AI-ready data is a higher bar than analytics-ready: trust, data products, and data freedom

David Sweenor 0:05 Hello, everyone, and welcome to the Data Faces podcast. I’m David Sweenor, founder of Tiny Tech Guides and your host for today’s show. In this show, I talk with the people who are actually making data, analytics, AI work in the real world. What’s exciting, what’s messy, and what’s coming next. Today, my guest is Matt Hayes. He’s GM of the Data Business Unit at Qlik. He leads data integration, transformation, and AI. That really all sits as the foundation for everything the analytics part of it touches. Matt’s discussed that getting data AI ready is much higher than getting analytics ready. He talks a lot about trust, so that’s what we’re going to discuss today. Matt, welcome to the Data Faces podcast. It’s great to have you here. Thanks, Dave. I appreciate the time today.

Matt Hayes 0:51 Good to be here.

David Sweenor 0:52 For those who may not be familiar with you and Qlik, can you just give us a rundown of who you are and what you do at Qlik?

Matt Hayes 1:00 Sure, sure. Hi, everyone. So I’m the general manager of the data business unit at Qlik. So I oversee our entire data portfolio of products. And if you followed Qlik at all, you know that Qlik was originally an analytics company with QlikView and Qlik Sense. And then six years ago, they made an acquisition of a company called Attunity. I was part of that acquisition. and then Qlik additionally bought Talend and some other data solutions, most recently Upsolver, which is our open lake house capabilities. So all of that together forms a data portfolio of solutions that I’m responsible for. So this includes data ingest, data transformation, open lake house, and all the fun that goes with that. So I lead product management, product marketing, and engineering teams that deliver those products for our customers.

David Sweenor 1:46 Fantastic. And Matt, the name of the show is Data Faces. So I like to get behind the people and their professional careers. So I’m just curious if you have any favorite hobbies that people may not be aware that you partake in. Because we see one side of people on LinkedIn and there’s whole other lives out there. So curious if you have any of those.

Matt Hayes 2:09 Yeah, what people do in the little bit of free time that you have in this world of AI and analytics. So basically what I do for fun is I’m a private pilot. So shortly after COVID, I finished a lifelong dream of achieving my private pilot’s license. And it’s the most fun thing that I love to do it because it’s very technical. It’s very process oriented. There’s a lot of buttons. I love pushing. So I love things that are technical, but there’s also a special feeling when you get above the clouds. And I live just outside of Chicago. Okay. When you have a cloudy day in Chicago and you get above that cloud layer and then you look and you can see the city off in the distance and you’ve got a blue sky above you and you realize you’re the only person in the city, you know, 8 million people that is having a nice sunny day that day, that’s kind of special.

David Sweenor 3:01 That’s fantastic. And is there a particular aircraft that you specialize in or that you fly over and over again?

Matt Hayes 3:08 Yeah, so I fly Pipers. So I fly a Piper Archer, which is a four-seater. That’s the one I got my license in. And when the family wants to go somewhere, I fly a Saratoga. A Saratoga, I could fit six people plus the docks and a couple bags. So those are the two planes that I like to fly.

David Sweenor 3:25 Okay, well, amazing. Well, let’s talk data and AI today. Previous to your current role at Qlik, you sort of ran the SAP business strategy. And so my question to you is, leading that SAP portion of it, what did you learn about why enterprise data is hard to move and create trust? And how does that influence how you think about things today? Yeah, that’s a good question.

Matt Hayes 3:55 So Dave, I grew up in the SAP world. I grew up as a basis consultant on the technical side, and I’ve been working with SAP since 1998. So I’ve been through a lot of implementations. I started my business originally around SAP consulting. I built a product called Gold Client, which is a test data management solution for SAP. So I’ve really just centered everything around SAP and SAP data. And the one thing that I appreciated right from the start is it’s complex. I mean, the data model itself is complex. Plus, when you factor in the different versions and industry-specific solutions and the high level of customization that people do, it really becomes a lot. So when it comes to doing more with that data, the world of analytics and AI, you need to have access to that data as well as data from all these other sources that are in your enterprise. So for me, it was about… tackling basically the hardest problem first, which is SAP. I mean, it’s the biggest market we have. They have most of the Fortune 500 and Global 2000 customers running their ERPs on SAP. So it was the area that the value of that data is sometimes the most highest value data in an organization. So my focus has been on that for many years.

David Sweenor 5:14 Okay. And I saw a bit of stalking you on LinkedIn and I saw a little bit, I won’t call it a rant, maybe just a point of view. about some announcements that were happening at that vendor’s conference. So I’d love to just give me your perspective on that. Just give people a sense of kind of what you’re thinking about data and how organizations should think about data.

Matt Hayes 5:38 Yeah, no, I mean, I think SAP is a wonderful company. I think they do a lot of things really well for their customers. But like any big software company, you can start to… broaden the scope. And as you broaden the scope, some of the solutions you bring to the table might not be the best fit for every customer. And when I struggle, the core principles that I have in running the business here at Qlik is context, trust, and freedom. And of those three, freedom to me is the most important. I think it’s really important that we empower our customers with the freedom to define their architectures, the freedom to do what they want with their data, and use it however they see fit. It comes out of their business. It tells the story of their business. What you saw in my blog was that SAP has been tempted and has executed on a strategy of lock-in. They’ve looked at things and realized, well, we own the customer’s data. or we possess the customer’s data, let’s say that. And they’ve made some strategic decisions to leverage and influence that with their customers. And when you look at things like, when you look at AI architectures, when you look at analytics architectures, these need to be flexible. Customers need to be able to choose solutions and architectures that fit their use cases and their business needs. And I struggle when I see vendors influencing those architectures by using a customer’s data against them. And that’s really what I was calling it.

David Sweenor 7:05 Yeah, I love those three principles, context, trust, and freedom. And I think that’s very important for people to keep in mind. And I totally wholeheartedly agree with you. There’s a lot of people, you know, that was the old sort of model. Get everybody locked into your ecosystem for years. And now people have choice. There’s more things out there. The world is more open and they want to be able to do whatever they want with their data and using whatever solution provider.

Matt Hayes 7:36 People want to do things with their vendors. They don’t want their vendors to do things to them.

David Sweenor 7:41 That’s great. I love that quote. So I have a question for you. You’ve talked a bit about AI ready versus analytics ready. First of all, I hear a lot of… vendors, and I see a lot of web pages and a lot of LinkedIn posts say, hey, you got to get your data AI ready. Question one, what the heck does AI ready data mean? And second part of that is… Is that harder than being analytics ready and why?

Matt Hayes 8:11 Yes, it is. It’s incrementally harder. And the reason I say that is because in front of AI, you kind of have to throw agentic in there. Because the goal in business is to leverage agentic AI to speed the performance of the business and make decisions, make autonomous decisions using the data. So imagine an analytics scenario where you’re looking at a dashboard. And I’ll just take this out of business context. Let’s just say you’re looking at baseball statistics. Okay. You happen to know, I mean, I’m personally a Cubs fan, but let’s just say I’m a Mets fan. Okay. because the Mets are struggling right now. If you looked at that analytics dashboard and it somehow showed you that the Mets were top five in stealing bases and top five in ERA and top five in batting average, you would look at that and be like, wait a minute, I know what’s happening and this doesn’t look right. So maybe there’s a data quality issue there. Well, you wouldn’t make a decision. You would pause your decision-making. You would say, let me dig into this. What’s going on? Oh, well, I don’t have all my data sources. Or maybe there’s an error in my application. Maybe there’s an error in the algorithm. There’s a lot of things that could be causing it.

David Sweenor 9:24 Someone typed in something wrong. Maybe they’re doing manual entry.

Matt Hayes 9:27 Exactly, exactly. Maybe the data load didn’t fall, like crapped out halfway through. Yeah. So there’s a lot of things that at me as a human, I could look at that and say, that doesn’t look right. Let me just make sure that the data is good. Okay. So that’s, that’s, that’s the definition in my mind of analytics ready. If I can look at it and say, wait a minute, I have to wait, see what’s going on. Now let’s look at an agentic scenario and agentic scenario. There’s no human involved unless you have an approval path with it, but you’re really, the benefit there is the automation. So. In that situation, if the data was causing an inaccurate result, the agent would execute with bad data. So now maybe the team is doing well, so now ticket prices have now doubled. Hey, look, we’re winning a lot, so let’s double the ticket prices. That could be a catastrophic error in strategy. If it uses bad data, if the agent uses bad data to make that decision. So you need to be able to trust the agents, but if the agents go haywire, if you get a bad result out of something that you’ve automated and it costs your business money, the boardroom is going to look at you and be like, wait a minute, I thought we were confident about this AI execution. I thought we were confident about this agentic strategy. Well, we got to take a step back because there was a data issue and we have to pause everything and take a look at that. So the cost of that is so much greater than in an analytics scenario. So that’s, you know, that’s the difference in being analytics ready versus AI ready. Will a human look at an analytics application and say, wait a minute, this doesn’t add up versus an agent that will just assume that the data is all good and then move forward.

David Sweenor 11:05 Right. Okay. Okay. And when you use that term, AI ready, what does that actually mean in practice for organizations? It’s like when we go back to big data. Well, what is it? It’s big, I guess. But what does it really mean, AI ready? And just a bit of context, right? We’ve been talking about data quality since data existed on clay tablets, probably. Yeah. And we talk about it every year. And it seems like, oh, this is the year we got something to really fix this. Is it a mirage? And what does it mean to be AI ready in your mind?

Matt Hayes 11:46 So to me, it’s an evolution. And when you talk about AI ready… The bar is higher, as we’ve just discussed, but you can only determine that the bar is higher if you measure it. So again, I said we’re centered around context, trust, and freedom. This is the trust conversation. The trust conversation is, can you trust your agentic outcomes, and can you trust the data behind it? And when it comes to trust of the data, this is a data quality conversation. It’s data quality, it’s data stewardship, it’s data governance. It’s everything that goes into delivering the data product that the business is going to use and the application is going to execute on. So when it comes to quality, at Qlik we look at it and say, okay, this needs to evolve. We need to not just talk about quality. We don’t just need to show a customer where there’s a quality issue. We need to give them a path towards remediation. And for quality, it’s about identification, remediation, and resolve. Identify the data quality issues you have. Can I resolve them for my use case? Or can I remediate them for my use case? And can I resolve the data back at the source so this isn’t an ongoing process? problem. So we’ve done a lot of things that collect not only functionality around data quality, data stewardship, and data governance, but we’ve also implemented something called the trust score. So the trust score basically is a customer-defined metric that’s a weighted metric for you to determine, does the data meet the quality standards for you to execute on that data? And again, the highest bar there is an agentic scenario that automates the critical business process. So you might decide that based on how you stank rank the qualities, if the quality drops below 95% or 90%, maybe the agentic execution pauses because there’s too much of a gap there and there’s too much of a risk in an error. So… I like it because it’s a leading indicator to the business and to the data engineers. Is the data ready based on what we care about the most? Availability, latency, quality. You could go on and on. Sure, sure.

David Sweenor 14:08 I love that notion. I think, you know, you said something, I don’t know if everybody caught it, but you used the term data products. And so when I said, are you AI ready? You put that in the context of data products. And what I’m hearing or inferring from that is don’t try to boil the ocean. You’re building a data product that’s built to do something. Focus on that. Get that right. Versus, hey, I got this random, like the old days. oh, we got all this data, everything needs to be pristine, but you’re saying a very specific slice of data for a very specific business purpose. Did I understand that correctly?

Matt Hayes 14:45 Yeah, that’s correct. I think, you know, when we look at the data integration business, that’s a manufacturing business. The data product is a finished good of that process. Right. So the first thing that we do is we take data from various sources or disparate sources. We collect that or we help you integrate that data into a data lake or a data warehouse, do some transformations on that data, do some data quality and remediation on it, and then make it available or publish it through a catalog to the users. And the best way to describe that is a data product because the business can give you the requirement. The data contract can define the service levels and the definition of what that data product is. So if I’m talking about… customers or sales orders, there might be data coming from SAP, there might be data coming from Salesforce, from HubSpot, from custom local databases that are on-prem. I mean, whatever makes up what a customer is, what a sales order is, that gets defined in the data contract, and then the data product is the deliverable. So if we’re in the manufacturing business, the data product is the finished good. Now, for the analytics and AI, part of it, that’s the business. So they’re the consumer of that. So the data product becomes the origination point of the experience. What am I trying to build? What’s the use case? What application am I building? What automations am I trying to enable through an agentic workflow? Well, the data product is the place that you start. And not to lead the conversation, but this is why so many companies and so many customers, all of our customers today, are talking about the semantic layer. Because that’s the context for the business. Sure.

David Sweenor 16:31 Okay. That makes a lot of sense. I like how you’re thinking about that. I wanted to, you know, you mentioned this notion about, you know, automating business decisions and there’s all sorts of business decisions. There’s these sort of little business decisions and, you know, maybe do I want to send a coupon to you as an example? And there’s biggest, bigger business decisions. So when you have these agents, are there anything like… have you seen any near misses, I guess, you know, you’re, you’re right. It can make decisions at scale and it can really screw up your business. So how do you, how do you think about this? Cause nobody wants to be on the upfront page of the New York times or wall street journal or, or pick your, your paper. Yeah. Like what, what, what sort of, are you seeing anything like, Oh my gosh, that was almost a near disaster. Yeah.

Matt Hayes 17:19 Yeah, I mean, personally, I haven’t seen much in the news. I can envision it because we’ve seen it all. Take security breaches, for example. These are things where you get these exposures that really garner a lot of attention because it creates massive risk for a business. Well, the same thing could happen around agentic workflows. Let’s say it has to do with supply chain. You have a customer that… you know, accidentally entered in an order where instead of buying 10,000 of a certain product, they accidentally put in that they’re buying a million. Now you’ve got some agentic workflows in there, and if that outlier doesn’t get caught as a data quality issue or doesn’t get flagged as a data quality issue, you could have an agentic workflow that then sources raw materials from 3, 5, 10 vendors, and all of a sudden you’ve got ships being loaded and trucks being loaded with way more raw material than you need. And if that stuff goes unchecked, if it doesn’t go through a data quality check, if it doesn’t go through a human in the loop check, there’s definitely the risk that you could have, I don’t want to say an exposure, but you could have a major error that materially impacts the business short term.

David Sweenor 18:37 Well, that brings up an interesting question then. So we talk a lot about this notion of human in the loop. And if we’re making lots of automated decisions at scale, how do organizations even approach human in the loop? Like, do I just keep stamping? Yes, yes, yes. You know, you kind of get blind to it a bit. So like, I’m just curious your perspective on how people should think about that. Because if you’re… reviewing every single decision that comes across your desk, that you’ve lost that scale, perhaps, to a degree. So I’d love your perspective on that.

Matt Hayes 19:23 I mean, I think everything around AI is about reshaping the workload. Yeah. and creating acceleration for the business. And you’re right, it can’t be blindly done. I mean, there are a lot of things that you can factor that are low risk where you can trust automation, but there’s a lot of things that really need a human in the loop. We look at it when we do our development because we’ve shifted to an AI-focused methodology for our software. So that means that we’re delivering software faster. We’re coding faster. People are less coding hands-on. Everything’s moving faster. So if things are moving faster, then everything that supports it needs to move faster. So if we’re delivering code faster… We need to be faster with our code reviews. We need to be faster with our testing. We need to be faster with our documentation. So I think everything around agentic and AI is designed to accelerate. So I think the businesses need to adapt to figure out if everything’s moving faster… What is it practical for me to have my hands on? Where is it going to slow me down if my hands are on it? And where is it critical that my hands are on it so that we don’t have a major error? And I think this is something that’s probably rippling through every organization in the world and every department in the world. Because when you look at adoption of this technology, it impacts everybody differently. You know, us as a software company, we’re intensely aware of how it affects our software development lifecycle. Sure. That’s because we build software. Even internally, when we talk to finance, when we talk to legal, when we talk to our product management team even, we look at that and realize that there’s different ways that we’re accelerating adoption and different checkpoints you want to put in to say, where do we want to safely automate? Where do we need to be involved? And how does human in the loop involvement actually enable that acceleration as opposed to hindering it? Okay.

David Sweenor 21:28 I’d like to maybe come back to this notion of trust score. I believe Qlik was an early pioneer or may have even invented it. Yeah. I’m imagining there’s a bunch of trust scores in your enterprise system. And I worked in manufacturing doing yield characterization. So we had this dashboard that said… How many tools are out of control? And everything was a red light. And if you calculated the number of engineers and the number of red lights that you had to analyze, there wasn’t enough time in the week to look at these thousands of red lights. So I’m curious, how do organizations set these in a way so I guess they’re not ignored? Because there’s lots of things that can go around. How do they know which ones are the most important and they need to be remediated or taken action upon ASAP?

Matt Hayes 22:28 Yeah, well, I think anything that is a KPI, so anything that’s a key performance indicator of how something is performing, you need to be able to measure it, and it needs to be meaningful. Like, if everything’s red, it’s not meaningful, right? Well, yeah, that’s my point. You just ignore it. Yeah, exactly. So this is why, and again, when you take the trust score, and yes, many companies out there have copied our trust score. We’ve had the trust score for a while. It was part of the Talend Data Fabric product that we acquired, and it’s a big part of what we do with Qlik going forward. Other companies have done it as well. It’s a feature that you could almost argue is a gimmick, but it really is a feature to let you create the benchmark. It’s not perfect, and obviously it’s not designed to be 0 or 100. It’s designed as a way to give you an indication on what’s the overall quality of my data for this use case, for AI readiness. And if I decide that any data that’s more latent than three hours, if it’s more than three hours old, it’s not good anymore. That’s going to hit the trust score right away. Bam. If I’m pulling my customer data from five different source applications and one of those pipelines is down, that’s going to be measured in this. That’s going to be noticeable. So latency, availability of sources, so completeness of data. You could get into the data quality and look for individual characteristics in the data. if I start seeing letters in my pricing, well, that’s a problem. I’m getting silly there, but you could actually get into some large things that move the big needle on it, and then you can start really drilling into the specifics of the data quality. to decide, hey, can I act on this data or not? Is this good enough or not? But if the company agrees on it, if the company agrees on the context of that data, the health and readiness of that data, the quality of that data, then you can implement measures in Qlik. to continually measure that data and make sure that it meets the quality standards. So the KPI is adjustable. So it’s really important that every company really looks at it and sets the data quality metric to something that makes sense for them.

David Sweenor 24:54 Okay, I love that notion. I guess when I was doing field characterization, we didn’t learn that lesson yet. It took a while.

Matt Hayes 25:03 Yeah, it’s kind of like when you said that, I started thinking about scrap because when I started implementing SAP, I would start getting educated on the processes behind why things are fresh. Well, you know, if scrap, everybody knows that, well, in manufacturing, if scrap is 2%, that might be perfectly manageable. But if scrap exceeds 5% or 10%, well, that’s cutting into your cost of goods sold. And that’s something that you have to look at and be like, whoa, this is materially impacting our margins. We need to improve that. And again, in an analytics world, that is a use case.

David Sweenor 25:36 Absolutely. So we had one of your colleagues, Brendan Grady, was on and we talked about trust scores and consequence management on the decision side. You sort of own the data and integration side. Where do these sort of have to connect and meet up for an agent to be, you know, trustworthy end to end, right? You got both halves of it. We have… sort of the plumbing, if you will. I don’t know. That’s probably not a great term. And then like the visible part of it, but they need to connect somehow.

Matt Hayes 26:07 How do you think about that? This is my favorite part of this business right now because I love telling customers that like we were built for this moment. Like right now, when you look at Qlik as a company, and I’m just going to paint the broader picture here because this is how I see it. We have the rich analytics history in our analytics engine. That’s the part of the business that Brendan owns. But it’s very, very feature-rich. Customers love that solution and love that product. The data integration is all about providing the data, getting that data in that… Getting that data in an area, in a format, in a way that you can leverage it for analytics workloads. Well, now the AI world comes along and now it’s like, okay, well, now the business cares about itself. Like we care about each other. It used to be that we would integrate data and then say, well, if you use Qlik Sense on the other end of this, that’s great. If you don’t, that’s fine too. Right. Now I care. Now I look at it and say, look, this data is better if you leverage Qlik Answers and click analytics with it because it creates the AI story. So a lot of our product releases this year are around agents. We just today released agentic data engineering. So it’s a series of three agents that we’ve released to the market that are productivity agents for data pipelines and for the data engineering community. If you look at analytics, we’ve released Qlik Answers, which is really the connective tissue between an LLM and your enterprise data. And those features are also available through MCP tooling. And then you’ve got the discovery agent, the predict agent, the analytics agent, all designed as productivity agents to really improve what you’re doing with the data. So this is the point in time where Qlik was built for this, between the analytics front end… and the data backend, and it’s just kind of grown like this. Well, now there’s aspects of our business that Brennan and I look at and say, this is the same thing. Data quality, data products, that’s a handshake moment. Semantic modeling, that’s a handshake moment. And if we’ve built our business for this and can message it correctly to our customers in the market, it’s ripe for what’s happening right now. And then I would back it up with a freedom concept because we don’t persist your data. I mentioned earlier with vendor locking. We don’t persist your data, so there’s no chance that we’re going to try to use it against you. So we look at it and say, we’ve got a customer-focused product mindset right now that we can back up. Everything that we’ve built is designed to create cost efficiencies in your AI investments. We can lower your storage costs. We can lower your compute costs. So we look at this and say, and we can prove this to customers because we don’t store their data. Right, right, right. So we feel that we’ve got a message that really resonates with the market right now. And it’s something that I think is unique and powerful. Yeah, I think so as well.

David Sweenor 29:09 I’d like to maybe… You’ve mentioned context. You’ve mentioned semantics. And sort of for our listeners who may not be familiar, but there’s sort of like this notion of you want to… Give all your knowledge of your business to the LLMs and these agents so they can not be generic, I suppose. But how does Qlik approach this? Because every organization has this giant spider web of data all over the place. So how does Qlik approach context and semantics with the agent and the work you’re doing?

Matt Hayes 29:49 Yeah, so I think the most important thing that we can do is meet our customers where they are. So as customers are looking to combine their enterprise data with data from an LLM, or sorry, not data, but combining their enterprise data with an LLM to get context-rich insights and answers… This is something where we have customers that are really concerned around data sovereignty. We have customers that are really insistent on on-premise solutions, banking, finance, pharmaceuticals. These are companies that are heavily regulated and they look at this and say, we’d love to be doing this, but we have to do it on-prem. We have to do it in an incredibly sovereign environment to reduce risk. And at Qlik, we’ve engineered it such. We have customers that are all in on the cloud. That’s great. We can meet them exactly where they are. We have customers where they tell us that even for the data integration, they want to do that all on-prem. They still want to do that all in their data center. We can do that as well. Another release that we had come out today, you caught me on a good day, David.

David Sweenor 30:57 I had a good day. I should have read the newspaper.

Matt Hayes 31:02 So one of the other releases that we had today was Replicate support for open lake houses. So Replicate is our on-prem data replication and change data capture solution. Hundreds, thousands of customers use that product. And they’re moving data from on-prem systems into on-prem data warehouses or cloud data warehouses. But our support of OpenLakehouse means that customers can actually replicate from those on-prem applications and databases into a lakehouse, into an iceberg OpenLakehouse data lake that’s on the cloud. So, again, we’re… All in the cloud, great. All on-prem, great. Hybrid, great. So I do think that a big piece of what we’re doing is making sure that we’re meeting our customers where they are. I hope that answered your question.

David Sweenor 31:51 No, it did. That would be a perfect answer to it. And we’re coming close to the end of time, but I want to ask this question. So there’s a lot of AI experimentation out there. A lot of projects, I’ll just say, have failure to launch. Yeah. So you have a prospect or an existing customer, a data leader, they come to you, say, I want to build this AI business case. Sort of what’s the metric that you tell them they should really focus on to prove that value versus one they may have in the back of their head that they should probably stop reporting?

Matt Hayes 32:27 Well, I think I’ve met a lot of customers recently at our Qlik Connect conference, and we’re doing this AI reality tour around the country right now or around the world. I’ve met a lot of customers that have talked about how their executive teams and their boardrooms are pressing them for results with AI. And a lot of them have defined use cases, but many of them don’t. And they’re just kind of flailing out there. And I think this is where we’re seeing a lot of reporting of AI failures, okay? And again, I don’t think they’re failing. I think it’s experimentation. I mean, I jokingly say that the evolution, the maturity level of AI right now, we’re at the part where the iPhones came out and we were all walking around with lighters on our iPhones, pouring beers. Yeah, exactly. And that was really freaking cool back in 2007. So we were all enamored with that. Well, that’s kind of where we are with AI. Yeah. So I think that the acceleration, maturity, and adoption is going to move at a very rapid pace. And I think that the experimentation phase is part of this journey. It’s part of this maturity. So I think that customers need to really look at this. There’s a couple big things. One is… Readiness of AI data. Yes. We need to make sure that we’re doing… We need to make sure our data integration solutions and transformations are delivering the value that we need so that we know that the data is good and ready to act on. So the trust is really key. Context. Make sure that you’ve got those data products set up so that you can use them, that you’ve got a semantic layer and a glossary so that the business is talking about the same thing. So the word revenue means the same thing to… five different departments, you know, you know, I know there’s probably people watching this nodding their heads right now, but that, that happens. Um, and then, and then, um, be willing to work through that experimentation to get to the true use case. So you might have to iterate something three, four, five times to get to the use case that’s meaningful to the business. The challenge in that is how do you get there without the cost scaling out of control? And we’re starting to see that. We’re starting to see storage and compute costs. start to scale too high, and you have CFOs going, you know what, you guys are having a lot of fun figuring out this supply chain issue, but just so you know, it’s costing us a million dollars a month now. So you get to the point where the CFO might stop. an experimentation project. So you have to be responsible with the architecture. You have to respect the freedom of data, because if you don’t really lean into the freedom of your own data, that will cost you. It will cost you in flexibility, it will cost you in infrastructure and architecture, and it’ll cost you more in the long run overall. So lean into that concept of freedom. Take the position with all your vendors that this is our data and we want to do a lot with it. So build those architectures that support that. Invest in things like Iceberg and Open Lakehouse that are going to reduce your storage footprint. Invest in solutions like Qlik that can help you do that. With Qlik Analytics, again, with the engine and analytics, we can process a lot of that compute power in memory within Qlik Solutions, which reduces your storage or your compute costs. So… Again, our message is really to try to help customers safely navigate that experimentation, do what they need to to make sure that the data quality is there so that they get good results so that they can move on to the next step and help them control the scaling of their costs so that they can actually get to the meaningful end result.

David Sweenor 36:17 Excellent. Matt Hayes, Context, Trust, and Freedom, General Manager of the Data Business Unit at Qlik. Matt, you’ve been an amazing guest. Thank you for joining the Data Faces podcast.

Matt Hayes 36:29 Thank you, David. I appreciate the time and great conversation. Thank you.