S3-E184 - 60% of AI Projects Fail. Here’s What Leaders Are Missing!

(upbeat music) AI is everywhere right now, but here's the uncomfortable truth. Most businesses don't have an AI problem. They have a clarity problem, messy systems, disconnected data, and no real strategy trying to put it all together. Today's conversation is about fixing that from the ground up. Welcome back to the Champion Mindset Collector podcast. I'm your host Anthony Dyer. I'm grateful that you're here listening, watching this episode. Please subscribe and share this episode with someone that could benefit from it. Today we're diving into what really takes to build future really organizations in a world that's obsessed with AI. My guess today is global director of strategy and operations at Intel E-Plas. Matt works with organizations around the world helping them untangle complex text-to-systems, integrate systems properly, and create strong data foundations that enable real transformation. What I love about Matt's work is that it's not about chasing shiny tools. It's about clarity, alignment, and building systems that actually support people and decisions. Matt, welcome to the Champion Mindset Collector podcast. Anthony, thank you so much. Great to be on. It's great to have you here. Matt, before we dive deep into these some of the questions, let's get started and rewind a bit. You've worked across strategy and operations and procurement and integrations. Tell us a little bit about your journey in life and childhood, and then what originally pulled you into this world that keeps you excited about it today? Yes. Throughout all my time, I tend to be in a lot of different places. That friends and business partners, et cetera, always start, hey, Matt, where are you right now? So this started actually pretty young. Initially, it aged 11 and then really continued from age 16. At 11, I was spending the first time four weeks overseas in Canada, actually, at a children's camp organized by an organization called CISB, the Children's International Summer Villages. Anyways, delegations of two boys, two girls, one leader, was sort of going into various different places. I happened to be part of a German delegation at that time and spent four weeks in Canada. That was my first exposure to the world at a fairly young age to sort of see, okay, well, there's more out there than just Hamburg, Germany, where I was born. Hamburg's a great place. I'm a proud hamburger. The, yeah, like I mean, very early, I realized the world is bigger than just, you know, my yard, the school and everything around it there. My parents then gave me the opportunity to spend time in New Zealand, where obviously, you know, this podcast is where you're from. And I owe a lot to New Zealand, to be honest, because that was really an opportunity for me to not just be alone on the other side of the world. I was with a host family that took great care of me and you know, I was then at school in New Zealand and ended up finishing high school over there went to Victoria University and while I came before then continuing studies in Germany. I just 'cause I wanted to have some international European degrees, well, as international, that then continued me being in Ireland in the US, in Australia for many years, and now for the most part, I'm all across Asia. All of that, and like if you're looking at it, it sounds messy, right? Because you've been in so many different places of officially lived in six countries. You can probably count a few more of late, because I've spent quite a bit of time many different countries. It's related to data integration, and that's what we do now. It's really all about creating order from all the chaos. And that's what happens really if you're trying to connect lots of different systems with another. Now, on the data integration side, it works as follows. You've got a lot of different source systems with little bits of pieces of information that change frequently, and you've got ideally target systems where all the information is supposed to be going into. But that changes depending on the use case. So it's really, for me, life is all about connections, and it turns out data integration, well, clearly is all about connections. I guess that's where they're sort of found like a nice little synergy. And everywhere I go, the challenges are uniquely different. Some of them are regulatory driven. Other of them are very much driven by internal processes. Obviously, a lot of things are gearing towards how can we get data AI ready? So as a result, a lot of transformation of data needs to take place. It's quite versatile. Well, what a journey. You have been to, was there six different countries, did you say? And yeah, absolutely. It does sound like chaos, but you've got a lot of experience from the different countries that you've worked on. And now AI is everywhere right now, and many initiatives stall early. And they say that most initiatives fail. From what you see, why do so many organisations struggle to move from intention to execution? Yeah, that's a great question, actually. The intention, especially when it comes to AI, I mean, that there's so much enthusiasm around it. Still is. It seems a little bit more muted, or maybe we're just a lot more used to it right now. So, you know, I guess there's less chatch-y-b-t headlines is what I'm saying. And comparison to just a few months ago. But that's actually part of the entire adoption cycle as well, right? Because we need to understand and get to terms with what new technologies can do as well. While organisations struggle to move from intention to execution, I believe your question was. And that's in my experience or in my opinion, its companies are stuck in pilot programs. They want to prove that AI works in a sandbox environment, but then often they fail, because they haven't solved the scalability nature around it. Basically, they haven't solved the plumbing at scale to allow AI to work across an organisation. Think a bit this way. If you're building your Ferrari engine in a lab environment, that will work very, very well. And I'm using Automotives as an example because we tend to work with Automotives quite a bit. Not Ferrari, but some of the other competitors. If you're building it in that lab environment, everything works, right? Everything works beautifully. That's basically your AI that you're building. So now you're trying to drop that into a golf cart. So that's literally the legacy infrastructure and architecture that you have. The moment that still looks shiny, right? But the moment you hit that accelerator, your chassis is going to snap. So the capabilities that you're building, your existing infrastructure is not ready for what's about to come and that's, you know, that AI engine that you put. So if you're trying to put that high speed intelligence really into your low speed infrastructure, so to speak, then things are going to become problematic. Now we argue you really need to start with the bottom. You need to make sure that your plumbing is fixed. No one goes into the house for a different analogy and just says, well, how amazing your plumbing looks here. But they will comment if they open the tap and no water is coming out. Basically, data is no different to water. You want to make sure that they're at the right place at the right time and it serves different purposes. In the bathroom, it serves a different purpose to in the pool area, for example. But in the end, it's only water and from our point of view, it's only data. And you just need to make sure that if you want to really move from this intention, making AI work for the organization to really driving that at scale, you just need to make sure that that underlying foundation is very, very solid. Yeah, yeah. And I love the analogy about the plumbing because I've just sold my house and I had some drainage issues. I was dealing with the council and we've hopefully got that resolved. But yeah, I totally agree with you. If your data's not plumbed into the right place and you haven't got it centralised, then you're going to end up with multiple sources of data as well. So is it more about technology, the leadership mindset, or organizational habits? Well, both, right? Like the leadership mindset on the one side is ensuring that AI is possible and that also risks can be taken when it comes to AI. Because it depends on how you want to deal with it. But some organisations are very AI first and this comes down to, I guess, leadership mindset. Some organisations are remarkably reluctant. By the way, this is not always just leadership mindset. This can sometimes be cultural mindset. And not just company culture, but also regionally. We see significant differences. So an example in Tally App has its roots in Germany, right? So all of our basically engineering and everything actually takes place in Germany. Now from here, Germans tend to be very risk-averse when it comes to it. So AI is still a little bit of a thing we say, OK, well, sounds really interesting. Would like to do it in a sandbox environment, certainly good. However, you know, what if personal identifiable information were to go into it? Basically, it's all these sort of like scary factor, seemingly scary factors, I should say. By the way, we can remove them. But results in dampened rollouts, to be honest. The organizational habits that to an extent also come in, but this again comes, that that is a pure company culture part. And because if you've got a company that's genuinely curious, then AI initiatives are not viewed as a scary. you roll out. If you've got a and recently had a conversation about that actually in detail about what do you do if an organization is completely against it. And to be honest, in some cases, you either got a wait for these people to pass or that company altogether will pass. And this is actually something quite interesting. Gartner said, I think it was two years ago, they predicted it was 2024. Gartner said that companies that are not ready for this organizational transformational change in the next two years. Again, they said this in 2024, now it's 26. So basically, Gartner back then said, you got to be ready today. And if you are not, you're at risk of basically going out of business. And I would say back then, and back then is only two years ago, that probably seemed relatively far fetched. However, look at the advancements of even just AI models that the big guys like OpenAI and Google with Gemini, etc, have been bringing out. And the advancement of that, the advancements of capabilities that are really coming with that as well. And it's still at its infancy really in comparison to what we're going to be seeing in the next couple of years. But again, you already have either hesitation from some leadership. You've got hesitation from the wider organization. In some cases, you've got general cultural reservations. I think these are the ones that are going to be disproportionately struggling when further advancements. And they are going to be inevitable when they're happening. So again, we just need to make sure that in order to be ready for all of it, that kind of like overall foundation is taken care of. And what you're saying before about the culture and the geographical where you are, a lot of people are scared of implementing AI, but purely because of the PII, the security and so on. And I think it feels that way at the moment that most people are scared of AI, but of implementing it because of those reasons. And that can be taken care of. I mean, that's literally just architecturally nothing stands in the way of that, at least from our point of view. So we can ensure that it's actually a project we've got going on in Korea at the moment. So Korea, maybe for the listeners that don't know, there's AI laws that countries are starting to bring out that are going to be very, very restrictive. Korea is one of the countries that's actually starting to lead that front. Everybody looks to Europe when it comes to this that actually turns out Korea is leading the charge here at the moment in this respect, at least. And it's good that AI regulation is there, but in some cases that create some challenges for the local businesses there. Now PII is a big part, right? Because you want to make sure that no personal identifiable information that's what PII is gets sent to a large language model and ultimately that data is used to train the wider more. So and especially organizations need to take that quite seriously, and at scale because there's a series risk of uploading information that may fall into the wrong hands and therefore could result in less competitive advantage. Now when it comes to sending requests, we can actually on our side do a cleansing process. So and that can be done in a static way or actually we can have on-premises AI agents that are sitting within it that act as kind of like the PII police inside of it. And so let's say we want to analyze customer data that's coming in. So on its raw front you've got an auto number, you've got a customer ID number, you've got the first name, last name address, what they purchased, etc. So the parts we need to then redact and tag out of it is the first name, last name basically and then basically provide a hashing to that which basically makes it identifiable again. But when it comes to understanding that AI output. So on the top level you really have the raw data, we can then clean it, you can either do it in a static way in some cases, that's the easier part. In some cases it's actually easier to have a trained on-premises because on-premise means no data goes out. AI agent that sits really on the organization's premise. By the way that can be literally on its own surga or it can be an a virtual private cloud like if you were using an AWS or a Google cloud service or and the data can then be processed there so it doesn't leave the organization before it then gets sent to maybe open AI or somebody else for further analysis. So that's actually certainly something you can set, should probably stop there because there's actually nothing in the way. It just comes down to mindset again, right? And are you going to be deploying this or not? Exactly. Now you talk about transforming the fragmented systems into strategic assets. What are the warning signs that a business doesn't actually have the foundation in place? The easiest way to actually describe this and talk about it is talking about the swivel chair. If you've got swivel chair processes in your organization and that is literally a computer on the left you're typing in some information or maybe you're reading some information, a computer on the right and you're trying to enter that information. If these are processes that are in place, first of all, you're showing that these are very much fragmented systems and that there is probably no foundation in place. So that's probably the first part. In the end, let's be honest, these are in many cases highly paid experts, right? You have very capable team members that in some cases are forced to just become the human middleware if that makes sense. I mean, we don't have to talk about this very long to know that this is not scalable. Yeah. To you, what does a system chaos look like on a day to day operations? Double entry of data. Not really having a central point of truth or your information on your day operations. The best example is always, so I mean, as companies grow, obviously, some processes fall behind. What we see in that environment really at scale is when company acquisitions take place and we've recently had another client that bought another business. They are in the printing industry, a little niche, but nonetheless. Anyway, so they bought another print shop quite large and they've done that multiple times before, but these were smaller ones. So in that case, they were just forcing their technology on it. In this case, here, they really need to combine it all. So you've got two CRM systems, two ERP systems, two customer success service management, two or basically everything. So it's like merging two households together, right? So like, you know, which fridge do I keep? Which, you know, is your microwave better than mine, type of stuff? And then, you know, one gets, what one gets, let go and the other one, you decide to keep. But for a period of time, I mean, that's easy in a household environment because you don't really need to microwave side by side. But in a business environment where everything is actually interconnected, you can't just rip and replace your ERP or CRM system overnight. So it's quite a tedious process to be honest, either merging that data or, and then now, because ongoing, like the business just doesn't stop only because you decided to buy another. So ongoing information still goes into that system. And that's the biggest, that's really the biggest problem with it, you know, like how do you manage that ongoing flood of information? So that's where data integration naturally comes in. You need to make sure that on the one side, you can focus on syncing existing records and even with our platform, you can sync records from the past. Basically, you just, you know, set your truth date to the past and then we just start syncing it. That works. Alternatively, there's various other methods on how you can import lots of data. Now, the syncing, constant syncing of that data is really important because otherwise, you've got one sort of system of truth and I'm using this in quotations because that is on the one side might be accurate, but it really needs to be accurate on both sides because otherwise, you know, business intelligence is going to fall there. That is the ultimate chaos, not knowing where your data is, not really realizing that it could be in one place. And in some cases, also not having a strategy that really fosters this. Yes, and integration helps with keeping your data current and in one single source of truth. Does your process also do the ETL for those who don't know what ETL is? It's extract, transform and load and then structure the data and does it keep the data current and is structured. In the end, for the audience, it doesn't quite know what happens on a data integration side. In the end, when you look at the outcome, it's magic. Everything is suddenly in sync and you don't have duplicate records anymore. So, it's got all that positive effects to it. Now, what happens behind the scenes is there's multiple different ways in how you can get the data, but you can kind of imagine it like, think of this as like the integration platform walking into an office and says either in real time or in some kind that this is pulled or sometimes it's scheduled. Nonetheless, we go into the office and say, hello team, do you have an update for us on whatever we're supposed to be updating? And then sometimes nothing comes back and we say, okay, well, no update. And then we go again after this said period of time. And this could be every five seconds. This could be network initiated that if something happens, then we get the notification. Anyways, but the same concept comes up and says, is there an update? And then maybe there is an update. And then we take that. We process it. That's and where business rules come in in many cases. Sometimes the sometimes data needs to be transformed. Sometimes it doesn't need to be transformed. But one example is if one system has first name last name or maybe title first name last name in one row, but the target system where it goes to has the title, the first name, and the last name is three different roles, then obviously we need to break that apart. That's a very simple use case, and in some cases it's a lot more complex. In some cases data needs to be translated, in some cases needs to be summarized, in some cases information needs to come from three different systems to become the information in one system, and that's where we either use static transformation, so then we know it's happening every single time, or we use AI powered integrations and transformations that are basically happening right within the platform itself. And that's why AI to transform that data to make it what it needs to be. Yeah, and for people who don't know what between static and AI powered, what does that actually mean? Basically, the static part is the, and that's the example with one line, first name, last name, being together, and in another case it's broken down, so basically you would just break that up. In some cases addresses are common one, right? So you've got, in some cases, you've got the entire address, including postal code, zip code, you know, country might all be written in one long line, but in some other system these are five lines, maybe six, depending on what it is. So in that case, no, we would need to break that apart. That's pretty static because in most cases we know how addresses are structured in the US versus in New Zealand versus in Europe, for example, right? And then obviously that changes in each particular country a little bit, but so that's fairly static, because the structure of that really doesn't change. Now, in other examples, summary information about customer records, right? Customer engagements. One example is we've recently done a use case for an insurance company. Well, basically they've got call center records that are recorded. Customer obviously knows that that's recorded because, you know, I mentioned at the start of the conversation that didn't opt out, transcript of that goes live, it goes into the record, but nobody really has time to look through that. So yes, the entire transcript is there, but that's really just a long list of text and nobody has time to go through. What we then do is we use AI to summarize that, it's a pretty simple use case, but you know, the conversation with the customers is done, we're summarizing that using AI. That's on premise AI, by the way, so because, you know, insurance is a little bit of a highly regulated industry, so therefore the data cannot go out. So they decided to process that locally, then this turns into exactly the format that they need for each particular customer. So they've set up a template and they said, give me all these options. And in some cases, certain items were discussed. In some cases, certain items were not discussed. If they were not discussed, these are basically just blanks. So, you know, like not discussed, not discussed. So that if you were to go back or the customer calls again, the next agent, which in many cases is not the one that had the conversation, might you, they probably wouldn't remember the conversation. They see the notes right away, who took the call? How long was it? What were the couple of points? And that's an example of AI power transformation of taking an entire record of text and turning that into a summarized output, but at scale, because nothing, no additional clicking, no nothing, the call comes in, you're having your natural conversation about what the customer needs. They get whatever they need. And in the end, once everyone hangs up, that's when we know the conversation is over, the transcripts taken, we then go into the transcript office, the example before and we say, "Hey, is there an update?" Yes, here's a transcript that's related to this particular customer we update this year. Yeah. And so in that processing of the transcript, does it also take into account things like sentiment analysis and what was the sentiment of the customer that was calling and talking to a person? You can pull out of all that. This comes down to how you prompt it, right? And in which models you use behind it to be honest. So there's various different models that can be used for different reasons and different purposes, I should say. So it really comes down to what you need to do. You'd be surprised. I mean, the use cases we often show customers are significantly more advanced than what they actually need in the real business world because you would be surprised how much data still gets sent via email. We were recently talking to a very, very large organization in the chemical space. So they use SAP like many others and if it comes to orders, and I couldn't believe it, but well, that's the reality. If it comes to orders, putting king in the information into their own SAP system, that's all fine and fair, then they print it. Print it. Then somebody goes and scans it to create a PDF, to then email it, to whoever needs to get it and said, okay, and how many people are involved in that? No, I'm not doing the printing. So Susie does that. Susie is a made up name. There's literally somebody in the office who is responsible for these sort of admin tasks. Yeah, she's great. She also scans it right afterwards. Well, brilliant. Okay. So why do you have to scan it? Yeah, obviously, otherwise we can't email it. Yeah, why can't you just, I don't know, print to PDF or do it that way or automate it, right? Because it's the same, it's literally the same partner. So guess what, that partner uses a, well, not SAP, a different ERP tool, but we could connect to that. And that's literally what we're working on right now. Yes. How can we make sure that the order gets approved and instead of hitting the print button and not not just the time being wasted, but all the trees that are coming on top of that because there's no regulatory requirements for them to keep the physical copy of that. They need to keep a copy of it, but that can be a digital copy. So and now you've got a record in SAP, you've got a record on paper, you've got a rep code and an email trial. And in the end, what's happening on on the recipient side is they receive an email with a PDF with then having to manually key in all the data and somehow on ERP system. So yeah, that's that that seems like a very obvious one. No AI involved here. This is like literally just, you know, connecting data with another and just making sure that it's relevant in both the source and the target system and that any updates that are happening, fulfillment, notifications, etc. are being pushed back because at the moment, yes, what, that's another email. Yeah. So this brings up the point of like, where should leaders focus first to regain control and to gain that clarity in their business? I have really honest conversations with yourself and your team. Yeah. I think that's probably the most important bit because no one at 50,000 feet, you've got no idea how many emails are being sent. Frankly, you probably don't care. But like like in that petrochemical company there, leadership was quick to agree. No one can disagree with that, right? I mean, no one would say, well, that's a fantastic initiative. Let's continue printing, e-fing and emailing orders. No one says, okay, it's 2026. That's probably a great thing to do. Yeah, that is probably the very, very first thing. In many cases, leadership has no idea. Like they know what's going on. They look at dashboards and that's all fine and fair. But as long as there's a positive net increase operational, these are your low hanging fruit when it comes to not just being more competitive, but also it's and it's not just always about reducing costs. It's about increasing team member satisfaction. Someone is employed to do the printing at scanning and helping on the emailing side. But that was my other question as well as what else is she doing? She's the office manager. She does this and this and this and that. I said, okay, well, I mean, when was the last time you had an office party? No, no, no, we don't really have time for it. Well, I mean, wouldn't that be a nice initiative to maybe spend less time on printing and scanning and maybe have an office party every once in a while? There's a lot of repetitive tasks that add no significant value that are honestly just result in a series of manual data entry errors that are highly avoidable. Yeah. So first up, talk to your team and actually have a really curious conversation with them. What's actually going on? Like, I don't know if anyone's seen the, have you seen the show undercover boss? It's CEOs that are basically pretending to be the new starter. Anyway, so that they're coming with a story because obviously there's a TV camera following them everywhere. But be the undercover boss in your own organization and actually see what's going on and that that might be one or two surprises. Instant automation and integration comes up. I hope you hope you heard the podcast. Yeah. And this is why strategy and people are so important to align them and it's important to make sure that you're constantly checking in with your people to see what are some of the challenges that are happening? What's what's going on and how we can improve it? And the people that are closer to the ground, they know what's going on. They know where the inefficiencies are. So, you know, just, and they feel and they, I mean, they'll feel empowered if leadership comes and suddenly has a conversation with, I know, the warehouse for example or customer services team. I mean, out of the booths to the their morale comes and that's really just a byproduct of giving people maybe even just the attention that they deserve in their roles. A lot of interesting insights do come out of that. Now AI ready is a phrase that we hear all the time, right? And in practical terms, what does it mean to have AI readiness in your business? I guess that depends on each particular organization. I think it really comes down to, you know what, it comes down to data sovereignty. You need to know exactly where your data is, who owns it, how to access it, how to process it also legally and obviously technically, and then making sure how everything sort of comes together. AI readiness for me is a little bit like the internal diplomacy between data. So we had the analogy of the house earlier with the plumbing going into lots of different parts of the house. You can also think of AI as different nations within the UN. Every nation and every in this case, the translation would be to, let's call it the Republic of SAP. I don't know. The kingdom of Salesforce, you can name it, however you want to. The federation of on premises legacy software. So they all live in their own sort of little environment, but at large in the international community and now translated to data, they sort of all need to come together and need to make sure that each one of them, whether it's the Republic of SAP data, I think we called it, or on premises legacy data or, you know, a cloud solution like Salesforce, that all of that comes together and make sure that the data can flow really freely without major border checks. Because that's what's going to hinder it. If you've got blocks, and by the way, these blocks can be not just at the application level, they can really be at the department level. So you want to make sure that these blocks and that's kind of a leadership comes in again. You want to make sure that all departments, at least from a technological point of view, can communicate freely with another. And that's when you get the biggest value at from a, you've got so much knowledge and data across the organization, you need to make sure that at least systems can access that for analysis purposes, if that's one of the goals. Or for synchronization purposes. It's a catered and locked that down, and that's important as well. So you need to have like a zero trust, role-based access control environment, 100%, that's very, very important. But foundationally, you really need to make sure that these jurisdictions all communicate with another free. Yeah. Now, your platform in TelePASS acts as the integrator across all these systems, right? So if you normally, you'll have a lot of systems connected to each other and they're talking to each other. And it can be quite messy because it just looks like spaghetti junction effectively. And so in your architecture, your platform sits in the middle. And then all of these other systems are connected to the, to, to, to, to IntelliPASS. And so data can then flow through and then you can, you can manage that data. Now, what happens if business and have got platforms where APIs are not available or there's no direct connection available? How do you deal with those sort of situations? So the party mentioned before is called point-to-point architecture. And maybe a quick little fun stack to this year. The average organization, we're talking, you know, more towards enterprise here now, but they've got 8080 unique different applications on average. That's significantly increasing at the moment because of a new trend over the last couple of years that's called best of breed applications. Best of breed basically means you don't just take the solution that happens to have everything kind of like you kind of like this all in one scenario. But you're taking the best HR software. The best warehouse management software. The best whatever communication software, etc. And as a result of that, the tech stack becomes disproportionately bigger not just at all the time, but actually quite quickly because there's so many unique new applications that are coming out that, you know, companies need to take on board where or want to take on board where that really burdens the tech stack much more. Now to what you said before, the spaghetti architecture, that's exactly what it is. Even if we just take our 80 different applications here, right? If you wanted to connect all of them with another, it's everything with everything. You've got 80 applications times 79 applications. That's all I was in. Points that need to be connected with another. It's impossible to manage. It's not possible. It's impossible to create, not impossible, but difficult and very costly. It's very hard to maintain. And it results in also to security issues as well because, you know, these are 6,300 vulnerabilities that you ultimately have at only 80 applications provided they were all connected with another. So to your point earlier, what that we do at Intel, you have as a data integration platform is we kind of sit in the middle of it. We don't store data. We just move data. So you have instead of 6,300 connections like point to point or connected with another, you only have 80 and you basically every time you have a new application, you don't then have to connect it to all your other 80 systems that you have in place already. You only have to connect it once. Once that's connected and more on that in a second, then it's interconnected to absolutely everything else, which now means that you can use a drag and drop flow builder that we have or we've got pro code, like literally you can code everything inside as well. So basically for the pro users, we've got a very strong story and for the ones that prefer more of a drag and drop sort of scenario, we've got transformation capabilities there as well. But all that data goes through one place and it goes from any source to any target system in real time and it can go from one source to multiple target systems in real time as well. Now, how do we get access to the data? Because you mentioning rightfully so there are legacy systems, there are modern systems, et cetera, and they all connect different. APIs, many have heard can very much help, no doubt. However, sometimes we need to connect to databases and that's then literally at a ground level. You've got to be quite careful with that because databases, you don't want to make changes there unless they are very much vetted. But ultimately we can connect to databases, we can connect to flat files, something like a CSV file. Something especially legacy applications don't allow or they're not built to have real time integration but what they do is they do scheduled outputs. So for example, they would do a scheduled CSV file at 1159 PM on every given day or 3 a.m. in the morning. They then give you that data and they drop it in a SharePoint file or somewhere else, some sort of accessible data source. So we can then take that flat file and then use the insides of that for whatever processing comes with that. And that could be live updates of how a bi dashboards, this could be updates of Salesforce records or SAP records or whatever it then is you need to do with that data you can do at that moment. So it's either in real time with the more modern applications or from a database point of view or flat files, for example, and a few other scenarios here that's typically how we connect to legacy applications. Yeah. Good question. So important, especially in these hybrid environments. Yeah. Now back to the security side of things, right? When we had a conversation previously, you mentioned that you can actually install that onto into an organization so that the data doesn't leave the premise as well as a right. So you can actually securely install that implement that. Yeah, correct. We've in fact, we even have customers in the high security space. I can't mention them for national security reasons, but they ultimately are deploying the platform in an air gap environment. There's not many integration platforms that can do air gap. Benefit of this really is is that there is no data leaving the organization at all. That really means that the data stays on premise. It stays behind the firewalls. It doesn't move anywhere. It doesn't have to be air gaped all the time, right? Air gaped is a bit of a special case, but it's one that actually is probably coming out a little bit more. Yeah, what does it get for those that don't know? Air gap basically means that no, like, think of it as like the integration in a vacuum. So no data goes in and out of the internet, so to speak. So everything is behind firewalls in a very secluded environment. So if you kind of most applications are connected to something outside of themselves and therefore data needs to pass in a secure way through the internet in many cases. So we can make sure that no internet connection whatsoever takes place and that the entire integration and all the processing, including all the AI processing, can be done in a completely air gaped environment, no internet in no internet out. And it's a little funny to think about a no internet scenario in 2026, but some organizations and some parts of organizations are either very secretive and therefore they need to keep that data very secure and it's about Making you know allowing digital transformation for them or in some cases there's strong regulatory requirements on top of that as well For various different reasons some data must not leave an on-premise scenario and yeah, that's where we can also help Yeah, now critical is data quality versus vol data volume well different things happen for different reasons, right? If you're looking at lots of different transactional data that's probably data Volume for the most part it comes down to what you want to get out of it in some cases You've got lots of different I guess lots of volume of data that you want to analyze and process no problem You can do that the use case better be clear And in some cases it's the data quality you want to get out of it like you want to make sure that only finalize records For example sync to a backup source might be one example One thing we tend to So this I guess this leads a little bit towards our data lakes important as well or not data lakes are Large server infrastructures really were you storing the data for whatever they may come you don't know what you're doing with that data Yet, but let's store it and eventually we might need it now That's that's good or well and good, but you still don't have that information ready at the target system We really need it so let's pick the transactional data as an example You've got lots of e-commerce data coming in that's all well and good if you're storing it in your data Like for example, and that's then your volume coming in But you really need to make sure that the quality data which is what it Peter what did Bob what did Stephanie What did all these individual customers buy to then update the CRM record to update the customer profile To allow better marketing to allow better follow-ups. I mean That's really the quality of the data that really needs to move through so I would yeah for various different reasons you need both but The key recommendations we always have for customers is take the data that you have put it where it needs to be Connected in as much as possible, but really on top of that make sure that real-time updates are taken place so that you have a source of true Yeah, I'm just saying that's that's really important. Yeah, and can a business be an organization be a already if they're not fully automated Maybe they've just got some automation, but can they still be a already? I guess you got to start right if you don't start You know, you're not gonna get anywhere That's it. I mean start with the sandbox, but you go want to make sure that to move out of the sandbox that's you know back to the Back to the AI high-powered engine that's being created you Really need to make sure that your entire infrastructure is ready for because otherwise you're probably setting yourself up for failure Many AI and this is another management consultant. I think it was Boston Consulting Group that said it They basically said that's over 60% of AI projects are gonna get abandoned only because the AI basically only because the data isn't AI ready So now picture that over 60% of In the time and effort and these are all good initiatives, right? And they're basically just being shot down really because of lack of Structure in the end because that's kind of what it comes down to. Yeah, so from your experience What sort of mind shift do leaders need to make when approaching Digital transformation to avoid short-term fixes and create more long-term value? Clearly becomes really important. What do you want to achieve? What's the desired outcome? Honestly, and then work backwards and always have somebody in the room Excuse me always have somebody in the room who asked the scalability question because that's gonna be your north star It's very easy to get distracted by one or two nice little features because You know the finance team wants to have a new AI powered or AI enhanced some cases not even AI involved dashboard So It's easy to get distracted by that because that's obviously nice and shiny and it's a new system and it's easy to get behind that right What really needs to be taken care of is the underlying structure of that because if Otherwise we're going to be back to the point that we discussed at the start of the conversation You will not leave your sandbox environment and everything will work beautifully well in sandbox environment by the way So everything's going to be great You know leadership's going to be very impressed But then you drop it into the real world and you don't have access to real-time information You know customer demands might change whether the System has been trained in a certain way and as a result. It's kind of like stuck in that way and analogy I often make is is humans and and data sorry humans in AI have have have really One similar need For us it's water we need water to think clearly and data needs AI To quote unquote think clearly if we don't have access to clean day sorry access to clean water Well, then we're going to start hallucinating and the same thing in a data sense is going to happen to AI now You need to make sure that some core knowledge is there and that's information you can load into your AI tools No problem, but then that needs to transform over time and needs to evolve and as a result of it you're going to get long long-term benefits on that this is various I'll give you one example another Intra actually out of out of Singapore they Wanted to completely automate or help automate their customer services team incoming phone calls with requests from Intra and active customers as well as some new ones so basically just a hotline the team was trained to Trust what's on the screen because in real time the entire conversation was transcribed and in Real-time the correct answer magically appeared on the screen that sounds pretty good right because you don't need really need to train the people anymore You've got the ability to just You know anyone who can read so to speak and has a pleasant voice can suddenly be the customer success or the Outbound call manager great the problem is it didn't work at scale because the entire platform was really I guess the knowledge was to an extent hard coded to what it knew at that day zero But then after that only periodic updates happened and no real-time information really came in which resulted in Well textbook use case of hallucinization and That project very much ended up on Boston Consulting Group 60 percent. They abolished it So great intention in a sandbox environment perfectly set up however That's the shield an environment right. It's a sandbox. It's a sand castle the moment the big wave comes and that's then taken your Sandcastle outside of it and taking it to the beach well that wave comes and that sandcastle is gone so That was basically their experience They don't want to give up on the project, but they want to restructure in terms of how they're going about it So yeah, this just comes back to getting the data AI ready making sure that they've got access to the not just the live transcripts But that any sort of new information that's coming in feeds the model makes the entire model better and then at scale It can be used you know across the organization Colin innovative use case there to be honest and hopefully something also other insurers You know other other services and just pick up over time. I just it just shows that data The currency and data equality is so important. Yeah, now how are you helping leadership teams move from the reactive decisions to more strategic ones? Yeah, look we do a lot of workshops right and with that a lot of discovery comes in as well back to clarity Like we just need to together need to be very clear on how on what needs to be achieved and for what reasons Don't just add AI because you think it's a nice thing to do in fact I would argue you're better off focusing on business transformation first and really nailing that down before adding AI on top of it And sometimes that can help sometimes AI can help accelerate it but unalignedly if you're still stuck and actually okay with it with a process like print your order forms Scan them and send an email with a PDF within leadership. You're okay with that and that's just the status quo That's been universally accepted then that is really the part that needs to be started with Um my best recommendation is just us Not just ask the team in terms of you know what yeah actually we often say ask the team right but it's really ask the team What they love doing about what they're doing and allow them to be brutally honest about what they really don't like to do And the answer is probably not going to be very surprising people don't like Repetitive tasks very much people don't like re-keying data people don't like fixing other people's mistakes And mistakes happen especially when people are involved right somebody puts in the wrong field somebody accidentally Sents the confidential information to the wrong person all these things are happening on a regular basis And most of the times that's because there's human error and we can't blame people for that because after a while if you're looking at the screen for eight hours a day, at some point arrows are going to be made. So just being really curious in terms of what's there and just allowing for real feedback to come through. I think that's that's going to go a lot. Yeah, 100%. And they're the people to talk to because they know what the boring task are, the repetitive task are. And those are the best candidates for automation. So yeah, these are the simplest use to be honest, complete. These are the simplest use candidates. That's where you get your runs on the board for yourself as well because you need to understand for your own company, how import like what value are we getting out of the data transformation. Right. So that that's really important. You need to make sure the business case in the end stacks up as well. And in some countries, it's obviously easier to throw people at a problem because, you know, global wages differ significantly, especially in Asia. You've got a lot of, you know, a lot of coders, a lot of admin people doing things, but yeah, it just comes down to global competitiveness as well because wages globally are rising. And again, it's not about that. It's about staying ahead of the curve with it all. Yeah, companies that are not 100% over time are going to be falling behind quite significantly. Yeah. Excuse me. Okay, last question before we go into the quickfire questions is the transformation often focuses on heavily on systems and tools, but not the people. How do you help the organizations keep the human side and the front and center while modernizing the operations? Very much related to what we've spoken before. We don't automate jobs. We automate the dull work related to some of these jobs, right? So again, very few people get excited by the opportunity of comparing spreadsheet entries. So for the most part, that's not a very desirable task. Now again, you want to make sure that technology overall is here to assist humans. So technology should elevate what we do. It should allow us to do more and better things. It should not compete with us. It should really just assist us in becoming better at it. I have always say if the AI involved or overall if technology involved makes the people feel small and insignificant all of a sudden, then it's implemented wrong. So you want to make sure that AI just amplifies what people are doing that allows them to do their jobs better that it allows them to maybe write emails quicker like that it some people are struggling with writing professional business emails. Well, great. Here AI can help and you know, with the likes of Microsoft and Google and so on have done very well when it comes to that. That's great. Now that's on the front end. Really behind the scenes and that's kind of what we're doing, right? Like nobody really sees our work. Like nobody really sees the plumbing of a house again. So behind the scenes, you really just want to make sure that that data that is being created in real time and that's naturally being created through orders or invoices or issues like we do a lot in the IT service management space. So there's a lot of you know issues tickets, for example, that are coming in on a regular basis and then making sure that on the other side, all team members can actually understand which which is a critical issue that I need to focus on and which is not a critical issue. If on the one side, people always want their issue, it doesn't matter how big or small, they want their issue to go away as quickly as possible. So if I've got a password reset request, I may put an urgent because my boss gives me, you know, tells me I really need that to have this particular report. Why can't I have this and I forgot the password. So I need a password reset request. So me, that's very urgent, but really in the scheme of things that just shows I'm unorganized, you know, that's not an urgent request. It's load medium at best. Now, how and by the way, I'm putting critical in there because for me, this is critical. Now in the end, how can we make sure that we differentiate that because there might be some real issues like security issues or service gone down or something else, right? That really needs to be taken care of. So that's where we then deploy AI and the AI does a analysis of what is actually coming in. It puts a score against that and it's semi-autonomous, it just says, okay, based on everything that I've seen, that is really doesn't like I'm ignoring the fact that it's that it says it's urgent or I may put in one level higher because of that, but really for the most part, I'm just looking at the requested face value and the request of my access token is lost versus that's a security issue potentially versus my password needs to be reset are in completely two different categories. The token lost needs to be deactivated immediately because if that falls into the wrong hands, you've got a significant issue. The password reset can be done well after that in that sort of order and that's where we really help behind the scenes. Nobody will see that, right? That that's where the systems and tools that are in place at the moment operate too statically. They only operate in the sense of in notification comes in on the one side. Somebody says it's urgent so that means it gets mapped to being urgent on the other side, but it doesn't have to be urgent. Maybe there's more to the story and certainly in the bigger picture of it and that's where AI really can come in and help. Things sometimes look a little bit more muted. So again, it's behind the scenes data transformation and making sure that then the teams that really have to act on that really can focus on the things that matter a lot and that they don't have to manually go and sort of tickets by a seemingly higher or lower urgency. Yeah. And when you implement these sort of solutions, do you also train the people in the organization as well to make sure that they are across this and they can manage that or is that purely managed by your company? We've got a professional services team that acts globally and we've got very, very good partners internationally as well. So we often help with the initial and/or the partners help with the initial setup of that, but 100 percent it's in our and everyone's best interest to understand how the platform works. So there's different ways in terms of how you can consume the platform. In some cases, it's a completely managed service that is then managed by a local partner. So we've got a very good partner in Hong Kong, for example, to use them as an example. They are actually deploying the entire infrastructure on their own service. And what that means is that we typically work with Azure, but in that case, so all of our data typically, you know, runs through various different data centers in Azure. We recently just opened up a new one in Malaysia and Kuala Lumpur, but outside of that, we've got them in Europe, we've got them in America, we've got them in Australia, we've got them in various parts of Asia, we're available, but in Hong Kong, for example, we didn't have one. So therefore we partnered with a partner over there. And I can mention them, we're working with ASL, and ASL does a great job at making sure that they locally take care of a certain group of customers and making sure, again, that that entire infrastructure is locally present in Hong Kong, and that, etc. What I'm hearing is that you've got presents right around the world globally across various different industries. So, you know, most of the times it's a mid market and lower enterprise, but we've got some, like in Europe, for example, and it ranges a little bit by region because of, you know, the more you have in one particular region, the more you tend to stick to that. So in Europe, we're doing quite a bit with the automotive space. So some of the world's most well-known German carb brands are using the platform, and their suppliers. We've got banking institutions. We've got government institutions. I mentioned the one that's in the high security space deployed in a no-internet out type of air-gapped environment. There's insurance companies. There's technology partnerships as well that we have. There's a new technology partnership with a point of sell provider that's operating in a very unique niche space for entertainment parks that are called Rola. So Rola has thousands of customers all around the world, and we're helping get some of the data, either transforming the data that comes out of the point of sale system and turning that into, in some cases, AI summaries. In some cases, in one case, we're automating responses to customers because, you know, obviously one of the metrics is net promoters scores. No one venues really has time to do that. And so as a result, we've built some agents that are ultimately doing that. They're getting the reviews, they're waiting a little bit, and then they're creating the response to that. So these are use cases we've built. There's various others around it related to marketing and to financial sales, etc. But all together, it's all about training the partners and training the customers because that's really where the use cases are coming out. Once a particular integration or once a particular application is integrated, unlimited use cases can come out of that. And that's really then done in this drag and drop flow environment. Yes, can be coded, but the flow builder is really good where you can then literally, it's a box and you drag and drop lines, you're literally drawing lines and the data moves from whatever source to whatever target system. It's quite magical when it works. That's awesome. Yeah. And what's the typical timeline or of implementation and organization? Are we looking at the project? It depends a little on the project, but you can go live in a couple of weeks. So it doesn't have to be months. And that's really because we've got plenty of out-of-the-box connectors. So these are pre-built integration adapters, but it's very easy for us to connect new applications as well. We've got MCP support as well, so MCP service, we can natively support and that really makes a big difference when it comes to connecting to especially these new best of breed applications. But honestly within a couple of weeks you can go live with the first results. When it comes to a large of transformations, this can take a couple of months of course, but it's always about getting a few runs on the board quickly because the organization just needs to see how it works. So everything is good in a POC environment, proof of concept, but then once the rubber hits the road typically we start with one particular department and then we move our way through it. Start with an MVP and get away from there. Or start with an MVP because technically and that's probably our biggest challenge Anthony. Anything is possible because really it's only data right back to the plumber. The plumber doesn't care if you're saying I want to connect my new fridge, I want to connect my washing machine, I want to create connect my new ice maker, anything will water at some point comes in or maybe goes out that's where the plumber comes in. That's basically us. We're just the IT plumber when it comes to it. So all we do is we just connect whatever endpoint that needs to be connected and that couldn't be more different and across different departments and across different industries. They're vastly different. So the part that's often hard to explain and that customers often don't really understand immediately is, but if we've connected or if we've got use cases with automotive and with government and with insurances for example, how does that then translate to us being a print shop is like this recent example I mentioned. And the honest answer is it's only data and it's unique different endpoints that you're using. For some cases we've got out of the box connectors. For some cases we can very quickly sometimes within hours, sometimes within days connect to a unique application. We've got something that's called the flow builder. So in the flow itself you can actually build an endpoint what unique no other integration company has that. And in that environment itself you can just create your own sort of adapters literally on the fly. So there's plenty of things we often show in POC environments because that makes it real for the customer that makes it real for the partner that gives them something to then showcase mostly internally. And after that it's really starting small getting internally some runs on the board because you need some, you know, you need leadership alignment at the end of the day if you don't have that is with anything. It's not going to go anywhere. Yeah. Awesome. Thanks for that, Matt. I've been listening to this and someone says, "Hey, I love what I'm hearing and I'd like to get some more information or have a chat with someone. How can they get out of you or your team for a further conversation?" Yeah, look, you can always invite people to go to our website and telliipest.io. So that's one part for free to connect with me on LinkedIn of course as well. And yeah, the team is there, certainly there to help. But when you reach out, it'll be great to say that you heard us on Anthony's podcast. I would love that because I would then like to take the initiative and talk to you rather than some of the team. But either way, you're going to get looked after. So head to inteliipest.io and that's certainly the place where you can find more about the adapters, about the use cases, about the deployment models that we offer. Do check out some of the videos that we've created. We call them the integration packs. So these are pre-built use cases and that gives you a really good idea in terms of what can be connected with another and what sort of business outcomes are really coming out the other end. Yeah, thank you for that. We'll set something up so that people can actually go and say, "Hey, look, I've come through the champion monster collector podcast and then come through to your calendar." All right. We're at the quickfire questions and the first question I have for you is, "How do you define success and has that definition change for you over time?" For me and I think that's, you know, my upbringing and everything, it's autonomy and freedom. The freedom to be who we are, where we spend time, where we decide to live and work and I guess to an extent, I said that in the beginning that translates to the data solventory to an extent as well that I'm exposed with through work right now. But yeah, being in control, autonomy and freedom. It's been your greatest inspiration and why? Tough question. Various people and it kind of feels unfair to name some and not others because they were teachers in New Zealand for example. Incredible, incredible people, lecturers along the way. Bosses that I've had over the, I'm really grateful for my parents actually. That's again them having the courage also to send a, well then 16 year old boy to New Zealand. But that really allowed me to kind of really lead them to the unknown so to speak, right? Like it taught me to be adaptable. It actually taught me that adaptability at least for me is adaptability is my only true security that I have. Because we all don't know what tomorrow looks like. I certainly don't. And in the end, being adaptable to that, taking that and transforming along the way, it's obviously slighted a little intention with data transformation and so on. But transforming along the way is really what I owe to my parents. So yeah, they've been a great inspiration to me. What is something that you believe that others may disagree with? Well, I'm popular opinions. I've got a few, but you said quick fire questions. Cloud first strategy is one thing that many organisations have talked about for many years and obviously been very much fueled by, you know, the big hyperscalus as well. I think cloud first strategy as we romanticise that it's dead. The future is really a hybrid and a sovereign architecture. And we're seeing this more and more with countries really locking down what you can do and what you especially can't do with data. So the day will come where personal identifiable information and others around it will be viewed as in the interest of national security. And we're starting to see this. It's not fully out there at scale yet that as AI is starting to evolve and more things can be possible and different countries can fall right or wrong reasons get access to some of the data and then do things with it. I believe the day will come where countries are really starting to lock that down. So for that, the cloud first strategy as we thought about it that we can just place the data anywhere. It doesn't really matter where it is as long as it's secured. I think that that strategy is dead. The future really is making sure that the data is sovereign. Yeah, 100%. If you could go back and give your younger self one piece of advice, what would it be? Should give it that advice right now. It's a breathe, breathe more. And any laugh more. No, it's yeah, look at it. Everyone is obviously impatient in one's things very quickly and especially as a younger person, I want it things right now. And yeah, kind of couldn't wait to get older and all of that. Look, you know, now we're sort of asking ourselves where did it come from. So yeah, enjoying moments. I work a lot. I must disclose that. Yeah, taking time to breathe more is probably the advice. I am definitely giving my younger self. I'm giving myself today right now as well. Nice. Yeah, it's always good to laugh and breathe. Absolutely, breathe more. Yeah, what is one message you'd like to share with the world? Do what you can't. People always say, do what you can, right? I'm saying do what you can't. Yeah, push yourself. Do the things that are unfamiliar. If somebody says it can't be done, maybe that's the inspiration you need to hear. 100%. There's plenty of people along the way that said, well, you can't do this. You can't move to New Zealand at 16, like your family is in Germany. Well, turns out I could. You can't live and work from a hotel full-time. Well, turns out I'm doing that at the moment. You can't add another list goes on in terms of what people tell you what you can and you cannot do. So probably my biggest advice is, I don't know if advice is anything, is my message, is do what you can't. Yeah, absolutely. Just do what excites you and nothing is impossible because in the word impossible, if you break it up, it says, I'm possible. Well, here we go. Impossible, as I'm possible. So there you go. That's right. And the very last question is, what does it mean to you to be a champion and have a champion mindset? Yeah, see, I think there's a stigma with how do you define champion? You know, we are the champions, typically the ones that are standing on stage and everyone's cheering them on. For me, actually being the champion is being allowed to drive the bus from the back and not being upfront. So if, and now this kind of ties in together with everything I do, right? Because underneath it, no one actually knows that we're doing the data integration. Nobody knows that we're doing the transform, transformation. People associated to that know it. But we're not really in the line light of it. It's the business users in the line light. It's the CEOs that can show, you know, improvements of its department heads that can suddenly see transformational changes at scale as well. To be honest, for me, success and being a champion is ultimately allowing somebody else to be in this spotlight. And, yeah, again, no one ever walks into the building and complements on the pipes, right? But without them, the place becomes uninhabitable. So, we're proud of doing that. And we're proud of being able to kind of make other people the champions. And for us to provide the stage. 100% yeah, agree with you. Matt, this has been a powerful conversation. And the word clarity is actually my word for this year. And started off to see it with like getting clarity. So, what really stands out for me is a reminder that clarity always comes before capability. AI, automation and data are only as strong as the foundations that we sit on. And thank you for sharing your insights and your experience. Your grounded approach to transformation for everyone listening to this episode. If this episode is resonated with you, check out the links in the show notes and learn more about Matt and his work at Intelli. I pass and also share it with someone who might benefit from this. You know, it might be a business owner, might be an organisation, might be someone that might benefit and maybe hey, we can connect you with Matt. And it may transform the way you guys work. And so always remember the champion mindset isn't about chasing the next big shiny thing. It's about building strong foundations, making intentional decisions and playing the long game. And thanks for tuning in. And I'll see you in the next episode. Lastly, I just want to say you are loved. You're worthy. Champion your life. Champion your greatness. And have an amazing day. [Music]