What happens when enterprise AI moves faster than the data foundations meant to support it? That question guided my conversation with Sumit Mehra, CTO and Co-Founder of Tredence, who joined me while travelling between customer meetings on the US West Coast. Sumit has a clear view of what is coming next, and he believes we are entering a phase he calls data Darwinism.
In his view, the next stage of AI advantage will not be won by the companies with the most models or the flashiest demos, but by those with the strongest data habits. Clean, governed, connected data is now the primary fuel for autonomous decision systems, and the enterprises that fail to address this will struggle to move past surface level gains. As we unpacked this shift, it became obvious how much of the real work in AI has only just begun.
Over the years, Tredence built a reputation for solving the last mile of analytics by bringing insights out of slide decks and into the hands of the people doing the work. Sumit described that early chapter with a sense of pride, but he was quick to point out that another transition is already here.
With agents now influencing and making decisions across supply chains, forecasting, and customer experience, enterprises are moving from reviewing insights to reviewing decisions. That shift demands stronger data platforms, tighter governance, and a cultural adjustment that many organisations are still wrestling with. Sumit spoke openly about how teams need support to trust agent driven outcomes, and how the leadership layer plays a major role in closing the long standing divide between business and technical groups.
Our discussion also moved into the rise of real time decision systems, the move toward unified data platforms, and how vertical AI is reshaping expectations inside industries that rely on precision. Whether it was supply chain visibility, marketing personalisation, or the growing need for credible governance models, Sumit emphasised that organisations can no longer rely on siloed data or fragmented strategies.
As Tredence expands deeper into regulated industries through its acquisition of Further Advisory, the work ahead touches everything from finance to healthcare. It left me thinking about how ready most companies truly are for this next phase, where every agent is only as reliable as the data beneath it. Where do you stand on data Darwinism, and how prepared do you think your own organisation is for what comes next?
Tech Talks Daily is Sponsored by NordLayer:
Get the exclusive Black Friday offer: 28% off NordLayer yearly plans with the coupon code: techdaily-28. Valid until December 10th, 2025. Try it risk-free with a 14-day money-back guarantee.
[00:00:03] What is it that separates companies that are thriving in the age of AI from those that seem to be quietly fading away or running into problems? Well, as businesses everywhere continue to rush to embed generative and agentic AI into their operations, the next very real differentiator might not be algorithms or compute power, but something far more fundamental, data readiness.
[00:00:30] So in today's episode, I'm going to be joined by the founder and CTO of a company called Tredence. And together we will unpack a concept that he calls Data Darwinism. And we'll also talk about how this new phase of enterprise AI is reshaping the competitive landscape. And why only the data fit will survive.
[00:00:55] And also how vertical AI models and infrastructure strategy, how these things are changing what it means to be AI ready. And before we bring today's guest on, I just want to give a quick shout out to our sponsor, because it's their continued support that helps bring this network of more than 60 in-depth interviews every month to life. And thanks to them, we can keep connecting innovators and ideas across every corner of the tech world.
[00:01:24] As you all know, I often talk to founders and business leaders who spend months building incredible products. But it's easy to forget how that one overlooked login or unsecured Wi-Fi connection can bring everything crashing down. That's where Nord Layer comes in. It's a business security platform built for the modern hybrid team.
[00:01:47] No hardware, no complicated rollout, just a simple toggle ready solution that lets you secure your entire network in minutes. And it does this by combining VPN, access control and threat protection, so your people can connect safely from any device and any network and anywhere.
[00:02:06] So whether you're five people or 500, you can scale it within a few clicks and sleep better knowing that your data is protected by the same team behind NordVPN. So if you've been putting off network security because it feels too technical or too expensive, don't wait any longer.
[00:02:27] If you've been putting off network security, don't wait any longer. But enough teasers from me. Let me officially introduce you to my guest now. So a massive warm welcome to the show. Can you tell everyone listening a little about who you are and what you do?
[00:02:54] Hi there. I'm Sumit Mehra. I'm one of the three founders of Treatance Inc. My role, my title says I'm a chief technology officer, but I also play the role of chief practice officer and also lead the M&A mandate for treatments. That's me. It's a pleasure to have you join me here today.
[00:03:15] One of the things that attracted me to you, one of the reasons that I invited you on here is I was reading how you were saying that data Darwinism is the next critical phase of enterprise AI. And when I heard that phrase combined with a few things I've heard at conferences lately, which is essentially no data, no AI. I thought I've got to get you on here and find out more about your beliefs. So tell me more about that data Darwinism and why is the next critical phase of enterprise AI?
[00:03:43] No, absolutely. Look, the future we know is already happening with the agents and the bots transforming enterprises and helping enterprises make autonomous decisioning. But agents and robots will not be going to be as good as what the data the enterprises keep or actually maintain and actually already have.
[00:04:10] So, hence, it's all about data in future, creating a great data repository, a clean data repository, good history of data repository, and using that and enriching it in future to help your enterprise functions get automated is the way to go.
[00:04:33] So hence, most of our conversations with our customers is all about how they bring in the silos of data into clean data set, data repository onto data platforms, help govern this data management, if you will. In fact, there's a lot of terminologies coming up in governance called data. You have to prevent your data from getting poisoned.
[00:05:01] There are bad actors out there who can actually, you know, add, pull bad data into your data, which actually makes you completely inoperable. So there are a bunch of ways corporations are working towards maintaining and cleaning and setting their data structures up for the future. So the future is more autonomous using agents.
[00:05:25] And I should also highlight that you've built treedance around solving what you call the last mile problem in analytics, which is something we don't dig in deep enough on this show. So can you explain what that means and also why it has become such a sticking point for enterprise AI adoption? Absolutely. This is one of the things that has really driven us, inspired us and inspired our teams to provide last mile of analytics solutioning for our customers.
[00:05:54] That's been the war crime since we started the company. Well, what it means is, you know, if you go back 15 years, which sounds like ancient times, frankly, the world was running on BI, primarily driven through business intelligence reporting, which is also was for a lot of, you know, our customers non-existing. A lot of the decision was done based on PowerPoints.
[00:06:21] Analytics was hidden or embedded or highlighted on PowerPoints, which basically ended in a board meeting with C execs. So you do all your math, you put it in a PowerPoint, take it to a C exec and then not heavily.
[00:06:39] But that's where it ends, because analytics never went or was not never taken to the to the line manager, the supply chain, you know, manager on the, you know, who has to make decisions about the forecasting or or inventory optimizations or frankly, freight, right? Decisions, if you will. They never got any value out of that. It was done by most corporations.
[00:07:01] So what we decided initially on was that as opposed to us delivering PowerPoints to C execs, we'll deliver buttons, applications with buttons to the line managers. So and embed behind those buttons, the analytics that they need. So think about a freight manager deciding whether they need to use FedEx, UPS or the internal trucks, as opposed to figuring out in an Excel sheet.
[00:07:29] Give them a button in the application they were using, where the button actually has all the embedded AI and they click and they get to know what they need to do, right? In terms of how, what, what decision they need to make. So taking insights out of PowerPoints, which was primarily considered by C execs to embedding insights and analytics into applications consumed by the line managers. May I quickly add one more thing here, Neil?
[00:07:59] Sure, go for it. That's the world we have been wanting to drive for last 15 years. And we've been pretty successful in that. That's been the work I have treated for so long. But we're in the midst of another transformation. That's the age intake transformation. Now we don't even need that button. The button needed a human being to make a decision as to what freight, right, transportation mechanism they will use.
[00:08:28] Now agents will make that decision based on, based on analytics that you feed into them and the data that you feed into them. And humans will validate whether that data is actually the right, whether the decision that agents have proposed is the right data on, right decision not. So the world has changed and we ourselves are changing that philosophy or rather that statement.
[00:08:55] Last mile of analytics remains the core tenet of what Tredence does. But it now has to change with this new, in this new world of agents where it's not about making a decision. It's about reviewing the last decision. That's what we need to enable our human line managers to do. I hope that makes sense. It really does. And I think we should also recognize it was, what, three years ago since OpenAI released ChatGPT.
[00:09:23] And in that time, many companies are still moving from fragmented and siloed data strategies to unified AI native decision platforms. I've got to ask, from everything that you're seeing during this time, from 15 years and indeed the last three years, what are the biggest shifts driving this transformation and how are you helping enable it at Tredence too? Yeah, absolutely. Look, as I said, we are living in exciting times. Yeah.
[00:09:50] 15 years ago, people were just getting onto cloud. Now about, now that's the given. It's now about, as I was saying earlier, about building a data strategy and building data platforms.
[00:10:07] And these, so most of our, most of our customers are right now really heavily investing in building, you know, unified data platforms on data, data platform providers, such as Google Cloud Platform or Databricks or Snowflake or Microsoft. These are the core leading ones. And we partner with all of them.
[00:10:28] Depending on where our client is in their journey, we recommend one of these platforms or sometimes multiple of these platforms, which speed up the creation of the core data asset that our customers need. So think about 15 years ago, people were just, you know, getting onto cloud and most of the data and systems were on on-prem. Log data still is on-prem.
[00:10:56] But that's what we're working on, on helping our customers take their data from on-prem or even on cloud and putting them into one of these data platform providers. So we're going to do a lot of data, you know, and create data applications using agents and agentic platforms. So there's an instant value.
[00:11:24] It's not just about building data platforms for, for the future's sake. It's about building data platforms for today by creating an, creating applications which give you ROI today. And that's what we've been working on. We, it's phenomenal. Last, last 24 months have been extremely exciting in Tredeans' journey. We had to transform ourselves, our skill sets.
[00:11:48] And our customers had to transform themselves from how they were approaching the data asset built. And again, great partnerships have been formed. We are very excited about what future brings to us. These are very exciting times, Neil. Very excited with what we, how we're able to help our customers. When talking about delivering value, and value is a huge topic right now, especially around ROI of AI projects, et cetera.
[00:12:15] One thing that organizations often miss is they get distracted by shiny new technology. And actually, cultural change is often harder than the technical implementation itself. So, Kuis, from what you're seeing here, and to bring to life what you're seeing, how are you seeing forward-thinking enterprises using AI centers of excellence to, to bridge that divide between the business and technical teams? Because that, that divide's always been there. And it almost, it needs a bridge between the two worlds, doesn't it?
[00:12:45] No, absolutely. It's a great question, by the way. Again, let's start with some sort of history. Yeah. To your point, the divide has always remained. I mean, tech has always been ahead of the cultural shifts within an organization. Let's start with even, you know, maybe 100 years ago when, you know, the car, the automotive revolution started. There was a culture shift then. Cars took over horses.
[00:13:11] There was a lot of angst about what humans and horses will do. Life moved on. Life moved on for good. And, and automation brought in a whole lot of productivity and new jobs opened up. But there was a cultural change that happened, which was resisted by a bunch of folks at that time. By that same change happened again with the internet revolution in early 2000s, late 90s. Once again, that just speeded up and increased the productivity of the, of the humanity.
[00:13:41] And that's happening again. That's the context I want to build up on. There is definitely a culture change that has to be brought in. And, and just like any fast tech change, we are facing resistance. Perhaps rightly so. But what needs to happen and what is happening is that the, that the, the CIOs and the top business leadership,
[00:14:07] the folks who see clear value coming out of this new tech revolution, which is, let me just call it the agent tech transformation revolution. The top leadership, if they are committed to this, they will drive the change. And we're helping our customers drive that change. So let me get to the specifics, specifics points now. Our first job is to show the value of the agent tech transformation to the CIOs in the business world.
[00:14:36] And a lot of everyone has heard about it. Examples help real, how we'll make it real helps. And that helps convince the top leadership this is the way to go. From there on, it's about helping the line managers, the supplier chain analysts, the marketing leaders, as to how their life will transform dramatically with this agent tech transformation.
[00:14:59] And all those conversations, all those changements really helps in driving the culture towards adoption as opposed to resistance of this new tech, new tech world. Again, easier said than done. It'll take a few years for people to start evolving, adapting the new change. But humans have always evolved. Humans will evolve again. And new opportunities will show up and come up. And as they do, the culture will shift towards adoption more than resistance.
[00:15:29] Again, early days, but very exciting days. It's a question of how we evolve, how we create new opportunities for the corporations and for human beings. And when I was doing a little research on you, you've also spoken a lot about the rise of probabilistic systems in enterprise AI. So how does that shift improve decision making compared to more traditional deterministic models? What kind of difference is there here? No, you're right. Look, I mean, let me state it differently. Yeah.
[00:15:59] Look, even in traditional systems, right? The decisions were still made based on probabilities. Nobody knew what future would be, right? You had human beings making decisions using deterministic systems. But ultimately, they were also in their head driving probabilities. I see a report. I see why my revenue is down. But there's a probability it's down because of reason A, reason B.
[00:16:27] But there's also probabilities because there could be other reasons why things could not be working right. Humans were making judgment calls based on the data that was provided to them in a deterministic form. But decision was still being made based on probabilities. Now, the decisions will be made by bots based on the same sort of data.
[00:16:51] But now they have created reasoning models and AI models to calculate the probabilities and make decisions based on the deterministic data that they already had. So essentially, in a way, the enterprises are moving from humans making probabilistic decisions to AI making probabilistic decisions for enterprises.
[00:17:14] The underlying decision making systems remain the same, if you will, in a large way. Having said that, that's what is happening. That shift is clearly happening and it's speeding up. We are just scratching the surface of where the world's headed, of how AI will make decisions based on existing data using probabilities of what can happen. And it'll speed up.
[00:17:40] But the good thing is this will open up a ton of opportunities for the future for humans to become far more productive using the AI and AI decisioning systems. And Treedance is also increasingly focused on real time and autonomous decision making. So again, I'm curious from everything that you're seeing here and all the conversations that you've been having with your customers. Are there any trends in the kind of operational decisions that companies are beginning to automate?
[00:18:09] And also, what kind of results are they seeing right now? You don't have to mention any names. I'm just interested in what you're seeing and hearing out there. No, I appreciate the manner, of course. Look, as I said, the enterprises are moving towards autonomous decision making. And a lot of it is being done through AI and agents. And all of this is done real time, right? I mean, the data is available. You don't have to wait for humans to review and understand and make decisions.
[00:18:35] A lot of these decisions are made real time in an autonomous way with data and AI. Some of the examples that we see that are actually happening. I mean, again, we in Treedance are focused very heavily on two core functions. Back-end supply chain functions for our large customers. For our customers.
[00:18:59] And front-end, marketing-led, customer-facing, customer-loyalty-facing decisions made by the marketing or customer retention teams of enterprises. So these are the two broad areas, supply chain and marketing, that we have built strong capabilities around. So the decision, if you will, in terms of what's happening real time in the world of supply chain.
[00:19:29] Let me back up a little bit. So supply chain is always in the problem, has always been a challenge for human beings because the data has always been in silos. And sometimes not visible to you. So you probably have access to the data that you have of your plans and your inventory and your orders. But you don't have information about your vendors. What is their supply?
[00:19:56] And when are they going to get the parts and inventory of the parts in their systems and all? So the decisions were made based on siloed data, based on not fully available data by human beings. Now we have been able to connect data, not just of your enterprise internally, but also your enterprise with your external vendors.
[00:20:25] So the visibility and the connectivity of the overall supply chain is for the first time in generation coming up. And that has helped AI make real-time autonomous automation on optimizing inventory in warehouses. You know, stockouts, helping improve demand forecasting based on the order visibility.
[00:20:52] A lot of these decisions are now done real-time as opposed to something that used to take weeks or months in the past because of lack of visibility in terms of data and data connectivity. So that's one sort of example. The other set of the example I talked about is marketing, right? The marketing decisions. So human beings have always had, you know, the enterprise have always tried to market their customers in, you know, using...
[00:21:18] And there's been, you know, there's been push towards personalization, but personalization was limited based on what you know about a customer. And so you had large segments of customers, right? Now we're moving towards segments of one. If the enterprises have data and enough understanding of customers where you can target an individual customer, consumer slash customer,
[00:21:43] based on their specific needs, hyper-personalization, all of this is done real-time now because you have a lot of connected sources providing information visibly into consumer behavior and customer behavior and customer needs and demands, which is helping marketers and make real-time autonomous decision, hyper-targeting, hyper-personalizing, right, for the consumers.
[00:22:12] Also, it has helping a lot in terms of campaign creations with this generative AI. In this generative AI world, we're able to create campaigns far faster at much, much lower cost than what we used to do in the past. So all of that, campaign creations, campaign executions, campaign measurement, hyper-personalization, email targeting,
[00:22:39] all this decision marketing side as well has become extremely targeted and real-time. Again, super excited with where we're headed, but these two areas, supply chain and marketing, are becoming really heavily real-time and heavily targeted towards decision-making. Thank you, Goshno Neil. Go ahead. Yeah. And as AI systems continue to make or influence critical business decisions in real-time, as you mentioned there,
[00:23:08] what kind of governance models are needed to ensure things like transparency, accountability, and trust? Many don't get here until it's too late, but I think this should be the foundation. So what kind of governance models are needed here, do you think? Data governance and data security is absolutely paramount in future. As I said, with AI coming in and depending on data to make decisions,
[00:23:37] data is now the most important asset you have, which absolutely needs to be available, visible, clean, and governed for AI to make the right decisions. And as I was mentioning in the past in this conversation earlier, there are always going to be bad actors, and bad actors will try and disrupt your functions in some ways by even introducing
[00:24:06] bad data slash poisonous data into your applications, which can completely disrupt your operations. So it is extremely important for enterprises before and while they're building their data systems, consider data security and data governance as one of the top priorities, something that we're working very heavily with our customers. So as and when we build any data asset or any data application,
[00:24:35] the first and foremost thought and design is based on how do we secure this data? How do we make sure that this data remains well governed, well understood, well explained, and clean for the whole enterprise? It's about building transparent systems. It's about building trust at multiple layers of the data. And this is where human in the loop comes in,
[00:25:05] where while you have different systems reviewing the data quality, humans come in to provide the trust factor. Let me end by saying the following. I know we're running out of time, but look, AI will automate a lot of stuff fast. But the trust is something which is always going to be human. Humans trust humans, if you will. So AI will help you create the trust systems,
[00:25:35] but we have to build that ability in humans to be able to trust what AI is producing. So anyway, in a nutshell, this is a very critical question. We are using a bunch of tools provided by large providers in creating secure data assets and systems for our customers. Sorry for the long answer. I hope this answers your question. It really does. Perfect. And finally, to bring it all full circle, a question I've got to ask as we're looking into the future.
[00:26:02] Earlier this year, Tredence, which is known as a leader in data science and AI solutions, you announced the acquisition of Further Advisory, which is a Pittsburgh-based management consulting firm. Feels like a great fit. But tell me more about that and also how you're bringing decision intelligence to highly regulated industries and everything from banking to finance and the kind of impact you're seeing so far. Oh, yeah.
[00:26:30] So we're excited for the Further Advisory team joining our journey and credence team. It's been roughly about eight, nine months that we've been working with them. No, they bring in special expertise in helping our customers, advising them on strategic decisions. They are a strategy advisory firm for the regulated industries of banking, financial services, insurance, and healthcare. And we're an AI implementation team. So it's phenomenal to have that skill set as part of our overall skill repository.
[00:27:00] Further Advisory team is right now leading strategy engagements with our customers. And as the strategy takes shape, they introduce the Tredence team for the downstream AI implementation. So it's been a brilliant partnership. And it's not just in the existing customers that they bring in. We are now able to add this as a special practice, special skill into our overall portfolio. And we're taking their expertise to our existing customers who are also asking similar questions
[00:27:30] about creating data systems, about how to secure data systems, cybersecurity, AI advisory. All these are key questions our existing customers are asking. And I'll be able to leverage further advisory team in helping them answer that question as well. So yes, great. Very excited with this new add to the Tredence capability. Still early days, still a lot of work to do, but very excited as to where we added. Love it. And we are just about out of time.
[00:27:59] And I know you're traveling at the moment, so I don't want to eat into too much of your time. So for anyone wanting to continue this conversation that we started today, where can everyone listening find out more information about Tredence and indeed anything we talked about today? Where would you like to point them? Well, you can always reach out to info at Tredence.com if you have any questions for Tredence. But more importantly, careers at Tredence.com.
[00:28:25] We look forward to growing a smart team that we already have. So if you are interested in actually joining Tredence in the future, please do send your profiles to careers at Tredence.com. You always can read about us, learn about us at Tredence.com and also on our LinkedIn page. Look forward to connecting with you guys in the future. Awesome. Well, I will add links to everything. And we did cover a lot there from why data Darwinism is the next critical phase of enterprise
[00:28:52] AI to how vertical AI, infrastructure strategy, data readiness, how all these things will determine who pulls ahead next year and indeed beyond. But just to thank you to bringing it all to life today in a language everyone can understand and real world examples. Really appreciate your time today. Thanks, Ian. Thanks. Thanks, everyone. Thank you, guys. Check it. Bye.
[00:29:27] Thank you, guys. What do you think? Are enterprises truly ready for this next phase? or are many still underestimating what it takes to compete?
[00:29:56] We'd love to hear your thoughts on this one. So please join the conversation and let me know what you think after listening to this episode. techtalksnetwork.com. You can leave me an audio message. You want to send me a DM on socials, LinkedIn, X, Instagram, just at Neil C. Hughes. And my email is techblogwriteratoutlook.com. But that is it for today. It's time for me to take a rest now. I'll be back again tomorrow morning with another guest. Thank you for listening as always. Speak to you then.
[00:30:26] Bye for now.