Turning AI Pilots Into Measurable Business Value With Tredence

[00:00:04] Why do so many AI pilots prove the technology works, but still fails to change what everybody does on a Monday morning? Well, today I'm going to be talking with the Chief Operating Officer at Tredence. And together we're going to discuss the last mile between generating an AI insight and then turning it into measurable business value.

[00:00:28] And my guest will argue today that the missing piece is usually execution rather than another model. So we will discuss why AI must be built into the workflow where decisions happen, why employee resistance can be perfectly rational, and how leaders can replace model counts and demo applause with measures such as margin, cycle time, loss reduction and customer satisfaction.

[00:00:58] I also want to learn more about Tredence's AI native operating model. And my guest will also make a persuasive case for giving data hygiene, evaluations, governance and runbooks a much bigger place on the enterprise stage. My kind of language. So enough from me. Let me introduce you to my guest now.

[00:01:25] So thank you for joining me on the podcast today. Can you tell everyone listening a little about who you are and what you do? Yeah. Thanks, Neil, for having me here. It's a pleasure to be here. I have a long Indian name, but I go with Jit. Easy to remember just in time, if that's easy for the people, the audience. Been in the industry for almost now over 25 years. Spent all my life in data analytics and AI, setting up the businesses for large companies,

[00:01:54] primarily helping the Fortune 500 organization. Spent significant amount of time in Cognizant, LTM entry, Wipro, before taking this role as a COO at Tredence. It's interesting when I started many years back, it's all about decision support system, data warehousing, which is probably in the sidelines. And from there today, excited to see agent TKI systems where data becomes a very, very important thing. I'm very happy to be here in this place.

[00:02:24] Coming to my job, just to give a little introduction at Tredence, as a chief operating officer, I'm responsible for global delivery and building global practices. And I also run all the operations. And most important thing, which I'm super excited these days, is about making the organization for AI first, which means whatever we do in the organization is AI first. And sometimes when people ask me, can you describe an airline?

[00:02:49] I talk about at Tredence, we talk about a last mile promise to companies. And my job is to make that last mile promises an operational reality. In one line, that's what I do for it. And at Tredence, I've been reading how you described the last mile of AI as almost a gap between creating an insight and actually realizing business value from it. And return on investment from every tech project is a huge topic right now.

[00:03:16] So why do you think so many organizations are still struggling to close this gap? Yeah. So I think when we started a few years back, this has been a gap. And interestingly, that gap continued to be there. Different organizations are struggling. Look, Neil, many enterprises are continuing to solve a couple of problems broadly, if I say, while many are there. One is about we all have an enterprise that has a lot of legacy. I mean, they've built over the period of time, lots of dashboards and lots of presentations and excels and all.

[00:03:46] They never reach the nth mile. What I mean by that is where it has to reach a supply chain manager or a store associate. They need to get those real-time insight at the right time. That's been still a challenge. That's number one. Number two, an interesting phenomenon last few years all of us are dealing with is an AI. AI is going super fast. It's just bringing things, insights much faster. But the challenge there is about the trust and governance part of it, right?

[00:04:15] I mean, we get some stuff, but we don't know whether how to trust that information. So in a way, everyone has access to models. Everyone has access to whether you call open-weight models or open-source models or frontier models, etc. But what we don't have is an execution rhythm. In short, what I would say is about we all buy, we all rent models, and we continue to build capabilities. But what we don't have is an evals or a culture and things like that.

[00:04:44] So in simple terms, the gap is not because of a technology problem. Gap is purely an execution problem. A very, very clear execution problem where we are not able to translate faster to the right people at the right time. So that's been the challenge is the last mile problem we're talking about. And I think everything that we've seen over the last two years proves the point that you've just made there. Because we've seen countless AI pilots demonstrate that the technology works.

[00:05:12] But when many companies execute, many of them never become part of everyday operations. So what is it that typically prevents a successful AI experiment from becoming something people consistently use to make better decisions? What's happening here? Yeah, I know it got lots of attention thanks to MIT has done a study and talked about. It's been most sought after all of us discussed enough.

[00:05:41] While there are many reasons why some of these pilots don't get into, you know, the day-to-day operations while we all get excited. In my mind, one of the most important is we have not redesigned the entire actual workflow for this new AI world. What I mean by that is, imagine we're going to get some kind of an output which we need to check. And then we need to go to another workflow where we need to remember that and then try to do stuff. I don't think it's going to work, right?

[00:06:10] I think it has to be a seamless into your pilot and your project has to be seamless into your existing workflow where you need to redesign. So in addition to that, we continue to have the trust gap, right? The AI pilots have given up beautiful results, but when you get into an entire workflow, people don't quickly adopt to something which has not been very, very explainable, which doesn't have a clear governance, which basically you can't contest.

[00:06:35] In a human-to-human, we contest the decisions, we contest the data, but whereas in that aspect it is not there. So that's been a serious issue. And another thing which I observed is about we all got super excited when a lot of these capabilities came. We started building lots of stuff. But what we lacked was about the evaluations of those outputs continue to be still manual, right? Because when you're doing the pilots, it becomes very easy.

[00:07:02] A bunch of people can come and look at it, but at an enterprise level, you want to do that. The build versus evaluation gap continue to go, right? So that's another reason why your fast-track pilots, which happens and some of the people are building in hours and days, it takes a lot of time getting into enterprise. Then you start uncovering a lot of them. And most often, Neil, if I can add a little more color to that, it's a little bit of afterthought. You start jumping into something available, a new toy is available.

[00:07:32] You jump into that and start playing with it and suddenly realize it just needs to do a lot of different things. So just to make it real, I can even throw a couple of examples which we had to deal with customers of our side. So one of the retail customers, when we were working with them, they just got excited and built some kind of an emergent capability, which they wanted to rewire with a pilot. But then suddenly they realized it's not going anywhere, which is not yielding the results, and they've been questioned.

[00:08:00] And so the way we looked at it is about, can we reimagine this entire workflow? Can we build an entire merchant AI ecosystem where it will just directly get integrated into their existing workflow? And then naturally you start seeing is about what analysts used to take days to complete something. It's just happening in seconds and they were able to save 70% of effort. And then it just moved into the happy part. The same way, another manufacturing company where we work with them in a supply chain area,

[00:08:30] instead of building one more dashboard for the supply chain insights, what we said is about a let's build Coinnovate, a supply chain agentic platform where your supply chain managers and supply chain associates will be able to really leverage the capabilities of research that is required. And suddenly they started seeing a significant value from their inventory. And in the very first year itself, they saved $10 million.

[00:08:55] So when you're able to take these pilots into either reimagination or embedding them into the core of your workflow, the pilots are translating to value. And wherever the organizations are not able to do that, they continue to struggle. That's what we observed over the last few quarters. And one of the things I love when I was doing a little research on you is I was reading how you often argue that the culture change in an organization is central to AI adoption.

[00:09:23] And this is music to my ears in my former life in my own IT career before AI. I saw so many different examples of projects that were having trouble because the technology was there. The solution was great. But the mindset and the culture change that was required to make a success increase adoption. These are the things that were missed. But if we fast forward to present day and we're looking at AI adoption, what does this mean in practical terms?

[00:09:49] How can leaders listening overcome resistance without simply forcing employees to use these new tools? What are you saying here? What methods have worked? Yeah. So I know in a way this resistance is not, in my opinion, it's not really a stubbornness, Neil. It's more about people are trying, people are actually rational in their mind because they're still trying to grapple with what does this mean to them and how they need to think about it.

[00:10:16] I think when we go back to a little bit of history, I think sensemaking is what probably people need to understand. How does it make sense to them based on the history? In my mind, to drive this culture change, first, I think we need to start showing to the leadership within the respective organizations, talking about what's the value of this, right? Is it like a pet project? Somebody got excited? What's the technology shift that's happening? I think the very first thing which we have to do.

[00:10:43] The second thing in my mind is about we need to just translate this back to how does this change their life? Again, I'll go back to the same example of a supplication manager or a store manager or a line manager. How does your Monday morning looks like with this change? I think you're able to translate that. I'm sure the chances of this, you know, the resistance will get away and they're all eagerly looking forward. This is not something, a very difficult path, which is going to make life difficult.

[00:11:12] Rather, this is going to make my life easier. So one is about at a leadership level, getting the attention. Second level, able to articulate that in my mind, which will be a very good beginning, starting point, right? Then, at Trident's, we actually reframe this entire team structure, right? I think, how do you make it? And what I call, just to make it easy, is about a Trident's tandem. What I mean by that is about the humans and agents get together. How do you put a team structure in such a way? There is a space for everyone.

[00:11:41] And all of us like to do some stuff which is much more faster with the agents. But at the same time, we need to do a lot of things which is about decision making, reimagination, and putting together the initial set of our vision. How each of us have our roles within the Trident's tandem becomes very, very important. So I think when we start building these pieces to the teams, it becomes very easier for the people to start realizing this is the culture I'm going. And just adding a little more color to that.

[00:12:11] And the day one, we can't just expect we put together some vision and expect everyone jump in. Just start trusting this. I don't think anyone is going to join there, right? We need to put some kind of a ladder to that where we'll say is about this is the way, you know, the phenomena of sense giving, I call. What I mean by sense giving is about we need to help the people. How does this translate and what will help?

[00:12:34] I think once we have it and then we are able to expose our entire knowledge base as an organization, all of us would have built over a period of time and make agents realize that. Then suddenly we start seeing is about the humans within the organization start realizing this is the future and this is the way we work and translate a big picture. I guess that's where the culture can startling. And there are many methods over the years all of us followed Neil.

[00:13:01] And one such thing we are experimenting, which is working for us and even for our customers, what we call internally as AI foundry. There you bring a bunch of people on a particular day, expose them the tools and give them a problem statement. All of them get together. And sometimes we don't expose this in a day-to-day job. And you expose your vulnerabilities. You really learn from each other. And then when you come back on a Monday morning, you just start appreciating much faster than just going through some YouTube tutorial and then you lost somewhere.

[00:13:31] So that's what I would say, Neil. Your agentic AI might not be secure even with real-time data and proper guardrails. But Denodo makes sure your business has every avenue covered. By placing all your data platforms under one AI data layer, your business can reach semantic consistency safely and securely. So get your agents on the same page by visiting denodo.com.

[00:14:00] And you can learn more about how to start trusting your agents to make business decisions. But now, back to today's guest. And we will have people listening inside organizations that have invested heavily in data platforms, AI models and infrastructure. And for those people listening, how should they be moving the conversation from creating value with AI to actually extracting real measurable value from some of those investments?

[00:14:28] And I appreciate that is probably a question for an entire podcast episode. But anything that you're seeing here to advise people listening? See, I think end of the day, any of our enterprise businesses is one about creating value. And most importantly, after creating value, how do you extract that value? And we need to translate that value into all of us who have measured on the top line, bottom line, customer satisfaction kind of a thing. This is the most sought-after topic.

[00:14:58] I think we've been challenged in questions in every organization, the value of AI investment. In my mind, I think we need to probably, some organizations are forward-looking, doing a decent job. But quite a few organizations still struck with the KPIs, which are a little bit of, in my mind, I call them vanity metrics. From a technology, maybe we focused on how many models we build, how many dashboards we build, how much accuracy we build. We are just struck with that.

[00:15:25] So maybe from there, we just need to quickly start looking at that. And while it takes time in some cases, it may take a few more weeks or months, we need to start talking about what's a measurable value with this to the bottom line. And what's a measurable cycle time reduction, which is happening. If you are probably building a product or even launching a new campaign, cycle reduction becomes very important. So we can move away from a typical technology metrics into some kind of a business metrics. The real value will start realizing.

[00:15:55] And it gives an opportunity, if you're not seeing value, it gives an opportunity to even go back and calibrate that than getting super excited with number of models built, number of check-ins happened to the GitHub Copilot kind of a thing, right? Here, I would probably give a couple of examples what we've seen with some of our customers who got benefited with this.

[00:16:17] So one of our customers, they were a leading supermarket chain and they were deploying the planner first kind of an approach to their platform. What I mean by that is embedding AI into that. And certainly that forecasting platform, I know with lots of changes in the world, the forecasting becoming the most important for every organization. They were able to extract almost a $200 million value out of this.

[00:16:43] And interestingly, the replenishment match rates increased over 90%, which is generally considered very high, very good for a supermarket where you're able to manage that, right? A second example I would also add to that is about an organization which is trying to do modernize. These days, it's very important for organizations to start modernizing to take the value of data and AI. So they modernized their platform using predominantly AI ways of working.

[00:17:13] And they started realizing almost $100 million in their loss reduction. And particularly their customer fill rates increased from 70% to over 90%. The reason I'm talking about this, once you start measuring the lens of the business value about whether it's a margin improvement, whether it's about your full rates, whether it's about loss reduction. Then naturally there is more hungry. I think people start tasting that value. You start getting more extracting value of that.

[00:17:42] I think I can go on examples across the industries. One of our TMT customers, we were able to help them in networks where they were able to do a $50 million saving. In another place, we were able to embed another wellness retailer kind of a situation. We were able to put a wellness advisor as part of the consumer shopping journey. And you're able to embed your egg in the shopping journey. Naturally, that increased your top line incremental revenue for your e-commerce business significantly.

[00:18:12] And most importantly, it's actually improved the conversion. I think in the e-commerce journey, you don't want to lose your customers. You want to make them a complete transaction. So these are all the measurable values which we talk about at this point of time. And that's, I know there is some more work to be done. As an industry, we need to get into more meaningful or standardized metrics for A and A to metrics. I'm sure collectively we'll get there. But these are all good beginnings some of the organizations started doing. They were able to help them.

[00:18:42] And again, when doing a little research on you, I was reading that you said that vision without an operating architecture is merely ambition. I absolutely love that line. Real killer line. But for people listening, what does an AI native operating model actually look like? And how do things like governance, accountability and decision making, how are all these traditional methods? How do they need to change in this AI world that we find ourselves? Yeah.

[00:19:11] So listen, the reason for that is, Neil, I know it's a little bit of provocating some of the audience may be able to appreciate. It is, I think, thanks to the frontier models and other things these days, building a vision is becoming much more easier. Everyone has a vision. Everyone has a strategy. And we're able to get together lots of them. But I think what we are actually, and by the way, interestingly, no longer we have a problem with accessing models. In the past, we are struggling to build models.

[00:19:40] Now we can access the models also very easily. But what we are lacking is an operating architecture. What I mean by that is, let me break this into a couple of parts. The first is about your AI native maturity, your AI native operating system require three clear layers. Okay. With interference, what we call is about, one is the foundational. The foundation is more about you need to have your data infrastructure. You need to have your data governance. All these foundation layers has to be in place.

[00:20:10] Which is a foundational for your operating model. The second is about your intelligence. Now you have data. How do you transform that data to decisions? And that's where your intelligence part of it will come. The third most important layer, in our opinion, is about experience. Which I said a few minutes back as well. Experience is more about your intelligence will start meeting the human workflows. You're not leaving the intelligence outside, but you're bringing intelligence part of your human intervention, human workflows.

[00:20:40] And where the agents are co-creating and co-pilots are working together. So, in short, if you think about your foundation, you know, your intelligence and your experience, they have to come together. And most organizations have done reasonably well on one and two, which is about your foundation and intelligence. There's still a long way in terms of driving the experience. So, that's why we talk about only less than one out of five pilots get into a real value.

[00:21:07] So, that's where operating model, in my opinion, is a very first step that, you know, start hitting you, Neil. That's the first one. The second one, to make that operating model, while again, it looks like pretty easy, just building foundation, doing some intelligence and experience. What makes a big difference for organization? You need to have a building blocks to do that, right? One is about, I spoke to him, you know, about, you know, Trident's Tandem, which is about having humans and agents to come together.

[00:21:36] The second is what I call execution rhythm. No longer we have an opportunity to wait for some humans to complete, humans to evaluate. Our rhythm has to be equally fast, like how much ever we are building, we need to be able to, you know, test it or deploy it or run it kind of a thing, right? Then we spoke about the third dimension about A native metrics, which is very, very important part of operating model. I think organizations start looking into that.

[00:22:03] The fourth one, always important, but it's even more important now. We need to have a strong technology products ecosystem. What I mean by that is, I mean, your frontier model companies, your AI data product companies like your Snowflake, Databricks, GCPs, and then you need to have your hyperscalers, right? Your AWS of the world, Google of the world, Microsoft.

[00:22:26] So, sometimes I'm a little hesitant to box some of these companies in one or other, but broadly, your hyperscalers or your frontier models and data products specialists, you need to have a strong, you know, ecosystem to be part of your operating model because you can't deliver everything customer looking alone. And then final, final one, I talk about some kind of a transformation.

[00:22:50] The organization should look at it in your own way because your culture, your values, your trust, what you build in your company should be part of operating system because we can buy a front-end model. We can buy an infrastructure, but what we can't buy is about what works in us to make that operating model real. And I'm sure this is not the first time, Neil. I'm sure you've seen in the past decade as well. Some of this transformation works only when you have a solid operating model in place.

[00:23:19] Otherwise, this continues to be a challenge question every time about what's the value of this investment. So, that's what I would say why operating model is super critical. Love it. And finally, if we do have a CEO or a business leader listening that believes their company has plenty of AI activity, but they're just not quite seeing that business value they were expecting, any practical steps that you would leave the people listening if they were that they should take and help identify whether the problem is technology,

[00:23:47] culture, adoption, or the way the organization itself operates? What would you advise that person listening? No, I think the very first thing, the amount of paranoia every CXO is going through, in my mind, is real. You know, I think all of us are dealing with in many ways. The very first thing I know, there are quite a few companies started doing lots of good work. And over the years, we, along with our customers and partners, were able to move the needle.

[00:24:16] So, those of you listening and would like to pick it up, I would break this into a few parts. The very first thing is about create your own diagnostic to start seeing is about how you want to drive your AI journey. The very first step I will talk about, let's get out of or let's escape from this demo trap. These demos are beautiful. They continue to be nice. Can you come out of the demos and start asking the questions about how do you build this into your own workflows,

[00:24:44] either reimagination or bolt-on into your thing, right? That actually performs, you know, operational work inside your enterprise systems. They have to work, right? Because particularly the Fortune 500 or, you know, the 2000 companies, they invested heavily over the years. So, it's important for you to get out of the demo trap. That's the first one. The second one is about we need to have entire organization ready for it.

[00:25:10] What I mean by that is we can't just one particular group started building. As I said, the rest of the organization, whether you're a deployment engine or you're a run engine, they're not ready for it. Then you have a problem of in the olden days, which is to talk about on the shop floor, there is a lot of accumulation happening in one process and we're not getting the final product. So, you see the same now, you know, some portion of your business, some portion of your process is working super fast and others are not catching up.

[00:25:39] My second advice is about can we re-look at the macro picture before we start trying some of this thing. The third thing is about we need to make the entire, I won't say entire organization structure per se, Neil, but at least the people working together are the type of parts required to deliver value. As there is amount of good work has been built, every company is building lots of agents.

[00:26:05] Do you have agent marketplace in your own company where people can leverage along with some humans? That becomes a very, very important playbook or a blueprint which is required for the CEO, but an ECE at source for that matter. And then the last one, which my favorite one, some of the people often underestimate is about, I think boring work deserves the bigger stage. What I mean by the boring work is about you can't get the real value of AI continuously

[00:26:34] unless your regular execution work. Imagine like your data hygiene, continuously valuation of things, change management, your agent governance and your run books. This and quite often people get super excited innovators. They often underestimate and they don't like. And to me, it becomes very important for the CEO to have a clear mandate to make this boring work the main stage so they get equal attention in the organization

[00:27:03] so that you're not just building, you're actually making it real and you continue to re-innovate yourself. So that's probably, I would say, will make this real and this make repeatable in an enterprise. I think that's what organizations look for, repeatable thing, not just one time wonder. So that's what I would say, Neil. Fantastic advice. And for anyone listening that would look to carry on the conversation and talk a little bit deeper on a lot of the things that we've raised today, where would you like me to point everyone that want to connect with you

[00:27:33] will find out more about all things Tredant? Yeah. Well, feel free to get into www.tredence.com. That's where lots of latest and greatest information. And those of you would like to catch with me, I'm decently active in LinkedIn. Now, obviously, that's a default platform these days. So my name is Chitendra Pucha, J-I-T-E-N-D-R-A, P-U-T-C-H-I. You can find me in the LinkedIn.

[00:28:03] And some of you are exploring the, you know, transforming yourself into the new opportunities. You can also check it out in Tredence, carriers.tredence.com. So lots of interesting stuff is there. But all in all, very happy to exchange views, very happy to learn. I think this is the time for AI. And your best time to be is right here. Yeah, I completely agree with you. We covered a lot today for that adoption imperative.

[00:28:32] Why last mile AI often fails without culture change. And also looking beyond value creation, the art of value extraction in this AI era. And by that, I mean, not just models and foundation, but focus on how to extract value from these investments, how to measure that success as well. I will include links to everything that you mentioned there, including your own LinkedIn profile. And encourage people to carry on this conversation. But more than anything, thank you for starting it today.

[00:29:01] Oh, it's a pleasure. It's my Neil. Thank you so much. I think my guest left us with a useful test for every AI program. If the technology produces an impressive answer, but then asks an employee to leave the workflow, check another system and remember what to do next, the last mile remains unfinished. Business value appears when intelligence reaches the right person at the right moment.

[00:29:28] And employees understand how it improves their work. And leaders, well, they can measure the result in terms that the business already respects. So I loved his defense of the supposedly boring work there. Because data hygiene, evaluations, governance, change management and run books, they never receive the loudest applause.

[00:29:53] As an ex-IT guy, I know how they help turn a clever demonstration into something reliable and repeatable. But over to you, is your AI program producing activity, AI theatre, or is it genuinely changing an outcome that somebody can measure? Whatever it is, I want to hear from you. I want to hear your story and your experiences. So techtalksnetwork.com.

[00:30:22] You can leave me an audio message over there. I'd love to hear from you. Connect with me on LinkedIn at neilchughes2. But I have eaten into far too much of your day. So I'm going to go now. I'll be back again tomorrow with another guest. But thanks for listening as always. Bye for now.