Good afternoon everyone, and welcome back to theCUBE's live coverage of Qlik Connect 2026 here in Orlando, Florida. I'm your host, Rebecca Knight, alongside Rob Strechay. Rob, we're in the home stretch.
This is awesome. I've been so energized. I think the combination of customers, partners, and the folks from Qlik looking at how data becomes AI and becomes real outcomes has been wonderful.
It has. It has, and we've saved some of our best guests for last. I would like to welcome Mike Capone, the CEO of Qlik. Thank you so much for coming on the show.
And Ed Dunger, director of ops, tech, and automated cluster, and analytics enablement at HelloFresh. Welcome, Ed.
Thank you very much for having me.
So up on the main stage this morning, you were talking about the questions around AI and data that you're hearing from CEOs and boards right now. Is it real? Can I trust it? Can you prove it? And how much is this going to cost at scale? Talk a little bit about what you're hearing, how the conversation is evolving today.
Well, look, the big conversation really has pivoted. AI used to be a novelty, but now, it's really around value creation. Companies are spending a lot of money on AI, and it's a lot of experimentation. And that was fun in the beginning, but now, CFOs are weighing in, and they're saying, 'Where is all this money going? How am I getting value?' So what I hear from customers a lot is, 'Help me achieve value with AI. Help me show some real business results.' And what they're also saying is they want their existing partners to actually help them. They don't want to go rip and replace everything that they've already done. They trust partners who've built with them before. And so, saying, 'Please, please, please bring AI to us. Help us navigate this journey, and help us create business value.'
So, Ed, I mean, from an operator's perspective, when you take a step back, everybody was like, 'Wow, AI is just so impressive.' But now, it's like, 'AI is accurate. It's got context. It's connected. It's scalable.' Getting to that point, how was your journey, and how did Qlik really fit into that journey?
So the journey that we've been looking at is how we manage our exceptions in our logistics network, and we've been using Qlik Predict to help us power identifying where deliveries are not going well, and then using that to turn it into actions. So rather than just someone being informed, actually, how do we then retrigger production to send someone another delivery, or how do we send them communications so that they're aware that there may be an issue with their delivery that day? And that's really made us more effective. And I think one of our core principles at HelloFresh, one of our DNAs is customer centricity, and I think that really weighs into that, that if you can communicate with a customer faster and quicker, you're going to retain that customer for longer.
That's great. One of the things that we were talking a lot on the main stage this morning was trust. How hard or how challenging was it to make your workforce feel that they could trust the data that they were getting enough to be able to take action?
So back to our logistics example, the trust has really been built across not just tech telling our operators what's going on. It's been a two-way engagement with our business stakeholders really being involved, and then just proving really reliable data. The business has to understand where the risk it wants to take with AI. So for us, if we trigger a certain recovery process, we know that sometimes it will predict it incorrectly, but as long as that far outweighs the business cost, and those business stakeholders understand that it will occasionally get it wrong, it doesn't need to be 100% perfect for our business use, that trust gets built quite quickly. I think it's when people come with a narrative of it always has to be 100% right. And also, the secondary question is, how good was your process it's replacing? Because you sometimes need to measure that and see... You make an assumption that your existing process is infallible.
That's a very good point.
People forget fast though, right?
Yeah. What it used to be like.
Humans are also prone to hallucinations and mistakes.
Yes. There you go. Well said. But just going back to what you were saying, Mike, about this growing pressure where the CFOs are saying, 'Hey, wait a minute. Show me the money here.' Where are leaders going wrong when they try to measure the return on AI? Is there something that they're not thinking about, some sort of blind spot that they have, would you say?
Well, I think it's actually how they actually start the process, because I think you guys did a great job, because you identified a problem in the business and somewhere where an operational improvement could really move the needle, both in terms of top and bottom line, but also in terms of customer centricity and satisfaction and building trust and repeat customers. I think a lot of times, people don't do that. They want to skip to the end. It's like, 'I'm going to plug in an LLM, and we're going to put all our data in it, and all magic stuff is going to happen.' Right? And what Ed did and what HelloFresh did, which is the right way to do it, is, 'Let's take a measurable business problem, and then let's figure out how we can apply AI to this and then work backwards from there. Okay. What data do I need? Then how do I get the quality of that data right? How do I build a trusted AI data pipeline to do that, and then actually hold yourself accountable? Did we really make progress?' The problem is, people go out and they spend a lot of money, and then they're afraid to admit that it didn't work, because it's like, 'I got to justify all this spend.' Right? But if you actually start with a process, you invest, you measure, you test, you improve, and you get a good outcome, you're going to be fine. It's just all the pressure. All the pressure. And look, I'm guilty. CEOs and boards are super guilty, because they're like, 'Do something with AI. Make something happen. Why aren't you moving faster?' So everyone's running, but the reality is measure twice, cut once. It's the same story over and over again.
Yeah. I liked it. We were talking with James earlier about sometimes you have to go slow to go fast and things of that nature, which I loved that I've used it a couple times since, because, to me, it is all about that, especially, I mean, HelloFresh. Right? I mean, your scale has to be a key consideration of what you're doing. So it's not just about connecting all the data, but it's connecting it at scale with governance and trust, and so that people can rely on that. How did you challenge with that?
So if we look at the exceptions management in the logistics network, I think, again, it's that framework behind it. So we understand what those SOPs are if the customer is not in, what happens with that box now. The box gets damaged in a courier's depot. As long as you have that good framework and guidance around it, AI can then drive and power off of it, and I think we invested in standardizing that terminology across the business. And quite often, you're both talking about exactly the same thing, and we're now currently globalizing our Qlik solution across all of our GOs to help bring that standardization across the business. So we all do it in a similar way, but get the benefit of that AI helping to speed up decision processes. So what we found was, when we left it to a human, they would quite often wait, because they wanted all the facts to make a decision, whereas AI sped that decision up, because it goes, 'Well, I know the decision I want to make. Let's execute it.' And quite often, by delaying, you therefore didn't actually give the customer that great experience that you could have given them if you just let the AI act within, again, those boundaries that you've set for your risk parameters.
Interesting. So the human is hearing the customer issue, but needs to gather more information in order to make it right for the customer, but the AI knows, 'We're going to make it right for the customer, and here's how we do this.'
Yeah. That's what we found, so that the staff were not feeling confident even though they had enough information to make a decision that would, nine times out of 10, come out with the outcome that we'd want to achieve in that exception workflow. So yeah, it's surprising.
Well, really, because it's a segue into my next question, which is this idea of a chief context officer that you brought up, because it is about knowing the context and having the guardrails and knowing how to make decisions. Talk a little bit about how you see that. I mean, there really is no role like that right now in most organizations, but maybe there should be.
Well, look, I think everybody should be chief context officer. Right? I mean, look, it's somewhat fundamental. Every data point in and of itself without context is pretty useless. Right? So you can say, 'Look, we predicted this with 90% accuracy.' And you say, 'Well, that's not very good. It should be 100%.' It's like, 'Yeah. But last week, it was 30%.' Right? So, again, that's good context since we've made a massive improvement to our business. And yes, it's not perfect, but contextually, it's like, 'Okay. We're doing pretty well.' And that's what actually Qlik's superpower is, is that we're able to bring context because of the way our technology works. We can see patterns and data that other people can't see, and we can look at things, and we can put things inside of a context and not just present data in isolation. And when you start getting into a world of agentic where now you're getting insight and you're automatically taking action on it, context is really, really important, because without the right context, that agent could do something you don't want it to do. And then all-around data quality. We're going to talk a lot about data quality, but you have to have good data quality in order to make these decisions. And so, one of the things that we do really well is we monitor the quality of the data, and we're to a point now where as these models are built and the agents are fed, if the data quality starts to drift, if we trip certain things, we can turn it off, say, 'Okay. Stop making decisions on this. We no longer think the data is of the right mindset.' And again, it's about context. Is this data really good enough to make an automated decision that can fundamentally take action in your business, change pricing or redirect to delivery? That's where the world is going.
So what are some of the characteristics? I'll ask you, and then from your perspective as well, starting with you, Mike, is, what are some of the distinguishing factors that people lean back on that organizations can say, 'Okay. I think we're starting to go the right direction with agentic and starting to think about it, and here's how we need the platform to help us going forward'? What are some of those key aha moments... Because, to me, it's about understanding, to Ed's point, around the process and understanding that and, 'Hey, we can rely on this.'
Look, I mean, I think it starts with just the realization that there are no shortcuts. Right? The aha moment for most customers is, 'Oh, wait. Claude didn't just fundamentally transform my business.' Right? And it's all about the foundation. Oftentimes, customers actually have to fail at something before they actually understand, 'Okay. We really do need to go back and build a trusted platform with our data. We have to build all the AI data pipelines and put all the governance and quality controls around that before I can start really using AI responsibly.' And you can see the lights go off with customers. Unfortunately, it's usually after some pain, but they come back, and they approach it the right way. And all of a sudden, they start to see repeatable success. One of the things I said on stage today, in addition to, 'Can I predict the cost?' et cetera, is, 'Is it durable? Will this continue to work over the long run, or is this just a point-in-time solution?' And durability is something we talk to customers about all the time. It has to be durable.
But also flexible and adaptable too over time.
Yeah. Well, if you build your foundation right, if you build the foundation of a house wrong, the house is going to collapse no matter what. Right? If you build it right, then you can add a room. You can redecorate. You can do all those other things, but it's hard to change the foundation after the fact. Right? And so, that's kind of where we remind customers of, 'Look, LLMs, it's all great. It's all awesome.' Look, they're going to leapfrog each other. There's going to be change and evolution, things like that. But someone gave me the analogy. This isn't mine. I'm sorry, I'm quoting somebody. I don't know who you are, but somebody told me this. The car's changed, but the gasoline stays the same. Right? And data is the gasoline, and good-quality gasoline is going to run cars, and you could have a Ferrari this week, something else next week. That's kind of how we think about this problem.
I'm going to take a slightly different tack to Mike, and I'm going to say, I think quite often, people will try and overengineer and try to write massive code from scratch. One of the amazing things we had with Qlik Predict was, within a day, we had a working model that it had refined, deployed itself. So using platforms which have already done the investment legwork for you, I think, is the right strategy. Quite often, people just get drawn into going down over the complicated architectures and approaches, or they want to show how well they can code. So they want to code it all from scratch. I'm quite happy for Qlik Predict to do 90% of the work for me. And within a day, I can have an operational, deployed model up and running and starting to deliver impact back to the business.
I don't need to justify six months of a data science project team. So that's probably my different take on it to you, Mike.
Yeah. Well, it's different, but it's the same, because, actually, you must have done something right with that data before you put Qlik Predict on top of it to be able to get a good outcome. Right?
We did invest in structured data, understanding that data. HelloFresh has grown exponentially from COVID. So we do have tech debt at times. And so, our data isn't always perfect underneath.
