Latest / Architecting the AI Enterprise / Episode 2: From Tech to Business Impact—Rethinking the CIO role in the Age of AI
Transcript
- Marketing: Welcome to Architecting the AI Enterprise. Hi, I'm Gary Huberman, CEO founder of Unquark. Dave Ferrucci, Chief AI Officer and CTO at Unquark, long time AI researcher. Thanks for joining us again. In this series, we're going to be exploring Architecting the AI Enterprise. What does AI truly mean? How is it being deployed? What do we think the future holds as best as we predict it? And today we have a really exciting guest with us. Ala Whitston is the CIO for MUFG Americas. And Ala, I got to see you on stage at a AI event in New York City, which is where we met. But, you know, first thing I want to start with is just like, when I was an engineer, I wanted to be CIO. That was the goal. There was like, how is the path to get to the career roadmap? And in many cases, it sounds like the Odyssey book, The Hero's Journey. It's like, what's the story to get there? So. First off, if you could introduce yourself and what's the journey that got you to where you are today? Well, thank you for having me here. ⁓ It's really actually, I'm looking forward to this conversation because we talk about AI all the time, but we rarely focus or refocus ourselves AI for enterprise because the journey and the story is very, very different. So now about my story. ⁓ No, I never thought that I wanted to be a CIO. What I always wanted is to be a technologist. solve problems and then when I actually learn how to solve problems through technology, I decided I want to solve problems through technology for the business. And the journey to CIO, I think it was more along the line. I wanted to solve problem. I only as application developer or infrastructure architect only can touch and influence portion of and getting frustrated about not being able to really embrace and think about system and architecture design. specifically for business enablement. So my journey to CIO was, let me figure out how I'm going to actually get to the point in my career that I can talk to business, understand what they want, how to drive the revenue, how to be trusted by them so we can implement technology to enhance their productivity, enhance their product and services. And the best way to do it is to become a CIO. Amazing. So you wanted to learn the business. always used to tell my technology team, you should learn the business and know the business better than the business. So that's what you believe is basically just understand the problem to solve first. Absolutely. Every single time. I think as we start talking about AI, bringing it back, how AI is actually going to help business and how is that knowledge by the technologists will become one of the most important one of. the most important actually qualifications to be successful in this evolution of our technology field. I really believe that is going to be critical. That's amazing. And so you learned the business which helped you gain more knowledge in technology, how to apply it to the business. It's a circle. It keeps going up in infinite. So you get to the point of CIO. Absolutely. mean, the business wants people in technology roles who understand what they do. So when we as a technologist get to the table and we speak only our own language, we really don't reach them. We really don't build the trust with them because we speak language they don't understand. You have to figure out how to actually reverse that, how to get to actually say, I know how your process works. I've been there probably five years ago when we designed it in the first place. I know where your shortcomings are. I know now with the new technology, I can do so much more. What do you think about that? Yes, right now the table sitting at the ECA tables and we talk about growth, we talk about product and services, part of the organizations approve product and services. So we're not just being brought to the table to decide how do we improve in world old world, how do we design application to do X, Y, and Z. Like what can we do to compete? That's got to ask the word architecture. So is there a business architecture function that's equivalent to your technical function and how is technical architecture work? So that's a brilliant question. I don't think we ever called that business architecture function, but what we really called in very mature organizations, business organization is the product. So chief product officer in my ideal world lives within business and those products and services architected by the people who live within the business report to the head of business are actually the architects, business architect. I like that. No. And do you actually have product managers? have product owners? You have that role within the business today? We have some of them. I worked in the previous places, previous organization that was more mature than in others. So I think that's kind of the evolving field, but I think it's going to gain so much more right now with AI coming to the table. Well, we don't need to really talk about BRDs at the granular level. We need to talk about business architecture, business workflow, business product. The outcomes. Absolutely. That is so. I've seen outside in now as a vendor selling into customers when they don't have a product owner, the product officer, and they have a program manager, the outcome is not the same. So the product owner from what I've seen in the product manager role, they have a date to get something done by that's meaningful. Yes. And they push the date, they push for that. Is that similar to what? Absolutely. It's not just the date. Date is irrelevant. I mean, I always talk about the date being irrelevant. is working with the client and making sure that clients are ready to take the product and services that you just developed, you're delivering and knowing why would they be willing to change. When I worked very, very closely, when I worked very closely with the product architects, let's say on a business side, the biggest conversation, it's not about how do we develop them. The biggest conversation is how do we entice our clients to test what we are delivering and then adopt it. Because they live with that or they live without that or they live with the similar product and service by someone else. What is our comparative edge to actually ensure that our products were actually going to be used? what you're describing, there's always the concept of the CIO as the service partner slash order taker versus the business partner. And you are definitely on the business partner model. That is When it comes to budgeting, do you sit with the business and budget for the projects and say, here's what we should be investing? Absolutely. mean, again, we sit with the business. They come to the table saying, we wanted to invest in this. I go, okay, go ahead, go get the money. Then we sit with the business and then we talk about, did you really mean that that's what you wanted to invest? How about we invest in something else? And you do, that's amazing. That is, and that shows true partnership. That's, that's a very unique level. I don't see that all the places I've, we've, we've worked with. and that's amazing. Technical architecture. Yes. Is it a function in your group? It is a function in my group, yes. What do they do? What's their roles do you think? Really good question. So technical architecture in my world, so there are solutions architects specifically aligned with business products. So they understand the business end to end. They understand workflow. They understand what is available on the market. What are the comparative products on the markets from the business standpoint as well from the technology. So those are the solutions architects. So they... The good ones will know that and they will develop the strategy for the several years. In a bank that I work, we do the strategy for a year as well as we operate in the three year cycles. And we design our budgets and everything else in three year cycles. Our next third year starts planning starts in August. So in January, we build the strategy for 26. mean, what are we continuation of from the previous three years? And we started working on our three year cycle and then we will. fine tune it with our business and have all the approvals next year for that. So those are solutions. Then we have domain architects that are more on the technology side, cloud security network architects. So they ensure that our, not just the standards, but our current environment, current architecture is fit for purpose today. And hopefully if we are smart enough and we have the right people, then we are fit for the future as well. And where we are not, we kind of move towards that. through our strategic plans. Now I don't want to paint the perfect picture. Some areas we're good, some areas we're not. It's like in every organization, you have strengths and you have challenges that you are trying to build up. But it's a critical function. What I haven't mentioned is application architecture. That is probably for, in my organization is right now the weakest point because it is within the tower of let's say CIO for commercial bank or the CIO for enterprise functions. Ideally, you have application architecture and application platform across, it's standardized. And in a really mature organization that had a privilege of working, you have a global application standards for the board, infrastructure products, applications, as well as the business applications. And they're all connect and they all integrate and they all work together, long-term and short-term. So we're working towards that. ⁓ is helping and is going to help significantly more in that journey. I'm curious, and if you can share, don't worry, but like number of applications, it's because like people don't, when I, when you think about a bank, might say 30, 40 applications to run the bank. What number of applications do you have within the group? So let's say, let's just take subset commercial bank Americas. We are right now around 600 applications. But just a commercial. Just for commercial bank. And we were about 850, so we decommissioned about 250 over the course of the last three years. do a buy, hold, sell analysis and you say which applications are we going to get rid of and which ones we're going to invest. We had a long conversation with our, actually, you know, I work for MUOG, so with the home office, who is responsible for the life cycle of the application, especially making decision, hold, sell, end of life, end of service. And there was a thought. of some of our leaders that the role belongs in the business. And actually I don't agree. Yeah, I don't agree either. Right. Thank you. So we both know what's going to happen. If business will decide if application stays or goes, contains upgrades and so on and so forth. It's not going to happen. mean, it's not going to happen necessarily right way. So we had a long conversation that felt that business will be able to invest smartly and allocate money, funds, time. Right old fashioned way, UAT testing and all of those fun things where you need business engagement. Subsequently, they need to ⁓ drive this. My point is no, we need to partner and they can influence, they have to be part of the conversation. We always want them to be part of that conversation on ongoing basis, tactically and strategically, but they cannot possibly own. And that's when it comes to the application architecture and application life cycle. Especially right now with the direction we're going with AI, that's becoming more and more. much more elevated decision, much more strategic and consequential than anything else that we probably were doing before. like, so it's domain architects, solution architects and application architects, which I haven't heard the third. So that's something new for us. application platform, right? If we think about it, especially as we get through the AI coding, one of the biggest fears that probably I have and lot of my peers and I'm sure you would have is how to decentralize. ⁓ in AI coding world, it's going to be, if you allow that, compartmentalize, decentralize, because you're going to have anyone and everyone, rightfully so, including citizen development, developing applications. So what guardrails are we going to have? How do we stop? And I don't want to stop anyone. I want to enable everyone, but I just wanted to have a right guardrails around that. And for that, you need application architecture. You need application, standardized application platform. almost like a plug and play, go develop your thing, I'll give you the template format guardrails. Then you can plug in into my AI platform and it's going to just work because you followed all those standards and including security standards, including identity and access management standards for humans as well as for agents. agent should be treated like a human in a company. What entitlements do you have in software? What's your permissions to execute services? What's your audit? trail and tracking. Although that is the same thing as a person. It should be no different. You know, it's interesting. So Dave, like, so when Dave came into On Quark, he said, I'm going to mandate that developers use these coding tools. hope we're the last group ever using those tools because that's what Algold would be. we'll get a long journey to get there. But Dave, just your experience in the coding tools. I'm curious just to give some perspective. What you see, you've used them yourself, you said in your garage way before you joined us even. Oh yeah, mean, I'm an active user. You know, look, I think they're incredible. They provide incredible acceleration. But to the earlier point, where are the guardrails? I think you have everybody creating code much more easily. In some sense, the whole process of writing code has been democratized. The access to the ability to take an intent and generate an application or a function that performs an intent is just now available. you know, almost anybody that you give access. And so the challenge does become all the things we hear about, codes, fraud, code, draft, things getting out of sync, tremendous amount of redundancy, not just redundancy and intent, but in methodology, you know, you're doing it that way, you know, one person gets it done that way, another person has a different intent that gets done a little bit differently. So you now you have a lot of replication, different implementations for no good reason. Look, I think that the interesting thing about architecture is architecture is about constraints. It's about governance. It's about putting rules and constraints on top of the process of coding so that over time, maintenance and management are possible. As an architect, you sit there and one of the first questions you ask is, what changes together and what changes independently? And you define interfaces and modules based on your understanding of how things change. But how things change is not something that comes from the code. It comes from the external environment. And often you have to make bets about how will the technology change? How will the business change? How will the role of humans and technology change as things evolve? How will performance requirements change? You talk about application and technical architecture, right? One deals with function, one deals with non-functional things like performance and scale. So one of the interesting things, you know, when you were going through the various kinds of architects, which I agreed with, one of the interesting things is where do they intersect? Everywhere. Exactly. And so now this notion of how does the business architect role interface or intersect with the solution architecture, with the application architecture, with a technical architecture. And I think the reality, these are big decisions that these folks make, enormous trade-offs. There's a lot of deep knowledge. It's not about the generation of the code. It's about the accountable human who's going to make a decision, who's going to choose some point in that complex space of trade-offs. that is going to determine what your long-term costs are going to be and what your long-term capability is going to be. Those are tough, those are decisions that have to be made insightfully with a sense of accountability and trust. So it's very interesting to think about how AI helps with that. How it might bridge and support the understanding and execution, especially where the intersection of all those architectures are. I think what you described. of fears associated with the change, with the maintenance, with the decision making. It's a hundred percent relevant and costly and impactful to the business in a way we have operated for last 25, 30 years. We thought we're to have a little bit of a reprieve from all those pain points when we went to the cloud and we said it's elastic, it can go up and down sideways and everything else. got little reprieve. We didn't really get that much reprieve from that. I think that from the architectural standpoint and how we think about it, as far as the opportunities in the next, not 24 months, I'm talking about three to five years, I think it's going to drastically change. So change itself are going to be significantly less scary than what I think we are right now in all of our worlds perceive it every single day. Right, this whole notion of the change management, change record, that we can't wait, something breaks, what did we change? That's the first question we asked. That's such a key point. think that's an important thing to realize. That's such an important thing to realize what you just said in the role of AI, right? Architectural decisions are hard and their impact is absolutely dramatic. The wrong architecture can bring you down very easily, can be extremely costly. It gets calcified over time. It gets very difficult to change. Customers become rely on it and it could have fundamental flaws, flaws would slow down the business. So very often what happens, and I've seen this over and over again, is that to make a big architectural change is very hard. And I think that it's even very hard because there's such a commitment. to these architectural decisions, it's very hard even to get people to open their mind about making those changes. And I think that with AI, what happens is you can experiment much more rapidly with different architectures. And the speed at which you can convince people to change, it changes. It's faster. That's dramatic. To me, that's dramatic for the business. But what percent, I got to ask, what percent of error is allowed within a bank? because these technologies are probabilistic. with AI, specifically with AI, let's not just idolize AI. I think this is really, really important because there is a significant risks and we don't even know how to deal with those risks yet, systematically. Correct. We haven't defined them yet. as far as the errors, we are a financial institution, so with the customer, shareholder financial transaction, there is a zero tolerance for the errors. You have the regulators, I've worked with the OCC, you have them all. And they're going to come and say, oh, show me the history of this transaction and why did you approve that? Absolutely. Every single, so when it comes to our industry, there is a zero tolerance for the errors. If we talk about productivity or things like that, yes, of course, humans are not perfect. We're going to have some errors. If we have the right guardrails, we have right error handling, hopefully the right control frameworks, we should be able to catch them before they become. systemic and then subsequently spill into our financial transactions. But the way we think about it is we have so many stakeholders with zero tolerance. That's, I would say it's So it's interesting. you're, the view of empowering the business, which I love, I love the idea of actually let's let them build what they need to build, but within our guardrails, what does the future CIO function look like in your mind, three to five years? Is that, is it an architecture function or what did they? does that look like? I'm just curious if you have a vision of that. As I was thinking about my current environment, we are a manager of the human capital, human intelligence. I think with AI taking significantly stronger stance in our environments, we become the owners and accountable parties for human and artificial intelligence for the enterprise. Wow. I like that. So how do we ensure that both operate properly, human intelligence and artificial intelligence? How do you keep each other accountable for the outcomes for our customers, shareholders, ourselves, society? It's critically important. And I think this role of AI, CIO in the AI world, it's going to be increasingly that. It's governance. It's governance. It's architecture and governance. It's the, as the line of code comes down to the cost of the electricity that produced it, which is what we're hearing and we're seeing and we're all, we ourselves are using. You know, I like to think that there's a measurement of what's the cost of the line of code to maintain once it's created. You know, that's the burden. I'm using the word liability of that code. Do you think that's something down the road people will measure? Absolutely. They're going to be the measurement of the liability of the code. think there is already measurement of the liability of AI that we should think about how to embrace. Liability of AI is from the AI usage today by us, AI usage by our third parties. Yes. And AI usage by the companies that we buy specifically for AI models. So, and I don't think that has been measured enough. I was talking to our chief legal officer yesterday actually about that. Like how do we think about it? Yeah. Let me, let me comment on that because I don't, you know, we want to be careful to get fooled by sort of superficial error rate counts because I think what you're looking at too is what is the type of error, because there's higher risk errors, right? in some sense, with a workflow of agents that can compound and amplify an hallucination, depending on how much control you give them over the enterprise or what tools that you give them access to, you can have incredibly asymmetric risks. So it's not just about error rate, it's what's the nature of the risk that can unfold from that error. We talked about lot of things, but we have not mentioned is the data. Models that we buy or create, right? Very important. Agents that we buy or create. How do we orchestrate those agents? How many do we use autonomous or not? And then it comes to the very, very important part, portion of it is our data. Quality of our data, knowing the lineage of data and subsequently leakage of the data that can happen. And the data that we bring from outside in or those models that we just bought have been drained. I don't think we have thought about it yet enough or formulated what controls as enterprises we need to add to and build. Maybe large banks to have been doing it for the last 10 years, they have it, so I'm not going to speculate on that. But as I talk to the industry, to my peers, I don't think we have yet that level of the comprehensive thinking or tooling. or understanding and to end enterprise level. how to embrace it? Because we don't want to stop innovation. We don't want to slow down innovation. We can't, we can't afford it. I think not single bank or single organization, single company technology or none can actually slow this down. So the question is how do we ensure that we minimize, eliminate, can we eliminate risk? We are in the business of risk taking, right? So we're not going to eliminate the risk. But how do we measure the risk and how do we develop our risk appetite in the new world of the AI in all those stripes of our own agents, our own data, our own models, as well as the external. So we have three stripes that we need to think about. I think it's incredibly exciting time right that's interesting. I think I know the answers. The importance of data models to you and understanding critical. How about reuse in the organization? Because there are... There are technologists that are publicly stating we should eliminate all open source libraries and let's just let the AI write the code and just replicate the code for every application. Just keep replicating code. Why? Good. That's what I figured you. Based on what you said, a BIS is critical. Absolutely. Does an application exist in the future or is it just, is there, there's a view from some that a database and an app will go away because it's just an agent. No, I don't think anything in in a foundation of how we do things is going to go away. Absolutely not. I think what is really going away in the software world is how do we interface with software. So we may not write as many GUIs as we are writing today. We may not think about user experience because there will be no user dealing with the vast majority of those workflows. It's going to be just API-based and agents driving that flow. Just if you think about payments or you think about, I don't know. lending or trade finance or the trading or whatever it is, rarely the humans are going to be there pushing the button to move the process from point A to point B. Like why? How much? I'm just curious, from what you've seen being in the industry, those processes that move from point A to point B, how much were deterministic that could have been automated without AI had the investments been made? How much of them is truly probabilistic that need the agent? We probably could have automated. close to 40%, but it's expensive. expensive. So this rule is very, very expensive. like that. This is the way to reduce the expense of the automation. Absolutely. We don't know how to collect tacit knowledge of the individuals. And we're really hoping this is one of the biggest advantages upfront of AI is going to be. Red BRD is sitting down with business analysts, writing down how the process works and what if scenarios and how to, if yes, then no, can just, we can teach itself. and interact with every single one of those process owners to actually collect that information and then automate the process. And that's going to be truly inexpensive. That is a really good view. think about when RPA came out so many moons ago and years ago, ⁓ the biggest problem I personally saw was not the technology, it was understanding how to automate the process because you have to document it. then you, I remember we were one time deploying it to a team of just five. and all five individuals worked differently. It didn't work. It didn't work, exactly. That was the biggest problem. We didn't get the value from it. So Dave, any last questions from your side? Otherwise, we're get into speed round. Anything from you? Yeah, let's go for the speed round. Let's do it. Okay, so actually, Dave, you asked the first question. I'll go to the next one. We'll do this while ⁓ we're all today. I think the first question is, what is the show you've most recently been watched? What do you think? You know any shows? ⁓ Drops of God, Dockular. I haven't seen it. Do I need to just stop everything and Yeah, you need to stop and there's two seasons. Second season just came out. Start from season one and go to the season one. So it's beautifully done. It's artistic. It's amazing. What is it streaming on? It's streaming on Apple. Okay. I'm going to go watch. I did not watch that. What app do you use most frequently in the day? Message and communication. I think I'm the same there as well. What was the show last time? I think I said Landman was the one I was watching. Fantastic. It was good. Dave, you had a different one. What was the one? Beasts and Mail. I started Landman because of your recommendation. Crops of God is about winemaking, isn't it? Well, there's more than that. Yeah, more than that. I understand, but that's the... There's so much more than that. If you like artistic movies and that cinematography when it's really, really good, then you will like that. Yeah, made my own wine for many years, so at least I would enjoy it. Did you really see? I didn't know that about you. And I've never seen a bottle. So I feel like I'm missing out here. I mean, and the company's called Encore, Dave. We should be getting bottles of this. How appropriate. How appropriate. When we used to pitch customers, pitch investors back in 2017, we'd go to see the investor and they'd look at us and go, Encore, you're... Aren't you wine company? Like what do you mean B2B enterprise software? Like I disappointed hundreds. Maybe some of our swag, some of our swag should be a wine. You know, we did wine tastings events before. We should probably bring that back up again. It'll be a time to do AI wine tasting with Dave's. Dave, is there a name for this wine? Every year I did, I had a different name. So one of them was elemental cognition, you know, one of them was Jeopardy. And so I've had, I've had Every year I named it something different. That's awesome. That's so let's see. Next question. Tech product you wish you had invented. Is there anything that when you look back, you're like, I could have done that. It's something. I wouldn't say I could have done it. I would say I would have loved to be part of it. Smartphone. Smartphone. Yeah. Why? I'm just curious. ⁓ my God. It democratized data. It democratized information. Democratized. payments, it democratized life for people around the world. It became maybe not equalizer, but it came as close as possible when it came to the data, the information, access to the goods. know, what I like about the smartphone answer is it's a controlled runtime, controlled. it's the way it became with millions of apps is because The app sits inside of something which controls it governance. Back to a conversation. Governance, the architecture, the domain, the business. ⁓ That's awesome. I keep thinking back, the future state is architecture is one of the most critical functions both on the business side, but you called it the product, the chief product officer role, which I truly agree with. I think that's the domain architects to understand the cloud, the solution architects to understand what's the right way to solve the business problem. ⁓ an application architects, which is within this world of AI that is going to be absolutely critical. the goal of a global application architecture, which is amazing. This was incredible. I really appreciate you joining us here. And for those, didn't say this before, but Allah is in our, he's our first guest ever sitting in that chair using the microphone with the headphones because you know, it's the first time we're actually doing it in person. This, this is a brand new build office that we built out as a podcast room. It looks cool. It actually looks really cool, but But I just want to thank you again for joining us here. It's been a pleasure and we've learned a lot from you. Thank you so much for having me. This is fantastic. So good luck with your podcast. I will be listening. Awesome. Thank you all. Thanks again. Thanks Dave.