Why are companies investing heavily in AI, analytics, and data platforms while business leaders still struggle to see what is happening across their operations quickly enough to make confident decisions?
In this episode of Tech Talks Daily, I speak with Massimo Merlo, Vice President for UK, Iberia, and Italy at Elastic, about why the next stage of enterprise AI adoption will depend less on who deploys the most advanced models and more on which companies can give people and AI systems access to relevant, trusted, and secure information when decisions need to be made.
Massimo describes the problem as a lack of decision-grade visibility. Most large companies are not short of data. They have spent decades building data platforms, analytics systems, dashboards, cloud infrastructure, and reporting tools. Yet information remains fragmented across departments and applications, insights arrive too late, and employees often struggle to find the small amount of information that matters among enormous volumes of data.
The result is a growing gap between having information and being able to act on it.
Massimo explains why simply adding an AI model to this environment does not solve the underlying problem. If an AI system is connected to fragmented, outdated, poorly governed, or irrelevant information, it can produce convincing answers without providing reliable business outcomes. The quality of an AI model matters, but the context available to that model increasingly determines whether AI becomes a useful business asset or an operational liability.
This leads to one of the biggest technology conversations emerging around enterprise AI: context engineering.
Massimo explains how context engineering provides AI systems with the relevant data, tools, permissions, organizational knowledge, and guardrails required to complete a task safely. Rather than sending ever-larger volumes of information to AI models, companies need infrastructure capable of retrieving the right information and making it available at the moment a person or software agent needs to act.
Fraud detection provides a practical example. An AI agent evaluating a transaction needs more than access to a powerful model. It requires customer history, behavioral patterns, company risk thresholds, permissions, compliance requirements, and the ability to recognize activity that falls outside normal behavior. Without that context, the system could block legitimate customers or approve fraudulent transactions while presenting its decision with complete confidence.
We also discuss why digitally mature companies can still struggle with real-time decision-making. Massimo shares lessons from Elastic's work with organizations including Reed, the Met Office, and Rightmove, explaining why having sophisticated technology systems does not automatically make a company context mature. Information can still remain trapped between applications, teams, and databases, preventing employees and AI agents from seeing the complete picture when it matters.
The conversation challenges another long-standing enterprise technology habit: adding more dashboards.
Massimo explains why dashboards often provide visibility into what has already happened without helping people decide what to do next. Companies can continue adding reporting layers while employees become overwhelmed by information and remain unable to identify the actions that will improve customer experience, productivity, security, or business performance.
A healthcare example demonstrates what becomes possible when companies solve this problem. Massimo shares how CogStack at King's College Hospital brought together unstructured patient information during the COVID-19 pandemic and made it searchable using natural language processing. Clinicians could find relevant information without waiting for technical teams to build new queries or systems, helping medical professionals access information when patient decisions needed to be made.
For CEOs, CIOs, CTOs, data leaders, and technology teams trying to improve AI ROI, Massimo offers practical advice on where to begin. Do not start with another model, tool, or dashboard. Start with a business decision or workflow that is currently too slow, unreliable, or difficult to execute.
Identify what information that decision requires, where the data is stored, who or what system needs access to it, which permissions should apply, and where information currently becomes delayed or disconnected. That process can reveal the visibility gaps preventing companies from turning their existing data and AI investments into measurable results.
We also examine why search and retrieval are becoming infrastructure concerns for companies introducing AI agents. As software agents begin making recommendations and taking actions across business systems, their performance will depend on whether they can securely retrieve relevant information at scale.
For business and technology leaders facing pressure to demonstrate returns from AI investment, this conversation provides a practical framework for improving enterprise search, context engineering, AI agent reliability, real-time operational visibility, and decision-making.
The companies that gain the greatest value from AI may not be those collecting the most data or deploying the most models. They will be the companies capable of finding what matters, understanding its context, and getting trusted information to people and AI systems quickly enough to act on it.
That is where better visibility can become better decisions, stronger productivity, and business growth.
[00:00:03] What if your organisation's biggest obstacle to growth isn't the lack of data, but the inability to find the right information when it matters the most? Because let's be honest, businesses are now collecting more data than ever before. But many leaders are still struggling to make confident data-driven decisions in real time.
[00:00:27] Well, my guest today is from a company called Elastic, and he's going to share exactly why he believes the next competitive advantage is not going to come from simply adopting AI, but how it will come from giving people and AI the right context to turn all that information, all that data into meaningful action. But I don't want to reveal any spoilers here, so buckle up and hold on tight.
[00:00:54] I'm going to beam your ears wherever you are listening in the world directly to the UK, where you can sit down and enjoy the conversation with us. But enough for me, let me introduce you to him right now. 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? Absolutely. So I'm the AVP for UK, Iberia and Italy.
[00:01:19] I wasn't always in sales. I actually started my career as a software engineer, going into consultancy, pre-sales and then latterly sales. So I've seen the full spectrum, if you like, of the industry. From an Elastic point of view, we're the Search AI platform. So we actually have curated that into search, observability and security. We're actually used by about 50% of the Fortune 500.
[00:01:45] What we actually do is we sit at the data, context and governance layer of AI. So while most of the conversation in space, all the exciting conversations has focused on the models themselves, it's that layer, not the model, where really that from our perspective, the value is increasingly being created and awarded. And the reason that layer matters so much is simple. Organizations can't unlock value from AI through models alone.
[00:02:13] And what determines success is quality, accessibility and governance and context of the data and systems around it. So that's me and what we do. Love it. And there's so much I want to talk with you about today. A message I keep hearing at tech conferences is no data, no AI. And of course, most business leaders listening to this conversation today, they will have more data than ever before. But many are still struggling to make timely decisions.
[00:02:42] So why is access to information not translated into better visibility across an organization, especially with the help with AI? What's the gap here? What's happening? That's a great question. And it's actually a question that we get often asked because everyone's kind of jumped on the bandwagon. And like everything, you know, we think it's the panacea. It's going to solve all ills. But the assumption really has been that, you know, if we pick a good enough model, the rest takes care of itself.
[00:03:10] But the reality is that if you just bolt a model onto a broken foundation, it really doesn't produce transformation. It just produces a more confident version of the same problem. You know, AI is very, very good at getting the wrong answer very quickly if the data isn't, you know, accessed properly. So the technology has moved on. The conversation has moved on. But many organizations are really still focused on the wrong layer.
[00:03:39] They're grappling with siloed systems, legacy infrastructure and data that really is hard to access or act on. You know, a problem that's been around for ages. So if you add real-time AI adoption pressure, the budget constraints and talent shortage on top, it really is no surprise that the organizations have more data than ever and yet struggle to find what matters when it matters.
[00:04:08] And when I was doing a little research on you, one phrase that really stood out to me is how you often talk about decision-grade visibility. Absolutely love that line. But what is it that's separating useful business intelligence from data that simply creates more noise? And I suspect there'll be a few people that have experienced that listening. Yeah, no, I wish I'd invented that term. I love that term too.
[00:04:30] But, you know, long and short, it means that the information is accurate, it's current, and it reaches the right person or system at the moment a decision actually needs to be made. You know, like everything, data, information, insights is only useful if you can action it. So anything short of that is just more data sitting in silos or sitting in a dashboard.
[00:04:54] So, you know, if we take AI as being the brain, you know, that really powerful thing, context, you know, the data in the right format is the memory. It's the experience. It's the thing that governs the right decision-making. So AI ultimately is only as effective as the context it can access. So more data, which, you know, ever increased and doesn't automatically lead to better outcomes.
[00:05:19] What matters is whether that data is relevant, it's trusted, and it's trusted enough to act on. So organizations, as I said, that focus on the AI models, they risk overlooking that the harder problem is retrieval, its relevance, and its context. So a model can be excellent. It can produce a really bad answer, as I said, if it's pulling from the wrong information.
[00:05:45] So the difference between a useful AI agent and, well, ultimately and can be a dangerous one is rarely the model. The models are very good. You know, they've been studying for a long time. They're really good. It's the quality of what the model can see. So competitive advantage really won't come from deploying more AI. You know, it's like saying that we have a competitive advantage because we use electricity. It's here and it's going to ever evolve.
[00:06:13] It will come from connecting those AI systems to the right information. So that's kind of in a nutshell what I say, a big nutshell of what decision grade it really means. And, of course, AI is dominating almost every tech and business conversation right now. But you've argued that context, that would be the real thing that makes organizations stand out from the crowd. So why is providing AI with the right context? Why is that becoming more important than just simply giving it more data?
[00:06:44] So if we take the challenge, I think alluded to is that if we the challenge, real challenge is ensuring AI can access the right information at the right time. That's what's going to drive the right answers or the right insights. So that's where the context engineering comes in. You know, giving AI data permissions and guardrails, importantly, it needs to act safely and not just answer convincingly. I think we've all interacted with the various models.
[00:07:13] And what comes back to us sounds really compelling. It's really convincing. Now, as an individual, if it's a subject that you understand, you probably spot that maybe it's not quite as accurate. But, you know, in other areas where things are happening at pace, you can come up with really wrong answers that are very convincing. So if we take an agent in fraud detection, for example, you know, a subject that's becoming ever increasingly more important.
[00:07:39] Without the customer's history, without the understanding of the organization's risk thresholds, who's authorized to act, it either blocks a legitimate customer and damages trust. And arguably that could be, you know, just a nuisance or it waves through a fraudulent transaction because it had no way of recognizing that that pattern of activity was wrong.
[00:08:04] It was anomalous. It was out of kilter with what the customer typically did and obviously against all the compliance records. So that's not just a slightly capable agent, which, you know, we've all experienced. That's a liability and with a confident interface, which makes it arguably more dangerous. So context isn't supporting the layer underneath AI.
[00:08:29] It's the thing that really decides whether AI is an asset or a liability. That's such a great point and one that's often overlooked. And you've often worked with organizations with big names, household names, such as Reid, the Met Office, Rightmove, so many others.
[00:08:48] And without revealing anything confidential here or any information there, what common challenges do you see preventing even digitally mature organizations in the Fortune 500 from acting on information quickly? And what's holding some of these organizations back? Yeah, well, it's certainly not the fact that they lack data.
[00:09:08] They absolutely have huge amounts of data and they're really sitting on enormous amounts of really high quality data sets that's just constantly growing. The challenge is getting the right slice of it to the right person or system fast enough to matter.
[00:09:27] So the common thread that being digitally mature doesn't really automatically mean being context mature is you can have excellent dashboards and still have information trapped in silos that don't really talk to each other in real time.
[00:09:44] So the organizations that we've seen and we've interacted with that have moved the fastest are the ones that treat search and retrieval as the core infrastructure, not just a feature bolted on to an existing system, because that's what lets them act on data in the moment that it's actually useful as opposed to after the fact. And that's becoming ever increasingly more important because of the huge explosion of data. And there's going to be no end.
[00:10:11] That data is going to continue to grow and be valuable, but it needs to be retrieved in the right way. Do you need AI agents that you can trust? Well, with an AI data layer providing real-time connection within your data platforms, you can trust your agents to provide accurate solutions. So scale your business by trusting your agentic AI accurately getting the work done for you. Trust its capabilities with Denodo.
[00:10:40] And you can do that by simply visiting denodo.com to learn more. And context engineering is another term that's gaining momentum. We were talking about context a moment ago, but when referring to context engineering, what does that mean in practice? And how is that changing the way that people and indeed AI systems are beginning to work together now? Yeah, no, absolutely. And I often use that term and I used it earlier as well.
[00:11:07] So it's probably worth spelling it out, become part of the vernacular. And, you know, I assume I understand what it means and everyone does, but it's a fairly new term. So ultimately, it's the practice of giving AI systems the right information, tools and permissions at the right time. So rather than just throwing more data at the model. So AI agents aren't lacking intelligence, they're lacking content.
[00:11:34] So relevance matters more than volume. So it changes how people in AI work together day to day. So instead of someone manually pulling together what's needed, the system surfaces it automatically at the point it's needed with the right permissions already attached. So that's the context. You know, I think I said is AI is the brain. You know, context is the memory. It's the experience.
[00:12:00] It's the thing that really governs the right outcomes and the right insights. You know, a bit like a human being. You know, we could be super smart, but it's our experiences, our memories and what we've, you know, you know, the journey we've been on that allows us to make the better decisions. There will be many people listening inside organizations that have invested in dashboards, analytics platforms and reporting tools. And yet many still say that they don't have a complete picture of what's happening.
[00:12:30] I even hear people more and more say, I don't want another dashboard. So where does that disconnect usually come from? Yeah, it's a bit of a paradox. More dashboards, less clarity, right? It's almost information overload. So, you know, I think from our perspective, the challenge isn't really data collection or storage. You know, most organizations, they've got plenty of both and it's well understood and they've solved for it really pretty well over the years.
[00:12:56] But it's surfacing the relevant piece of information in the moment a decision actually needs to be made. So a dashboard shows you what happened. It doesn't really tell you what to do about it or hand that insight to the person or system that needs it right now. You know, we often get this, you know, in terms of what I do. I see a lot of dashboards and I'm constantly asking, well, what is this telling me? You know, what do I need to do?
[00:13:26] How relevant is it? You know, is it it's just telling me what's just happened? I kind of know. So the gap between visible and actionable is where that disconnect lives. So too often information, you know, it remains fragmented across systems. It remains fragmented across teams, tools. So even when the data exists somewhere, nobody can really see the full picture at the same time. So if you add more dashboards, you're just actually adding more more confusion, confusion.
[00:13:56] So growth increasingly depends on an organization's ability to connect people and increasingly AI systems. So with the information that they need to make better decisions and faster. So not adding another reporting layer on top of the ones that, let's be honest, already aren't working. You're just kind of, you know, throwing good money after bad, unfortunately.
[00:14:23] And just to bring to life everything that we're talking about here, and again, don't have to mention any names, but are there any examples or stories you can share of how when you actually give the right people or the right AI system access to relevant information at the right time, how that has helped organizations improve productivity, customer experience, or indeed business performance? But I think that's the utopia that everyone needs to get to. It'd be great to have some, I don't know, stories of inspiration.
[00:14:51] It will help people understand how they get there and what that looks like. One of the first ones that I came across in Elastic, and it was quite meaningful to me, and it really sort of brought home what that combining separate information. So this is CogStack, which is part of King's College Hospital. So during COVID-19, as you can imagine, you know, this was all very new. The hospital faced an enormous influx of unstructured patient data.
[00:15:21] They needed to track confirmed cases and all the symptoms in real time, and added to the fact or complication that for a lot of these, we know we had COVID, we were calling it various different names. I can't remember all the names at the time. So people didn't even know it was the same thing. So what they did at CogStack is they consolidated all the multiple data sources into a single searchable database, and they combined it with the natural language processing.
[00:15:50] So clinical staffs, i.e. not IT practitioners, the actual clinicians could quickly extract meaningful information from unstructured patient records without really needing any specialized coding or terminology. We can all remember, and it feels like yesterday, but it was a number of years ago, things were changing so fast. And, you know, we couldn't wait for systems to be recoded, reconfigured, re-queried,
[00:16:17] and trying to get a clinician to explain to an IT expert, you know, what they needed and what they needed that information, and they needed it combined in real time as and when they needed it. So really, for me, that's a brilliant illustration of the whole thesis, that the data already existed. It was actually already classified in a way that transpired was wrong, and the bottleneck was surfacing the right piece of it, you know,
[00:16:45] for the right clinician at the moment it mattered for patient safety. So CogStack, they've gone on to support clinical trial recruitment, population health management, clinical audits, what else, you know, and service planning across UK hospitals beyond the pandemic. So for me, that's something that's kind of quite personal, it's meaningful, and it obviously helps me directly and all of us directly, but it just goes to show that they had all of this data.
[00:17:13] At the time that they collected the data, they didn't understand the relevance until certain other discoveries, and instantly that data added more and better insights without any change really to the underlying data. And if we did have a CEO listening today wanting to remove one of the biggest barriers to growth inside their organisation, where would you recommend they start to create that visibility that they need for faster, more confident decision-making?
[00:17:43] Because it feels like the tools are there, but it's hard to see where that value is with so much noise. So where should they be looking? Where should they begin? Yeah, and it's never with the tools, and it's not with the model. Always start with the workflow. And this has been true, you know, going back through 30 years of my career, and probably beyond that, the workflow is important. Pick the decision that's genuinely slow or that drives low confidence today,
[00:18:10] whether that's, you know, to do with the customer interactions, regulatory requirements, security, etc. Trace it back. What data does it depend on? Who needs to see it? And where does it really currently get stuck? So that exercise alone usually surfaces where the visibility gap is. It's rarely that the data doesn't exist. It absolutely exists. It's that it's not really reaching the right person or even the system at the moment it's needed.
[00:18:39] And I think it was KPMG that did a global tech report, and it showed that 88, I believe, percent of organizations are already embedding agentic AI. So they, you know, they kind of got ahead of themselves, but only 24% are achieving ROI at scale. So those that don't see or do see a four times, I think it's four, four and a half times ROI against that 2x average, difference isn't the model.
[00:19:08] It's the foundation underneath it. So the organizations that move the fastest will be the ones that bring AI to their data safely and at scale, rather than the ones racing the deploy the most models and the latest and greatest tools. And I think that is a thought-provoking moment to end on, something that will resonate with people listening all around the world.
[00:19:33] And there will be, I suspect, many listeners who are leaders that cannot see what's happening across their organization in real time right now. And as you highlighted today, the problem isn't a lack of data. It's down to a lack of decision-grade visibility. And that gap is becoming somewhat of a structural break on productivity and growth. And for anybody listening who that did resonate with, want to find out more information about you, your work, and everything that we talked about today. Where can people listening find out more information?
[00:20:03] They can contact us directly. They can... One of the best places to start is actually the Elastic blog. So elastic.co, C-O, slash blog, would be a really good kind of area for people to see discussions on these subjects and various other topics around AI and beyond AI, in terms of really understanding the power of data and how to bring it all together and make it accessible.
[00:20:31] Well, we covered so much today from the rise of context engineering, how it gives humans and AI agents that relevant data, the tools and permissions at the right moment, and why platforms that scale relevance, not just storage, will win. And also, I think, how leaders can close the gap between boardroom strategies and day-to-day execution using this real-time operational visibility as almost a competitive advantage. So for anybody listening that this resonated with today,
[00:21:01] I'm going to add links to the blog, the website, and indeed your LinkedIn. I encourage people to check you out too. But just thank you for shining a light on this. Really appreciate your time. Thank you very much. I think today's conversation challenged one of the biggest assumptions in technology, and that is more data doesn't automatically lead to better decisions. And as my guest explained today, the organisations pulling ahead right now are the ones that are connecting the right information to the right people
[00:21:30] and AI systems at exactly the right moment. And I think this movement from collecting data to creating decision-grade visibility could become one of the defining business advantages of the AI era. And I'd love to hear your thoughts. Is your organisation now making decisions based on timely insights? Or are you missing out on valuable answers because they're still buried somewhere in your data
[00:22:00] that is scattered in so many different silos? And let me know. techtalksnetwork.com I want to hear all your stories. It's one of the reasons I record 60 episodes a month across every podcast on the Tech Talks Network is to try and share as many stories as possible. Let's pull all of our information, insights, experiences, and expertise, and all learn from each other. And that's one of the reasons why I enjoyed this conversation with my guest so much today.
[00:22:29] So keep your messages coming in. I'll be back again real soon with another guest. But thanks for listening as always. And hopefully I'll speak with you again very soon. Bye for now.