CDOIQ Symposium 2024 | Tony Cyriac | Accenture

theCUBE.net

Clips
News
More from CDOIQ Symposium 2024

Tony Cyriac

Managing Director, Data & AI - North America Banking & Capital Markets

Accenture

Data gravity in the age of generative AI: Navigating the centralization dilemma for enhanced efficiency and innovation

With the exponential growth of data, the pace of innovation and evolving governance needs, the concept of data gravity becomes more pressing than ever.Enterprises face the critical decision of whether to centralize their data or adopt an open format, particularly as generative artificial intelligence continues to advance. Choosing the right data gravity center is crucial in navigating this complex landscape, according to Tony Cyriac (pictured), managing director of data and AI – North America banking and capital markets at Accenture PLC.Accenture’s Tony Cyriac talks to theCUBE about why the center of gravity of data should be top of mind

play_circle_outline Introduction to CDOIQ 2024 symposium in Cambridge

play_circle_outline Exploring the Evolution of Data and AI in Banking with Accenture's Tony Cyriac

play_circle_outline Importance of picking a center of gravity for data in organizations

play_circle_outline Challenges and opportunities for CDOs in driving value, people management, and staying updated on data trends.

play_circle_outline Maximizing Success through Stakeholder Management and Cross-Team Collaboration in Data, Analytics, AI, and Technology

Info
Transcript

Tony Cyriac | Accenture

Tony Cyriac

Managing Director, Data & AI - North America Banking & Capital Markets Accenture

CDOIQ 2024 in Cambridge, Massachusetts, marks the 18th annual Chief Data Officer symposium. Tony Cyriac, managing director for data and AI in North America for banking and capital markets at Accenture, shares insights into the state of data in July 2024. The focus in the financial services market has evolved from structured data to unstructured data, unlocking untapped value and enabling new GenAI use cases.

One such use case is enhancing internal efficiency, like supporting contact center agents with copilot capabilities. Another emerging trend is ... Read more

explore Keep Exploring

What event is being covered by theCUBE for the eighth time in Cambridge, Massachusetts? add

What progress has been made in the banking and financial services market in terms of data management over the past 10 years? add

What challenges do many companies, including banks and financial services, face in terms of integrating internally built systems with SaaS vendors and GenAI capabilities? add

What are three key points to focus on when working as a CDAO (Chief Data and Analytics Officer)? add

What are the key roles that every CDO needs to focus on? add

bolt Powered by CUBE AI

Tony Cyriac | Accenture

search

>> Hi, everybody. Welcome back to CDOIQ 2024 here in Cambridge, Massachusetts, the 18th annual Chief Data Officer symposium. This is the eighth time theCUBE has been covering this event. I'm here with Sanjeev Mohan, my co-host all week. Paul Gillin is also here. We're really pleased to have Tony Cyriac, who's the managing director for data and AI in North America for banking and capital markets at Accenture, longtime practitioner in the financial services industry. Tony, welcome to theCUBE. Thanks for coming on.

>> Thank you, Dave. Happy to be here.

>> So give the audience a little background on your roles, your role currently, but also some of your previous practitioner roles. You have quite a resume.

>> Thank you, Dave. I would say I just joined Accenture three months back, so still learning how this all works, but amazing experience and learning so far, leading a lot of the data and AI conversations and actual delivery with large clients across North America. Prior to Accenture, I was the chief data and analytics officer at Charles Schwab, and great experience there in leading our data analytics and AI capabilities there. Also, I was with BMO, one of the large Canadian banks, for about five years as the chief data and analytics officer there as well. Lots of scars and experiences from work there as well. Overall, always been in data analytics and AI for financial services. That's background. Hoping to bring some of that to light here today.

>> So how has the onboarding process been at Accenture? Was it all AI based, just GenAI tell you everything to do and just talk to the machine, boom, seamless?

>> Yeah, I don't think it's quite there yet, but I should still say, for example, one of the things we had was something called Accenture Park, which was essentially a omniverse based capability.

>> Very cool.

>> We would actually then go and learn various aspects of the company as we tour around.

>> So physically tour or have 3D courses?

>> Have the tour inside the Metaverse, if you may.

>> Metaverse.

>> How was that? What was it like? What was the experience like as a new employee?

>> It was amazing because I think it gets you to the right topics of areas of interest. Learn a little bit more, learn some examples, learn how others interact with it and even kind of huddle in rooms if you may, and talk to each other.

>> Did you pop into Julie Sweet's office and say, "Hi, I'm the new guy in town?"

>> Remember there was this big controversy, Zuckerberg came out with his new glasses and then was it Elon Musk or some you was trashing it and said... No, Apple.

>> No. Zuckerberg came out and-

>> Zuckerberg came out, yes....

>> de-positioned Apple's Vision Pro.

>> De-positioned Apple's. Yes. What did you have, Apple or-

>> We have various options. So the Oculus one can be used, but also it's still on the laptop also available in case you're not wearing this, you could actually do that. So it varies by the batch and varies by what you are available at that moment as well. It's like that.

>> It's very cool. It's good to see a leading company like Accenture actually applying these technologies in such an innovative use case. So at any rate, let's get into data. How would you describe in part what you're hearing today, but the state of data in July of 2024? Where have we come from and where are we going?

>> I would maybe put a little bit lens on the banking and financial services market. I would say it's a journey that never ends in some ways, but it's a huge progress from where banks were maybe even 10 years back. I still remember talking about it 10 years back when after the financial crisis, everybody had so much of regulatory focus and trying to get things right in the data, recognizing the commonality of data, how to organize it, et cetera. So a lot of focus on the structured data, the governance of it and the reusability of it for multiple outcomes and then how do you drive value with that was the next swing into that. So it's gone well in many places. There are obviously missteps and corrections and you go back in time and some of these concepts keep evolving, but at the heart of it was what's the critical data, how do you manage it, how do you organize it for various outcomes in the company? And that still is the mantra, but what's changed in the last year and a half is with this focus on unstructured data with many of the GenAI use cases demanding that now there's a new light shown on that topic and there is untapped value there clearly, and there are new methods of now accessing it and using it in ways that was not possible in the past. So there's a lot of focus into that and how do I integrate with the rest of the structured data work that has happened and how do I get myself ready for more and more GenAI powered use cases. That's a lot of focus, a lot of work that we are seeing as well.

>> So what are some interesting use cases other than the obvious ones, summarization and translation, what are some in financial services use cases you've seen of GenAI?

>> So one is internal use cases of efficiency improvements and the most easiest one everybody gets to is the contact center agent. How do I support with the copilot capability? So that's a big one where real value is visible in a short timeframe and a simple example is somebody calls or call center and says, "Hey, I am moving from Illinois to Texas. What happens to my kids' college 529 plans?" Now, it's a simple question, but it's complex to answer because of the number of procedures and documents you got to look at from. So GenAI when applied the right way quickly gets you the answer. So a complex being answered in less than 30 seconds is possible and that's really productive and valuable. It's a client experience point, it's a productivity point. So that's an obvious one. There are many such co-pilots if you can think about within helping an RM, relationship manager, helping a wealth advisor, or helping let's say an AML investigator and so on so forth that you could actually do in the spectrum. And that's an easy one, no-brainer. But it takes readiness of data as kind of the key to this and then obviously having the right expertise and the talent to do so. But okay.

>> No. Please carry on.

>> The last point I would just hit is I think there are also client facing use cases we are starting to see emerging, which everybody was a little hesitant. But there are at least clients now I'm working with who are thinking about having capability to interface with clients where it could be a digital avatar if you may, where you can actually have a FAQ questions on typical things that banks offer and answer it, et cetera. So that's a new area, higher risk, but we are seeing some of that emerge in the market.

>> And that's exactly what I was going to ask you was the consumer facing, because if you look at things like credit card fraud, we've seen a market improvement as a consumer going from months to days to minutes and instantaneously. Okay. Certainly haven't seen that yet from a contact center perspective, at least from a consumer. You call your credit card company and try to get a resolution. There's no evidence today anyway, at least from my suppliers, that the GenAI is supporting those use cases. So my question is where are we in terms of early phases here, obviously?

>> Yeah, I think it's early phases. Last year I think everybody was in a lot of proof of concept and pilot and test out, et cetera, but I think things are moving into production this year. I'm seeing increasing number of clients putting such early use cases into production. And most importantly, there is also a conversation, "How do I do this at scale?" So that conversation has come up a lot this year where they want to say, "Okay, one use case here, another use case, they're great, but tell me how do I do this at scale? How do I reimagine or reinvent my business if this is truly something powerful to take advantage of?" So that conversation is where we are helping a lot in terms of thinking through what are the biggest outcomes you focus on? What are the type of adoption pathways? Do you just let a thousand flowers bloom all around the company or do it in a more governed and organized way? What type of data readiness exists and to do this at scale? Talent and readiness on talent aspect. Responsible AI aspect, how do I make sure that this is not creating new risks or it cannot be ungoverned and how does it connect into the existing risk management frameworks of the company? So all of those are conversations in the scaling agenda that we are seeing a lot as well. So lots of VOCs last year, but we see real production happening this year and a lot of focus in a-

>> In a phased manner.

>> In a phased manner.

>> Like a few financial services agents first, then some more-

>> Correct....

>> and then. I see.

>> It's a phased approach, but now with the eye towards what is the foundation I need to do 10 more of these, without breaking anything that is running in the company. So that's the shift we are seeing this year. Again, I should say early days still in that sense, the kind of potential that's possible versus where we are, but there is a clear action and there is real outcomes starting to be visible this year.

>> So you've been a practitioner, do you have a view of what is the progression of personalizing AI? What I mean by that is it still let's do rag or banks starting to fine tune or even train their own models just on their own data not all this noise?

>> I think of it as three buckets and we are seeing all three buckets. So bucket one is, let's just call it support, which is essentially, hey, things that are trained on world data available to everyone. It's a little bit of a commodity. It's in the teams meeting where you could summarize the notes and send the action item. Well, it's a no-brainer. You should take it, switch it on and you get it. It's not a competitive differentiator, but it helps you. The second bucket is where you go into the next level, which is to say, "How do I embed this and connect to my internal data and contextualize what the LLM is giving you better?" And this is where we see a lot of actions. So the contact center agent copilot is an example where we will take the LLM clearly exists as one of the components of the solution, but there's a whole application ahead of it to wire this to our internal data, make sure that we search and get the right content from it. And when the question is attached to this with a set of guardrails, the LLM now gives you a better answer, less hallucination, et cetera. So it's a rag idea but implemented in that. The last bucket is essentially what I call the transform bucket where you have truly unique data sets in the company that nobody else has and can you now maybe apply this and fine tune a model or even build your own model eventually to offer as a service that is monetizable to various products and services in the market. So that's the spectrum. So we see things happening in one, two and three. All three buckets happening at this time. It just varies by which institution, how large they are. The last bucket is mostly done by large institutions who are starting to go into different sections.

>> Compute intensive, you need GPUs-

>> There's a cost aspect as well.

>> . Right.

>> I want to ask you to put your practitioner hat on and Sanjeev and I have... Every time, we are together at a conference and we sit down with a practitioner like we did at the Snowflake summit. Remember breakfast? And we started peppering them with questions like, so irrespective of the technology, it could be AWS or Google or Microsoft or Databricks or Snowflake, probably all of them. Which of the following best describes your attitude? Remember you're a practitioner now, thinking about all the practitioner experience you've had, all the peers that you've known. Okay, try to consolidate them into one brain. Would you prefer to put all your data assets into a sort of single platform or would you prefer to have more of an open format, understanding it's going to be less mature. Given where we are today, what would be your strategy? Either put all your eggs in one basket, go other end of the spectrum, completely kind of open table formats or some kind of hybrid. How would you approach that question?

>> So I think it's an important question and many companies struggle with it, and I'm going to just add one more complexity to that question, which is that now what happens is I think most banks and financial services, they have lots of internally built systems. There are also a lot of SaaS vendors like Salesforce and Snowflake, and we have things like Workday and other things that are sitting. And everybody is also turning on GenAI capabilities within that. Now what does that mean? While they used to do a departmental function in the past, now to turn on GenAI and be effective, they need the full context of many things far more than in the past, not just what was required in the sales or marketing or whatever. They need the connected context being Q1 and everybody's vying for what I call where is the center of gravity of data for the organization. So this is a landscape we're working through. So there's a big question for organizations to decide and say, "What's my center of gravity for data? Do I need one?" And center of gravity doesn't mean you put all the data there and it's there, but you at least need to know that there is a reasonable amount of context being held together across departments, across lines of businesses that you can give to each of these applications and you can also get context back to add to this unified context. You just need to do that somehow. Now coming to your question. Now do I put the data in one place to achieve that or not? I think that's a little bit of a physical architecture debate. You could get to ideas of data products being managed by various areas yet have this context connected and held. Theoretically, that's possible. Practically, I think you still have to load data together into a place for many of the use cases where latency and other things matter. You have lots of things that are still done locally, yet there's a way to connect them. But there is still something that you got to do. In terms of a cloud, you got to pick some primary cloud for your data. If you scatter it all over, it's harder to stitch it together.

>> In current reality.

>> Technologies will improve and change. So I just come back to the simple concept that you got to pick your central gravity for data in some place. You got to have a way to have context that is connected and that is shareable and you can enrich it on a regular basis. How you physically implement it? It's a lot about, you have to think through the use cases, think through where the company is and think through the evolution of some of these technologies that are happening every day. And the GenAI puts a complete new spin to some of these topics as well.

>> Sanjeev, I'm coming to the conclusion that like many new waves and new innovations that it's just going to get harder and harder and harder with the exponential growth of data and the pace of innovation and add in the mandate for governance. I think it's a boon for Accenture and other companies like you.

>> I know analysts and for people like us .

>> And analysts, absolutely.

>> You need analysts to be mystified by all this because we've been talking quite a bit at all these conferences, but there's a whole new space that I don't think we've even touched upon, which is you take this data from unstructured data, unstructured documents, create vector embeddings, you store it in Parquet file and put duct DB on top. Instead of even taking it to Databricks or Snowflake. That's a whole unexplored-

>> New architecture, right? Very simple.

>> Yes. Very simple, very cost-effective. So we get into this... If you don't mind, I won't change tracks here. Because I've known Tony for many years and I have learned a lot of what I talk about. Like CDO stuff, I've learned from him because he's been very willing to share his knowledge. You've been a CDO for many, many years. What is the most important thing a CDO has to do and what are some pieces of learnings that you can share?

>> Yeah, thank you, Sanjeev. And yeah, I've almost over a decade now of having worked as a CDAO in the industry. So lots of learnings from that. And if I were to distill maybe three points out of it and say, number one, focus on value. If you can't articulate the business value and outcomes, all of this work is a means to drive and accelerate client and business value for companies. If you're not focused on it and you can't articulate and connect it, it's a problem. So focus on that, lead with value, and then work backwards everything you're doing to attach that. And it could be client value, it could be business value, but you are there to accelerate those in companies. That's idea one. Idea two is this is a people game of the highest order, which means the CDAO role is effectively about bringing the talent together in your own team, but also bringing the stakeholders together on a journey that they share in this and are part of. And you are actually working with them closely to bring them along the journey on the problem statement, on the type of issues you are solving for them and so on and so forth. It's a huge mandate. Focus on it. Don't underestimate it. And the last point is you just need to keep track with the moving motions on data and AI that keeps happening every day. You just need to know these things and the trends and have a point of view ahead. Otherwise, you just can't guide the company on which direction to go and what is right for the company in the current moment. The earlier question on where do I put my data? I think you have to be the leader answering that.

>> Number three is a full-time job. So you have three full-time jobs.

>> Three full-time jobs. No, managing stakeholders is a big job by itself.

>> Yes, it's a very tough job.

>> Very tough job. So it's three things you do in simple terms. Now we can put many, many more details to it, but I think those are the three I would say every CDO need to focus on.

>> Is the CDOs role at danger if they don't have CDAO analytics?

>> Great question.

>> I think it's not exactly a danger point. It is just about the organization. And the point is in the organization, how well are you working towards outcomes in a connected way between the data person, the analytics person, the AI person, the technology people who are supporting this? They just have to work together, increasingly the digital team too.

>> And CSO.

>> And CSO. So you have to have this group work together. Now one person can run many pieces of it. One person may be just doing three of it and the rest is done by somebody else. Well, it's an organizational construct. The question is do you have the buy-in at the leadership level for the mandate that you're trying to do, and is there a working chemistry between these roles? And is there a way for a shared roadmap that all of you can cumulatively show the outcome? If you don't have a shared roadmap, the CDO is doing whatever he or she's doing, the analytics person is doing what he or she is doing, it's not helpful. So you just need to have in the context of the business, what's the shared roadmap and then you're all pitching towards that. There's a working method, there's a leadership support, it works. If none of this is true, then yes. The answer is yes to your question. It's a problem.

>> And there's a new emerging role of chief data in AI officers. So that adds more confusion.

>> I really like what he said because actually Tom Davenport was with us earlier and he was like, "The CDO must have analytics." But what you're saying is as long as there's good chemistry and organizational support-

>> And the common shared roadmap.

>> And the common shared roadmap.

>> Then yeah.

>> You can make it work. The CDO doesn't need to own the ocean.

>> Yeah, it makes sense. I mean that shared roadmap sometimes is hard to achieve.

>> Yeah, but it's hard work. It's hard work. And it varies by institution. I did it at multiple institutions. The role is different at different places.

>> Guys, we're out of time. Anything else that we missed?

>> No. I'm good.

>> Okay, good. I'm glad you got it in. Tony, thanks so much for taking the time, coming to theCUBE.

>> Thank you, Dave. Thank you, Sanjeev.

>> Thank you so much. This was amazing.

>> All right, keep it right there. We're wrapping up day two CDOIQ in Cambridge, Massachusetts. You're watching theCUBE.