Hello and welcome to the ES Speaks podcast where leaders unpack technology shaping business strategy. I'm Corey Nolles and today we're tackling a big one. Enabling a Gentic AI. Why data strategy is now business strategy. Agentic AI or systems that can act autonomously to pursue goals promises huge opportunity but only for organizations that can get the data fundamentals right. Our guest today is Abby Visu Vasam, field CTO of enterprise architecture and solutions at Reto.
Abby helps enterprises bridge architecture, data strategy, and product vision to safely unlock advanced AI. Over the next bit, we'll demystify agentic AI a bit and dig into the business risks of weak data foundations and map the concrete moves leaders have to make today to win tomorrow. Obby, how are you, sir? Uh very good, thank you and really excited about this. Thank you for having me on and uh I think this is going to be a great conversation.
I think everybody's talking about this topic and it would be helpful to have uh another point of view on this. It is very much uh the topic of the week right now is a lot of talk around data. You know, I guess if we're going to demystify agentic AI a bit, I suppose we should start by defining it. What do you mean by agentic AI and how is it different from the AI most businesses are using today? Yeah, this is this is been a conversation that's been going on for a while now, you know, because everybody wants to define it a way that feels comfortable to them.
And that is, you know, how is what is what does agentic mean? Is it autonomous? Is there a human in the loop? And is it still considered a gentic at that point, right? And how are you building these things to to make it so that it's like a human that's operating on its own using a level of intelligence and making some decisions that you as a human would also make along the same path. And so this this concept of where do we draw the lines and say what is a gentic?
And I think it's all of it. Okay. Um I think it's a huge category. It's it's it's a category that's being defined right now. And I think at the moment it makes sense to throw everything in there and not to start weeding through it and start isolating one or the other. But I would say mainly it's about being able to make complex decisions in real time or near real time. The ability to do that. It's the ability for the process. I'm just going to call it a process for now.
This capability to learn continuously. Whether it's learning while you're engaging with it and continuously driving it or it's autonomously running and learning, it's doing a level of learning so that it's refining its output and refining its insights and what it's able to process. And last but not least, I think just like a human, I think it's able to walk uh work across multiple functions. it's able to make intelligent decisions not just about one specific area but bring in other factors like we would do as humans in in terms of making a decision and doing that.
So that's where that's where I see um what what I define as a gentic AI and what I think we're trying to solve in the business. Okay. So why should cos and CIOS care about this right now? What's the what's the business upside if it's done right? Yeah, I mean there's there's a lot of people out there that I think are focused very heavily on things like productivity and uh cost efficiency and being able to take away some of the grunt work out of certain roles. um there is a certain level of uh focus on looking at things and saying hey listen I've got a team of data stewards that are doing some some work for an organization and they are bogged down with for example just reviewing the quality of data in one system or targeting something in one platform or looking at one uh category of data and working on that and there's just so much inconsistency so much so many quality issues and things like that in that one area that they're just focused and and there's so much more happening in the organization that they should be paying attention to, they should be working on.
How do we scale these people? Does is the answer to just get more people? It it probably is not. It hasn't been for a long time. For a long time, we're just finding different ways of automating these workflows, giving them more insight so they can work work more efficiently. I think what we're realizing now is that with AI, we're able to get some of this uh productivity and cost efficiency in their game. What if instead of them having to go there and do the work a heavy lifting uh work that they would do to assess the data quality of a specific uh provider or something in healthcare and make sure it's the right data and there's they're comparing it with another provider record and making sure it's it's accurate that level of work they're doing that what if we were to provide them with a report automatically that's done that comparison and now they're just checking off and saying yeah that looks right that looks right that looks right so that's where I I think we're wanting to go in terms of efficiency um speed at which you can make decisions.
You know we can process probably 25 30 40 50 100 records faster than somebody can probably just individually look at one two or three. So efficiency in from that point of view is also a huge gain and at the end of the day we have humans in the loop and when you have humans in the loop there is a possibility of making some mistakes and if you get these AI um agents to be trained to a certain level and to bring them up to a certain level of quality you can be assured that you're actually going to get some insights that are really valuable and quite accurate for that's exactly right.
You know a lot of leaders worry about the hype around it. So what are what are some of the real business risks when organizations try to deploy advanced AI on a shaky data foundation? Yeah, I don't know if you if you recently saw this report that just came out from MIT about AI, you know, and and it's a lot of information. It's packed with information about there and obviously the big the big the line that everybody's talking about is 95% of the AI projects fail, etc., etc.
What what is it? And when you look at across the board what it comes down to and I'll I'll start with what it is uh at a very fundamental level it's trust okay everybody's willing to trust AI especially autonomous AI in a very small scale where a human can look at it measure its output and say yes I trust that this is perfectly great this is this is wonderful and it's exactly how I would have done it so let's go forward with it and let's implement it okay at the same scale let's not grow it any bigger than this because we don't the other factors that kind of play into this thing.
At a very fundamental level, trust seems to be the reason why these projects don't go from the smaller scale to the broader adopted production scale that people need in an organization to really move the needle on all the things like you know productivity, efficiency, risk management and all of those things. So the going through this explanation what I'm trying to get at really is that that trust comes from having a really solid data foundation that you can trust that people are already in the organization have trusted and said yes this is the data that we build our business on our core entities our core uh business um objects that we want to look at and say hey this is what we run our business on and we trust this data set if you can get to that point so one of the things I think that we've found really useful is having a solid data foundation, especially using a platform like an MDM platform like Raltto and hunkering down on having unified data.
First of all, it's brought in from multiple sources and unified and then it's it's cleansed and created into this trusted golden copy that everybody in the organization agrees on. Now, you build on top of that with something like an AI initiative. There's a level of trust that you automatically inherit and then you can go from there. And then from there you can then start to build in some other additional you know unique capabilities differentiating capabilities that you want to do but you can always ground yourself in the fact that you've got this agreed upon trusted foundation.
Yeah. Yeah. And you know um you've argued that u data readiness can be a competitive advantage and I know that a lot of companies aren't there yet. What is what does data readiness actually look like in an enterprise that wants agentic AI? Well, you know, data readiness is something that I think u our customers are are asking us for and and and they're saying that they they need it. But one of the things that we're able to tell some of our customers, the ones who have already invested in our platform and and are very um um adopted it across across their business is that the beauty of what Relio has done in its platform is that it it gets you data ready for this AI world already.
Okay. So you know there's there are people out there saying hey I want to be data ready I want to be data ready. I think the opportunity is that if you already have a platform like RTO unifying your data you are data ready. You are on your journey to being data ready. But what do you need on top of that? What you need on top of that is the ability to for that platform to be seamlessly integrated with what you want to do with AI. Okay. It's one thing to be one thing to have good data.
Okay. But then you need all these integration points ability to go and reach into this system and pull out the data that you need to then create your agents and create your uh um your your differentiating um characteristics in your agentic initiatives. So I think the key here with respect to data readiness is you need that trusted data platform provides you that especially if you have invested in a platform like Relto. uh you need connected data and that is data from various parts of your organization connected and built together in this you know this wonderful graph that we have and uh I absolutely love the concept of how we have invested so heavily in building out this data graph that has naturally speaks to AI. it just naturally speaks to AI because AI in general is is is something that requires context requires requires data to be stitched together so that it can then look at this build those use those relationships and draw more insights out of it.
So, I love the fact that we have this um graph-based context that's giving us all this uh rich data. And then, of course, you got to make this data accessible. Um anything that you're doing with AI, especially things like chat bots and all of these things, you know, there's certain level of human interaction and humans don't like to wait a long time. So, you want to make this data accessible. You want to make it as real time and as available as possible for that kind of millisecond type of interactions.
So across the board if you looked at this um a platform like Reelio will get you that data readiness right out of that box and then from there the you know the the sky's is the limit in what you can actually do with your AI initiatives. That's right. You got to got to kind of have your house in order first and uh uh that's been a struggle I know for for a lot of companies along the way and it's a topic that keeps coming up. So I think it's it's really cool that there's this engine that kind of handles that for you.
Um, so in this space like governance and trust are huge obviously. How do you design governance that enables agentic behaviors without stifling innovation? Yeah, governance is is as everybody knows in in the data world uh governance is is is a ne necessary evil and and that's the way people perceive it, right? They they perceive it as being something heavy and something um that kind of sits on top of all this fun work that we want to do. The buzz kill.
Yeah. Exactly. For a for a more accurate term. Yeah. Absolutely. Buzz kill. That's the way people kind of perceive uh governance and and and in this in this AI world and where we're going with data and that needs to be trusted and data that's going to be ultimately at some point in the future near future we're seeing um agents being autonomous, right? And when and something is running autonomously and it's starting to just work on its own and it's running hundreds thousands hundreds of thousands of iterations on data um you better be sure that it's coming to the right conclusions.
Yes, you better be sure it's it's it's making those decisions and especially if those if those insights that are being at that scale are being then you know disseminated to customers or you know partners or suppliers or whatever it is that that's consuming this data, you have to be sure that it's correct. So governance is now coming back into into vogue. It's becoming fashionable to be to have governance and there's multiple layers of governance now that people are actually actually wanting.
So one of the things first things that governance teams are looking for is what data are you operating on? What is the quality of the data that you're operating on? Are we elevating this quality of data? And when you look at something like RealTo, it immediately exposes your organization's quality of data. Okay, in a good way. Okay, this is absolutely good thing. When you start to pull in data from all these different sources and you're starting to bring it into into Reto and you're starting to uh work with that data in the single in the single um um platform immediately you start to see the quality of data coming in from various sources your governance practice levels up because you have insight and you have visibility into something that for a long time was either anecdotal or for a long time was just managed in silos.
So you have some of this now once you brought it here. Governance requires transparency. Governance says hey let's open up this box and let's show us show us what the data looks like so that we can operate on it. Relio allows you to kind of do that right out of the uh right out of the gate. And at the end of the day transparency and governance will ultimately lead to what is the most important thing in your organization that is ultimately having trust.
When people can see what's going on transparently and that they know it's governed properly and it's managed properly end to end, they will start to build that trust. And building that trust, again, we talked about it, is the foundation for that agentic AI initiative to start to take. And it's not just this this first wave of what you're trying to build. You're building the foundation for the path forward for years. You know, this feels like such a big important task that companies should be prioritizing.
Yeah. And and I think if you remember the history of how you know legacy platforms where legacy uh platforms and concepts of just MDM was just stuck in the back office. Yeah. It was it was pushed back there and it was a few a few u hardcore data data engineers and people that love data for the sake of being data engineers you know they were working on this thing and they were it was a hot project for them. Of course, we know that, you know, life sciences and the the financial services industry and banking and all of this that are insurance included that are all highly regulated.
They anchored on on master data and they started to use that very effectively because they needed that. They needed that for compliance. They needed that for managing risk. They needed that for all kinds of audit purposes, right? So, this was it was such a back office thing. But Relio all the time we in Relio we always believe that listen as much as you want to push it to the back office and say that's some department out there that's working on MDM.
We think having trusted data applies to the whole organization and it's going to come to the forefront at some point when people start to prioritize data. Yeah. Literally right now with AI and especially with a Gentic AI, this is being pushed to the front of the conversation, the front of every discussion and being said, do you have this in place? If you have this in place, you can do AI. If you don't, good luck is where we're we're at right now.
And and it's proving itself out for in in a lot of cases. So having this foundation is becoming very very valuable and companies are really starting to focus on that and starting to realize that um that time invested in this foundation has not gone wasted. It's accelerated everything. Absolutely. You know this is this is the time where if if you're not if your data is not in good shape, this is where you pause and you deal with that uh and move forward from there.
I totally agree with you. So from from an architecture perspective, what are some common antiatterns that trip up teams building agentic systems and how do you fix those? Yeah. Um, you know, we we've thought about this for a while. Uh, again, the agentic systems and what's tripping them up is nothing in in in essence nothing new. It's the same thing that's been tripping up organizations for for years. So we've been solving this problem for years and we've been providing tools and capabilities on our platform to overcome this for a very long time.
So as long as you are, you know, you've anchored on a platform like Reelto and you're building that deep data foundation, uh, bringing in data, unifying it, mastering it, cleansing it, and keeping it in a platform like this, you're actually solving some of those issues already. And and and some of those issues are, you know, an organization that doesn't have this data foundation is struggling with fragmented data. You got bits and pieces of the truth.
Okay, it's the whole story is everything connected together would be the would be the truth, but you've got bits and pieces of it all over the place. And fundamentally, you are then put in a in a position to have to figure out how to piece them together. Okay? And of course, there's many ways to do it. And there's right and wrong ways to do it. There's also the the risk of not knowing if you have um data you can trust. Just because a a certain business area has been using a certain set of data for a long time and operating on it doesn't mean the rest of the business perceives that data the same way.
You know, we we've gone to we've gone to customers and and they've had long debates about what the word customer means to that customer. They they you know, you'll be in a big meeting, you'll have plenty of stakeholders there, one person will define it one way and the rest of the team will just go off and start to debate it. And a place like Reelio when you're bringing in all this customer data into this one location and mastering it gives you an opportunity to kind of unify that and have that agreed upon uh agreed upon you know concept of what um a customer is.
So first of all for fragmented data silos and things like that you know mitigating that will definitely help uh help help with your u agentic initiatives. U making sure you have good clean unified data. Um, making sure your data is actually connected. That's another thing. Um, this is something that I think we uh is is so downplayed uh and I think the value of this is is uh immeasurable. When you have various bits of data that's connected in an organization and brought into a platform like RTO, there are relationships that are formed.
Yeah, these relationships are very valuable. When you have these kind of relationships and you especially have data describing these relationships, that's where the magic happens. If you just have a customer in isolation and a product in isolation, it means nothing to to to a business until you can connect them and say this customer bought this product on this particular date and any other attributes you have about it and this is how much use they've gotten out of it or this is what they're doing with it.
Right? That's when you really start to build. So, Rotoio, another thing about data is having connected data, being able to connect the dots between these things. Again, it's something that comes very naturally with our platform. And I find that um it just sets up the sets the stage for um doing anything agentic trying to unlock value from that so much easier. Oh, I bet it is. Yeah. And and having the right plumbing, for lack of a better word, in place to make sure that everything can communicate.
Uh that's that's a huge issue right now. Can you share a short example maybe or a use case where getting data strategy right unlocked some tangible agentic AI capability and maybe what changed? Yeah, I mean where we're going with this. So first of all across the board across all our all our customers and all our uh prospects out there um Aentic AI is at the cutting edge of what they want to do. Okay. So at any given point, it's it's rare that you find a customer that's completely automated uh their an agentic process and they're completely in production and they're moving on.
As a matter of fact, I think MIT clarified that for us there's 5% of customers out there that are actually in that production production scale. Okay. So having said that, what we're doing is we are helping our customers build these agents. Okay. We're just taking it to the next level. So I want to talk a little bit about exactly what how we're kind of playing into this into this in in this arena and not just giving them the data foundation. We're not just giving them data and say hey you leave you to your devices to go solve this thing.
That would that would be unfair. Uh and we realized that's not where the value is the value is actually helping them unlock the value from the data foundation that they have set. So we built this we built multiple AI capabilities right now and we're already in in production and it's GA and one of our our key products and I think that's this is just such a differentiator for us is agent flow. Yeah. And agent flow is our conversational AI interface that works with Reto and gives you the ability to use natural language to get insights and engage with our platform our intelligent data graph that we have built.
And so if you have the data that's built into Relia and you've got brought bring all of it in here and you start to u uh unify the data in here. You've got it in our platform, you can get into agent flow and start to build agents right there, right on it. You can also start to engage with the data in in natural language and it gives you a platform that can also be connected to other agentic platforms to start bringing all of that together and create this like agentto agent type uh engage engaged operations.
So what we're doing is since we have this platform out there we've got the data foundation of course we're also on a on a huge initiative to build a 100 agents in 100 days. So we're starting yeah and that's that's a bold initiative. It is, but we're well on our way and we're it's something that we're actively um actively doing right now. We're also working with our partners. We're also working with customers. We're doing uh doing things like doing uh forward deployed engineers, putting them out there in customers and listening to their pain points of what these personas are experiencing and starting to say, hey, that's an agent and this is how you can do it and starting to work handinhand with them and and and deploy these.
So at the end of the day, what we expect is to have some really a marketplace of agents that customers can look at and go and see themselves in and say, "Hey, I'm a life sciences company. My data stewards are facing this problem. This is an agent that's from the marketplace that I can readily just move over into agent flow and execute, get the insights that I need and the actions that I can follow." That's where we're going. That makes so much sense. sort of a templatization of them a little bit where you can hop in and pop in your own credentials which it can probably do on its own and uh absolutely and you know we we have customers that I know customers want to be unique they are unique they're unique in what they're selling they always want to differentiate themselves and do all these things but the data problems we're solving aren't that unique in that sense it's very common across industry right so these these agents that we're working on um in any form in some form or the other are applicable ac across industry.
Anybody can consume them. So we by creating this marketplace and creating this opportunity for people to to do this we think we think they're going to get instant productivity and that's our goal. Our goal is instant productivity and and and I'll add one more thing that I I want to talk about uh just briefly is I think adoption is something that is the biggest issue going on with with the agentic experiments and initiatives companies are working on.
They're great with an experiment but they want to now adopt it and and and move it along. That's where they're starting to struggle. We are heavily focused on adoption. We're not just about proving an an agentic flow works. We can prove that. A lot of people could prove that. Yeah, we want to talk about adoption. We want to talk about build putting it into production and at scale and getting it operational. That's that's where we want to be focused.
That makes sense and that's that's where you need to be. How uh building on that, how should executive teams measure progress and success when they are investing in data foundations for agentic AI? What KPIs really matter? Yeah, I mean from a KPI point of view, there's there's a lot around, you know, um agents, how quickly are you getting this this agent out to market or how quickly are you able to solve some kind of a uh problem in volume? How many of these problems are you able to solve?
For example, if it's a data stewarding agent resolving some kind of a um uh a specific specific scenario in in stewardship, maybe it's enriching a profile and they all have to go and do a web search really quickly and enrich a profile. The KPI would be how long does it take a data steward to do it? How much how costly is it for them to do it manually? And then if you're able to do it with an agentic flow at you know at a rate of 50 a minute or something to that effect where it's automatically doing this research for you then you're saving some money.
So those are the standard KPIs that you can use to to measure. I think one of the most important KPIs and we just talked about it is the adoption KPI. the the time you the time and money you invest to experiment and build out an agent capability for you and for you to get it in production at scale working for you and then resulting in that business outcome that you wanted that actual tangible business outcome in terms of saving money whether it was you know in resources whether it was in time that operational efficiency that you gain productivity they gain that's the measurement they should be focused on.
Yeah. and and all of these will get you to that point. And it's great to measure how quickly you developed an agent. It's great to measure how efficient the agent is and all of that. But at the end of the day, I think they should still focus on the most important thing is the business outcomes and what they've always been focused on. I agree. I agree. The truth is it's if if you're not seeing the differences there, it's not working yet. Absolutely.
Yeah. Exactly. So for leaders ready to move and hit the ground running, what is a pragmatic 90day plan to start turning data strategy into a business strategy for agentic AI? Yeah, I I thought about I was I was thinking about this like what would I tell leaders in general? you know in 90 days what can you do and how quickly can you get there and 90 days is is a long time believe it or not in our industry right now because in 90 days I guarantee you AI strategy and AI thought process and AI technologies would have evolved quite a lot in 90 days but you know jokingly I can very easily say in 90 days you can start from scratch and you can have relio implemented and working in your organization we we have a 90-day methodology we can get you on boarded and get you with this data readiness platform so that's that's part of our our strategy but you Uh putting that aside, I think what you really need to start on is understand where you are as a business.
Do you have the data foundation and do you have the technology in place and and and I'll say that thing, right? Say the thing techn you need the technology to make this foundation happen. You you're not going to build it from scratch. You're not going to do this DIY doesn't work in this case. Okay? And as much as people like to think that it does. Um do you have this technology? Do you have the ability to build this data foundation? take a really good quick hard look at that.
Okay. And the answer is going to be very quick. And that if you don't have a platform like Relio, if you don't have that kind of capabilities both everything from you know analytical use cases that can be driven by it to operational use cases and real-time capability. If you don't have that, you're starting off on the wrong foot. You're already behind. Okay. So I would say you start with that. Next I would say once you have that kind of a data foundation really focus in on the use case.
Hone in on the use case that you want to solve for and make sure it's a use case that will benefit from AI. Okay? Absolutely. Make sure it'll benefit from AI. AI is not the hammer for you to be walking around and looking for a nail. Okay? So, make sure it will benefit benefit from from AI. And I think the the the age-old philosophy of just, you know, start small and uh think big, of course, start small but scale fast, I think applies to all of these.
Agreed. Agreed. And you know, the idea of doing that all yourself, you know, I I can see why the temptation's there. You might think this is going to be cheaper. We can do this, but the truth is, you know, you're you might still be building it 9 months from now. Yeah. You know, and you need to move fast. Today is about speed. It's not just that they'll be building it for nine months from now, you know, and I just say one more thing about that is you spend a lot of time and effort and resources building it for a specific um use case or a specific scale that you know at that time and scale changes in business very quickly and all of a sudden you find that what you built can't scale right and there's so many other aspects to this that you need to think about than just building it for one specific use case.
Yeah, you know, not to digress, but DIY doesn't work in this world. It really doesn't. I agree. B, that was an excellent look at how data strategy has become such a strategic foundation for the next wave of AIdriven business. Um, thanks so much for sharing, you know, some concrete frameworks, real world examples that people can can run with and uh better understand what it is they need to do next. Where can readers go to learn a little more about Relio?
Well, I mean, we have our um our website, relio.com. Go right over there and engage there. We have um various ways to reach out to us. Um, of course, you know, you can always find me on LinkedIn and you can reach out to me and uh any anybody in the in in our organization and uh yeah, there's there's there's just so much going on in this space and we are at the bleeding edge of this cutting, you know, forging new territory, especially for a platform that has its history in in MDM. uh for us to be forging this kind of you know um agentic future for uh for data I think it says a lot about uh the way we think about the future and where we think about our flat platform and where we where we're going with this.
I agree. I agree. Well, Abby, thank you so much for joining us today. It's been a lot of fun. It's been an interesting talk. I've learned a lot as we went. Yeah, thank you Cory. Appreciate the time. Absolutely. Well, thank you to everyone watching today for joining us here on Espeaks. Subscribe. Leave a comment with your biggest data challenge. We want to want to hear about the struggles you're facing day-to-day, too. I'm Corey Nol, and we'll see you back here next time. Thanks for watching.
