their role anytime that you're going to have a citizen development project implemented.

(07:42)

Someone from my marketing team is in the room and I wanted to keep everyone happy here with a Gartner report that has us up top. But two things I just want to call out when we talk about the low-code space. First of all, we are up top when it comes to ability to execute and the question is always what can't you do with low-code? But there's something very interesting going on in this magic quadrant specifically. There's Microsoft, Salesforce, ServiceNow, I see Oracle. There's even Alibaba and Huawei in China, extremely different companies and that's just further proof of how everyone's getting a little bit involved in this low-code journey. But I don't think enough people are questioning how you're going to put the right governance controls in place. So we're going to talk a little bit about that specifically through the ai, the generative AI lens as well, making sure that people's experiences aren't just smooth, they're also compliant and not going to be a total disaster for you.

(08:42)

I'll skip the next Gartner slide for you guys. You guys deserve it. And then one of my last slides I just wanted to talk about briefly was this idea of build verse buy has evolved greatly lately. The idea is how much can I be configuring now? And so in the olden days, I guess it used to be pretty much everything's traditional code and it takes a longer time to build, a little harder to maintain, but you would get that specific ability to hit any of your requirements. There's also BPM tools now and that is a little bit of low code. So Appian, Pega, etcetera. Then there are your core systems that often have extension tools around them. They're often industry specific, industry focused, and there's also no code in the space. Something that I not many people really have an answer for is when you think about the no-code space, what's the difference between that and an off-the-shelf solution That's just a little bit more decoupled a lot of the times.

(09:43)

At the end of the day, no-code means you can't add code. So whatever the product management team at one of those companies has built is what you can use. And that is good from a governance structure because you can't have these runaway like ungoverned shadow IT disasters like the one that I mentioned. But can you have the flexibility that you need? And a lot of these solutions, the question is how are you differentiating your brand once you keep up, once you implement the system that every other bank around you is also using, how can you make sure that your customers will be able to tell the difference and have a reason to go tell their friends to go join your bank as well. So I guess without systems specifically, we are actually focused on low code for professional developers. So using some visual drag and drop capabilities and being able to ideally implement a solution faster than it would take to implement a bot solution while you were the one who built it.

(10:46)

So it'll be hitting all those requirements and this is where we're our presence in the banking industry. As you guys know, you probably have a lot of NDAs with a lot of your vendors. So there are logos that I would love to have here that unfortunately I cannot. But the point is there are a ton of vendors out there right now in the exhibit hall and we don't just do loan origination, we don't do loan servicing, we don't do any of these one use cases. Our goal like we're doing at Key Bank is a wide scale transformation where you're continuing to build on reusable parts that you've already built out and there are ways to be infusing chat, GBT and different generative AI elements into that as well as you build for your exact system. So five questions that I wanted to leave everyone here with.

(11:40)

If you are considering a low code evaluation, if you're considering things similar to low code, you're also going to want to probably listen to this. So one question you're going to want to ask is what types of applications are filling my backlog? You're going to notice some patterns. There might be certain departments, certain teams, certain types of roles that are continuously being underserved. Maybe that just boils down to their management above them doesn't fully understand the importance of that technology. However, everyone under their division might be having an incongruent experience with other teams. Then number two, how will I balance self-help with governance? To be abundantly clear, if you have a citizen development program, you need to get everyone in the room who's going to be involved, InfoSec the actual business users it and they need to understand their role. So things will not go off the rails.

(12:35)

What are my most complex app development requirements? The biggest reason that I've personally seen people come to OutSystems would be we built a say insurance, a policy management system. We got about 80% of the way and this no-code vendor wasn't able to keep up with the business logic and that was a sunk cost of 15 months of development. It can sour, it can leave a bad taste in your mouth for low-code in general, which I fully understand who will do the development. So I mentioned citizen development, I mentioned I have not mentioned there's a different vendors are going to have different delivery partners. GSIs, do you want your team to be self-sustaining within a year? Something to think about. And so I think a lot of no-code or low-code platforms, we'll play into that citizen developer space, but I liken it to would you give TurboTax to your CPA?

(13:33)

Absolutely not, right? They're going to lose capabilities that they would be able to provide. They're going to feel like it's a watered down experience. And that's similar to when you give a citizen developer tool into the hands of true developers. Lastly, thinking about what an orchestration layer across your business would look like, you want to be able to reuse your development, you want to integrate with systems and you want people's individual experiences to be as smooth as possible. And this will be a nice little gateway into the next slide, which Advait will present. Something I've sort of been reading into is the idea of super apps and how they can combine with generative AI. And I think five years from now, five to 10, we're going to see more and more sort of workplace experiences that are focused on really one single application where you are provided some information, you provide human responses, things that AI cannot do, your response is then processed and pushed to the appropriate systems.

(14:37)

And that's sort of what you do for a living at that point, right? Like RPA is getting there, generative AI is getting there so they will never be human and we kind of need to keep that in mind. Humans that will be running these super apps. And so are you ready to get there? Is that part of your five year plan? Realistically maybe 10, but are your systems moving in the right place for that type of experience? It's something to consider. So I'm going to hand it over for the time being over to ate. I'll chime in when I'm antsy and over caffeinated, but thank you everyone.

Panel Member (15:18):

Thank you so much Matt. Yeah, that was pretty much ground setting about how generally things are moving right now from the low-code perspective. And let's ponder over the question, what is generative AI? I mean everybody's listening at least 10 times a day that what generative AI is building there is generative AI is building applications. So generative AI is basically, that's nothing but an algorithm, okay? It's a set of workflows, it's set of procedures, it's set up processes, it's like which is creating some new content based on natural language processing, which is something like you can talk the language which anybody can understand and based on that you can create the magic around you. So that's precisely is generally where now when we are talking about the apps created by generative AI, so there are plethora of applications outside there. You open your LinkedIn and you'll say some AI is just building the application in no time. I've just mentioned this. And then there is, so you have practically no way to figure out whether is it really something which app did or it's something a video trick. So whatever we see right now around us is about development of application. What everybody is talking about is quicker development cycles, right? Faster debugging and troubleshooting. So that's how they can be am audible. Okay, cool.

Panel Member (16:54):

Is that great? Cool. Sure, thanks. So the applications which this AI app companies are claiming is basically just limited to the development or the troubleshooting part, but what about the accuracy of the results? So we have to sometimes stretch a lot about the results which we are getting from this kind of AI generated images, AI generated responses. They may or may not be true. It still lingers over the confidence level and that's why I name it as AI hallucinations. There is absolutely no governance structure. So there is nobody accountable for any wrong responses. There is no guidelines about how you particularly need to go for it and it is only limited to development. The enterprise application development is not just constrained to development. What about testing? What about the CICD? What about the DevOps? So still there is some knowledge or some development development needs to be happening over there.

(18:00)

And then even if you kind of bridge all these gaps, what about the talent? You still need some specially skilled developers or analysts to these kind of stuff. So that's where generative AI as it stands right now with the AI powered applications. Now let's see, what is the AI and OutSystems relationship so far? So OutSystems platform. So OutSystems has a dedicated AI team in our R&D and the product team since 2017 and that resulted in the platform having AI since 2018. So it's not something new when it comes to OutSystems. So we are not just jumping the wagon that okay, wow AI and MLR the buzzwords nowadays. So it was a well thought out strategy and that's where out system were actually preparing for their own service that is AI fusion. So instead of depending on any other third party tool for getting their own AI services, OutSystems invested focused and now we are at the stage where we can and we are reaping the benefits of having our own AI fusion.

(19:15)

Now what it does, it basically helps the developers to think when they have hit a wall from the development point of view, it also gives them option to analyze their code. It also gives IT leaders to look what is their technical debt looks like, what is their security threats look like? Where are their performance improvement options are how their per code maintainability is looking like because right now what happens is these particular tasks are like automotive or repetitive tasks are causing a lot of money and also having the talent gap where your developer needs to attain certain goals within certain timeline where they need some guidance, they need some education.

(20:07)

Also, the key part from OutSystems platform perspective is AI mentor, which kind of provides a gallery to the entire warnings, which developers we developers are notoriously known for ignoring because developers only focus on errors. So what OutSystems does is it looks at those warnings, it categorizes them from maintainability performance, security and architecture perspective. So based on the priority based on the team which Matt was highlighting before, like InfoSec team can look at the security, the product team can look at the maintainability aspect, the architecture aspect. So for all the teams that becomes a one-stop shop where they can analyze the code and that's where the code development can really be handed over to little less experienced developers. You know can create applications, you can create prototypes, you can create frameworks which later can be utilized by citizen developers where you are sure that definitely there are some security guard rails guardrails in place. So the development will be going smooth.

(21:24)

Now how that AI mentor is split in our systems is code mentor and then we have something called as AI mentor. I'll be happy to demo that at booth 414. I would really welcome you all to stop by and I can explain what it is and how it looks and how it can be made to work. So we have code code mentor in our IDE which informs developer that how accurately he or she can move forward from the development perspective, what are the aspects they should be looking forward to And we have some changes made from the NLP point of view to our ID itself. So next week we have next step that is OutSystems big event where the next technological advancements with our platform would be announced many of the product details about that would be given in that particular event.

(22:21)

Moving on, so without systems, how to build AI enabled apps. So you can build cognitive services, Microsoft, Google and a Amazon AWS. They have their own cognitive services which out systems can help you build better with because those particular services which are provided, they're really headless. So you need some platform wherein you can build UI, you can track the performances of these services and also from cognitive services point of view out systems has its own out systems.ai document processor, which can be used for OCR processing and that is really helpful. So with building application with out systems is like building with all Lego blocks, right? So you can build some reusable components and you can use those particular components for building your applications further. With that, OutSystems has its own open source repository, which is absolutely free for all the OutSystems platform users.

(23:30)

And there we have these kind of AI components. Again, please stop by booth four on four. I'll be happy to demo them more in detail. Next is machine learning. Now all of the IT developers or IT directors challenge nowadays is how to use predictive algorithm to build their applications, how to build that intelligence, the smartness in their system. So for that you can either use your own machine learning libraries or the platform can help you to integrate with other machine learning libraries which you regularly use. Next is of course enable conversational experiences. So OutSystems has partnered with AWS, so you can use AWS lex chatbot or if you want to develop your own chat bot using your own data even that is possible without systems platform.

(24:26)

Moving on now, OutSystems advantage to leverage AI first is empowering junior developers so you can make best use of the existing development talent what you have because all the development which happens without systems is visual, it's drag and drop. You can develop all process orchestration, you can build all IDE, your pixel perfect UI databases, the data models, everything without systems. Next is privacy and security. So we provide security infrastructure based on whichever way you want our systems to be deployed. You will be having, your data will not be shared, you will have your own VPC. And then again from the governance and user controls point of view, we have authentication and authorization checks in place and again it's end-to-end solution. So without systems you are developing full stack, full lifecycle application. So you will be getting the development, the QA and the production server along with the audits and along with the logs for all the applications which you would be developing.

(25:40)

Now I will just quickly rush through the slides because for lack of time. So it's like how are we approaching the future point of view? So OutSystems has AI mentor system which would be working along with your developer and that would help him or her to develop the application and which would be more related to more focused on user requirements and how user is using the application. So that will enable you to map the steps which user is following the screens, which they are visiting the buttons, which they are clicking, the tabs, which they are following. And based on that you can build a predictive algorithm and again redesign the user experience if you wish to. Again, OutSystems is investing in its own language model because that's where you know can do the code generation then entire and it'll, it'll not be just limited to development, but it would also be involved in the entire CICD cycle.

(26:39)

It is also investing in its own data engine which would supporting the SDLC as well as it would be focused on the quality part. Next is again, OutSystems is particularly working on automated reasoning because at the beginning I mentioned about the accuracy part. So the idea is you should be able to develop your own large language model, which is serving to the governance, to the regulation which you are serving to. And it still should make you agile or nimble enough to make the changes in your application because the world is changing all the time. So you don't want to be another legacy module or another legacy application. When we are mentioning here, OutSystems is also involving AI research as far as the large language model or automated reasoning. And all this properties is concerned. We have tied up with Carnegie Mellon University and along with them OutSystems is working on AI research.

(27:38)

So as it stands right now we have seven patent filings and we have presented papers in 10 peer conventions so far. So we are planning to extend, I mean there is a plan to extend out systems ai, mentor to ops mentor as well as testing mentor. So even the guidelines for the operations which you have from the DevOps perspective and from the testing perspective, not just from the development but also we are taking it to all the elements of lifecycle where it matters. Now this is a little less talk aspect of the technology that is web accessibility. That is, it's like if this is particularly means for people with disabilities, so even they should be able to use, consume and contribute to the web application. And I think that's where the application development also should look for. So when we are talking about people with disabilities or accessibility, it is not just for people with disabilities.

(28:48)

So there are some temporary scenarios like we are stuck at some station at underground and we don't have network or there it's, we have some simple things like we have forgotten our glasses. So even that can make us temporarily disabled. So even these kind of scenarios we should be able to utilize or we should be able to factor in when we are developing applications. So OutSystems has some components which are part of OutSystems out of the box. So we have 70 plus double AA, WCG components which are part of OutSystems UI libraries which can be used when you are developing application. So in a nutshell, to give you an example of how single A, AA and AAA is defined is if you have a video with subtitles that would be single. If you have subtitles with multilingual ability, that would be AA. And if you have a video with subtitles, multilingual capability and sign language, that would be AAA. So for now OutSystems is complied for WCAG AA kind of features which can be utilized for your application development.

Matt Hugo (30:15):

Me again, I just wanted to talk about a couple of things that came to mind as Advait was talking. So we are also seeing a major issue when it comes to security, not the governance side, but the fact that it's really easy to make an application right now if you are a bad actor. So if get someone in a basement somewhere and they want to build something that looks like a bank, a banking app and they're going to give you very specific details within an email about what you have to do once you download this app, I don't click random links, but if someone's going to mention an app, it kind of makes it sound more credible. They walk you through a user journey where you just signed in to your banking account on a fake application with the style of the same UI style as your bank.

(31:07)

It can get a little bit scary there. And then also when we think about the manipulation of big data, something that generative AI is currently struggling with is it's based on data that has been inputted over the years. It's struggling to predict what future data will look like. So within the insurance space specifically, and this is pretty similar to banking as well, assessing risk, obviously not in a political way or anything, but thinking about maybe changes that have happened to the climate and how that might impact future disasters. That the current models are not doing a great job of predicting what that will look like and maybe the impact of certain industries. So when you're making accrediting decision, maybe someone's entire employment and livelihood could possibly be in an industry that might be impacted by environmental changes, just as an example. So it's something to also sort of keep in mind as you go.

(32:14)

I think we've all heard about recent issues with chat GPT. There was a lawyer for example, who cited a case that was not really a case and they're getting in a lot of trouble as you do when you commit perjury accidentally because you were basically using the equivalent of SparkNotes for law. So just some things to be wary of. And so we will be at booth 414. We're happy to, if you stop by we can send you a copy of this deck right into your email. We intentionally made it so some of the slides look really pretty and can make you look good in a meeting at some point. So we would love to see you. And that being said, we did save a few minutes for questions. I also want you guys to have the chance to do a little bit of networking here too, so we think you can get some time back, but are there any questions that sort of jump out from the things that we've covered so far? Oh, I was talking fast, I covered so much. If people need other topics, maybe we can cover those too. But thank you. I think that's probably good. Hope you guys enjoyed your lunch. Hope you learned something. So we'll see you at Booth 414. Nice to meet you everyone.