Why Trust, Data Quality, and Governance Matter | Utilizing AI Podcast

We talk a lot about AI in, well, the real world, but also, of course, in our weekly podcast. But this week, we are at Qlik Connect in Orlando, and we are talking to a group of data and analytics experts about how AI is impacting the data and analytics and business intelligence field. " Welcome to "Utilizing AI," the podcast about applications for enterprise AI from the Futurum Group.

Every Wednesday, we discuss news and use cases of the ways in which AI is implementing enterprise IT and the industries it serves. I'm your host, Steven Foskett, organizer of the Tech Field Day events for the Futurum Group. And before we get started with our conversation, let's meet who's on the panel today.

Hi, my name is Sydney Jarrell. I'm a solution director at Qlik. I can tell you more about what that means, solutions.

We'll get into it. Otherwise, I'm an animal advocate. I'm absolutely obsessed with my dog and my cat and pretty much any furry creature.

In fact, we're at a hotel with alligators, and they're my favorite thing. But they don't have fur. I love alligators, too.

Hmm. I'm Frederik Van Haren. I'm the CTO and founder of Hyfence, and we provide consulting and services for AI and HPC applications.

So, as I mentioned, we are here live with this episode at Qlik Connect in Orlando, complete with alligators- Yeah ... and indoor alligators, and of course, also complete with a whole crew of data analytics and business intelligence-focused folks who have been living the life of enterprise data for their entire careers. Now, many of the folks who listen to "Utilizing AI" are not as familiar with data and analytics and what all that means.

So I think we should start off by just asking Sydney, tell us a little bit more, what exactly is the data analytics market and- Yep ... who do they serve? Well, the answer is everybody.

Because if you think about it, everybody is using data in everything that they do. Anybody who's ever opened Excel, which is probably every single person in this room or in this kind of group, is using data to make decisions. And so what we're really focused on is, one, how you get the right data to people.

The analogy I like is plumbing. There are a lot of different data sources. There are a lot of different places that data comes from.

Data comes from trucks. Data comes from fast food cashiers. Data comes from all over.

So how do you get that data into one place safely? But then secondly, how do you access that data? How do you make decisions off of that data?

We often think of the terms of dashboards. There's so much more that happens behind the scenes, and right now, we're at a conference of people who are involved with every single step of that journey. Yeah.

So what kind of challenges do you see in the market with people that are looking at Qlik? There's so many challenges. Data is complicated, and you want data to make decisions.

But sometimes getting data at the right time, getting data to the right place, getting data to the right people, it's really challenging. And so you have to make sure that you have the framework in place to do that. But then also, we've been talking about this for years, but data literacy is a real issue.

Just having access to that data isn't enough. You need to be able to understand it. And in the world of AI, that gets so much more complicated because if you think about it, when you go into ChatGPT, when you go into Claude or Gemini, where is it getting those answers from?

And do you trust it? So maybe the biggest answer is just there's so many challenges underneath trust. And that, I think, is one of the big takeaways for me coming into this.

When you're talking to data people, they have been dealing with issues of trust and issues of data provenance and data governance and just that whole sort of gardening and pruning and maintaining data infrastructure forever. And now everybody in the AI space is suddenly faced with these same questions. Because, as you said, when it comes to AI applications, a lot of us have a lot of questions about what's going on here.

Where is it getting this answer? How come it says this? Is this true?

Is this verifiable? And these are the questions, especially in the business intelligence world, that have been the focus of thousands, millions of people's careers for many years. You're seeing the collision of AI and the data and analytics world right now.

How does AI, I guess, challenge and change your own perceptions of your industry? Yeah. So I'd say that we talked about it this morning earlier, that before, data really mattered to a handful of people, the analysts, the data engineers, data scientists like you, the developers.

But now all of a sudden, everybody wants access to that data at their fingertips immediately. And so it's changed the question of accessibility. How do we ensure that...

It's just we want it now. In the world of smartphones, we want everything now. And now with data, that's exactly what we want.

I was talking to a sales manager this morning, and he goes, "I have five minutes of my day that I'm going to look at data. I need those five minutes to tell me everything that I need to know. Who am I talking to?

What is our pipeline? What are our deals at risk? And I've got five minutes.

So if my application, dashboard, platform, whatever it is, isn't doing that that quickly-I'm doing something else, or I'm not making the right decisions. I'm wasting my time. Right.

Yeah, data is important. In the early days, a lot of people spent a lot of time massaging the data and figuring out what's good and bad. The challenge with bad and good data is you don't really know unless you process the whole cycle.

So I presume that with Qlik, you have mechanisms to kind of help customers identify the incoming data. Is that good data? Is that bad data?

And then I think that the second problem is security and compliance, right? Yeah. The market is growing so fast nowadays that the data is already in the system- Yeah ...

when you realize that maybe from a compliancy perspective or private perspective, it's not okay. Are there mechanisms that you use to approach this problem? Yep.

And you hit on a really good point. You know what they say, garbage in, garbage out. And so we have a lot of structures in place, but two things that I can call out, one is just data quality.

So Qlik has a number of ways that we automate data quality, so we can feed that back to you. " But also now we're starting to leverage AI to do that for you. Because, the con of AI is there's a ton of stuff out there, and do we trust it?

But the pro is that, we thought Google was good for getting all the knowledge in the world to us. With AI, it's even better. So really focusing on data quality so that we know that the data that comes in is not garbage.

The second is governance. And what I would say about governance is this is where organizations need to take a beat. Because oftentimes, everybody is rushing so fast to get to the forefront and we hear it, you can't look at your phone without something about AI or agentic, and organizations are so eager to catch up.

But you need to make sure, especially with so many of the organizations that we work with, healthcare, government, pharma, that you have the governance in place to make sure that your data is secure and to make sure that the people who have access to your data are the ones that are supposed to have access to it. And that's foundational to Qlik. Ever since the forefront, we have number of ways, both on the data integration, that plumbing side, as well as the dashboarding or data analytics side, to make sure that the data is only going to the people who should be able to see it.

And not just the people, but the agents and AI. True. I know it's a shift.

And that's what's really exciting is that-- And when we talk about kind of taking a beat, and what's really exciting for all the people who are at this event right now, they've invested all this time in building out the governance to make sure that the right people have access to data. And now you can just turn that on for agents as well. So it's not completely reinventing the wheel.

And again, you need to think about what data does this agent need access to. Is it everything, or is it just a subset? Somebody in EMEA doesn't necessarily need to know what's going on in America.

But then they can turn it on, and now the agents have access to the right data as well. Yeah. It has been in the news, like with OpenClaw and others where security was kind of a- Yeah ...

secondary thought simply because the market- Yeah ... was pushing for the technology itself. It's for people to consume AI a lot easier, and OpenClaw does that perfectly where you don't know anything, but it gives you some kind of information.

I think the concept of agents are great, but they also accelerate all of these problems, right? And it's more like a personal question I have in general, is with each of these agents having its own wishes to have access to everything- Yeah ... how do people kind of allow all of these agents to synchronize with each other in making sure that the gold doesn't get out, so to speak?

Yeah. I wish I could answer. I feel like this is a huge question, and I can only talk about how we are approaching it.

But to your point, you have two challenges. One, everybody wants access to everything. And two, there are a lot of agents that are built out, but they're not scalable.

They're not reliable. You build out this amazing agent that does something really cool, and then it breaks on a different use case, or it breaks the next day because it wasn't built to work past a single use case. And so what we're doing at Qlik is we're focusing on how do we take our existing capabilities and add agents to those.

And that means that we're building at a very kind of reliable, scalable platform. So a good example, we have what we call Qlik Predict. And that is a very lightweight, nothing like you, Frederick, but a lightweight predictive or data scientist.

And so we're taking Qlik Predict and we're adding an agent to it. And what that means is now, A, you can call it from other platforms. But B, as a user, I'm not a data scientist, just a math major, but not a data scientist.

As a user, you can now ask questions in normal speak. What is our on-time delivery? What are we expecting next year or next quarter?

And it's built off of that existing foundation. So behind the scenes, somebody has done all this work to get the not garbage in. You layer this on top and you know it's going to work because it's built off of this scalable foundation.

And importantly, too, I think that the way that Qlik Predict with the agentic and AI interface, I think the way that it's working is it's actually constructing a very conventional- Yep ... query. And so it's not just, here's a flood of data, take your best guess.

Yep. It's actually doing what a data scientist would do-And constructing that query for you. So in a way, it's a lot more like Claude Code and a lot less like ChatGPT, because it's building something the way that a data scientist would build it.

And I think that's important because to people outside the data space, I think that they don't maybe have this understanding that there are people who have been very carefully curating this and developing these capabilities and developing the ability to ask the right question in the right way to get the right answer for decades. And now you have an easier interface than having to go talk to that guy. Yeah.

You can actually talk to the machine. And I love the example. I prompt engineering.

" Okay, it'll write you an email. " By giving it the persona, it gives you a much better answer. And I feel like that's what we're doing with the agents here too, is Qlik Predict is supposed to be kind of like a mini data scientist.

And so it follows a very logical set of steps. And so when you ask it to predict something, it's not just kind of like, oh, okay, that's what it's going to be. It's following this data science best practice because that's what it has learned to do.

And that comes back to how do we make our agents better? Well, we kind of build them to be smart and to be focused. And the more focused they are, the better they're going to be at that job.

So is it fair to say then that agents will control other agents then in this format? Oh, for sure. Yep.

We've got orchestrator agents. So you have this agent that sits in the middle and tells the other agents what to do. It starts to get really complex and occasionally really expensive.

But it's pretty cool. Right. Well, I think the advantage of all of this is that AI is indeed very complex.

Yes. But AI is penetrating all markets of consumption, and a lot of people who use the technology don't have an understanding. So there has to be something being provided by companies like yourself, where those agents do the work for them such that they can utilize it in a useful form.

Yep. And at Qlik, we really focus on three things. I mentioned trust before.

So trust is the data that you're working with, data that you feel confident, because somebody behind the scenes has made you feel confident is right. And also can you see the data quality? And that means, can you see how many empty rows there are?

Is there empty data? Is there null data? Is it hasn't been updated for a year?

We don't want to use that data. We want to use the data that is high quality. The second is context, because if you talk to somebody in a business role, they have business logic.

They have a set of terminology. And this is something that we could spend this whole podcast talking about, just kind of context and semantic layers and all that stuff. But you want to make sure that your agent is working within the context of your business.

And the third is freedom. We've talked about these different agents, but technology is moving so fast, and LLMs are moving so fast. And so as an organization, as a user, you don't want to be stuck with one specific tool.

You might want to have the flexibility to move to something else down the line. And so at Qlik, those are what we see are the three most important things to set this foundation for agentic AI, context, trust, and freedom. An interesting point about AI is it learns from its own mistakes, or it should be learning from its own mistakes.

Yeah. So at Qlik, how do you provide feedback to your consumers, right? If somebody uses Qlik for the first time and there's some issues, how do they learn how to improve and so on?

That's a good question, and that's one that we're still figuring out. We're releasing new content so quickly. One of our new capabilities is Qlik Answers, which is an LLM built into the Qlik product.

So really it's on the analytics and the integration side. But within Qlik Answers, you can start to flag, hey, this is good, or hey, this is bad, or hey, this didn't make sense, or hey, this was... Usually, people don't say when things are really good, they just say when it's really bad.

Right. And behind the scenes, we're gathering that information to continue training. But there's more that we need to do, that's for sure, and everybody, to make sure that we're able to train them to have the right types of conversations.

Right. " Yep. It would be interesting to see.

That's my view on AI, is that it's a full cycle, right? It's not one way. You don't just send- No ...

information in. And that's where we always say in data science is that the best data you can collect is the data you get from consumers. Yeah.

But of course, also, I think there's a lot of concern because outside the traditional data and analytics space, basically, LLMs are that annoying kid from middle school who will answer any question confidently, whether they know anything about it or not. And I think that there's probably a lot of concern among, generally, that we don't want to be in a situation where LLMs are generating data. Because you said garbage in, garbage out.

Well, if the garbage that's going out is going back in, then things are just going to keep getting off the rails. I think that this is generally understood, but especially in the data space. I've been talking to some of the people here at the conference, andThere's some real terror and anger among these people.

We cannot have bad data polluting our good data because these are people who love data. Oh, boy. They are committed to data in a way that you just don't experience all the time with people.

So true. How do you answer their questions and their concerns and fears about AI? So I think, one, I was laughing at an example of AI hallucination that I came up with the other day.

I was looking up the cocktail, the Paper Plane. I don't know if you guys ever had a Paper Plane. No.

It's bourbon, Aperol. It's delightful. A.

Or Mia. And I was like, "I think it might be a little older- I dispute that ... " Yeah.

But to address your concern, one, I definitely, I've heard that a lot. I was at a Qlik event a couple months ago, and we were talking about the cost of agents, of LLMs, just the cost. And somebody goes, "Well, we're really trying to control people, and so we can only let them use this thing.

" And somebody else goes, "Well, we tried to do that, and then everybody in our organization just made personal accounts. " So I think the worst thing that you can do is try to lock it down from the beginning. You need to give people some flexibility.

You need to give people some space to learn. And what Qlik has done is we've built out an AI panel, and we have somebody whose job it is to say, "Here's the tool that we're using, and here are some of the things that you should not be doing. You should not be putting press releases that haven't been released yet into Claude.

Don't do that. " So I think it's a combination of don't try to control too much because people will go find a way to use it. But also encourage them to find new use cases because this is how really great new productivity techniques come into organizations.

It's somebody had an idea and shared it, and suddenly we're doing something much faster and much smarter. So talking about learning, so I totally agree about the learning piece. Now, when you talk about sharing, do you see a future where your customers amongst themselves share data?

I hope so. I hope so. What I really hope we see instead of sharing data, and I introduced myself, the solutions team.

And what the solutions team, it's a very high-level word, but it just means that the Qlik, we have a lot of capabilities across this data integration and data analytics, but we're trying to focus on business outcomes. Historically, it's kind of like whatever your problem is, we can help you with that. But instead looking at what are the specific outcomes we can use.

And so what I hope, less than sharing data, is more so sharing these solutions, sharing these business outcomes. A customer saying, "Hey, we did this thing with Qlik and ServiceNow and Claude, and it helped us do X, Y, and Z. " Their data is different, but a lot of what's going on behind the scenes is the same.

Right. Those would be then use case templates. Exactly.

If you do it well, share it. Well, let's talk about those. I think maybe that's a good way that for us to sort of bring this conversation home for people.

You are literally out there working with customers- Mm-hmm ... working with people who are making this stuff, where the literally in the case of the NHS ambulances- Yep ... where the rubber meets the road.

It's so cool. Talk to us about some of the more interesting use cases where people have been able to use data analytics plus AI to build useful, really practical solutions. Yep.

So I mean, there's such a wide range, and I'm going to forget so many of them now. I'm kind of been myopic in the last few months, and I've been focusing on the very sexy world of FinOps. Mm-hmm.

Do you guys know what FinOps is? Okay, so for anybody who doesn't know what FinOps is, it's cloud financial management. I'm sure this is not what you were running towards, but cloud financial management is cool because everybody's doing it.

Every organization, probably even if you're in the government, you've got stuff going on in the cloud. You've got costs, and they are running rampant. And so we have customers that are using, we won't even talk about what Qlik is doing, but customers who are bringing in all of that data, and now kind of putting it in one place, and then building dashboards on top of it, and putting, let's call it even anomaly detection.

We have something called Qlik Discovery, but it's essentially anomaly detection on top. And I had an internal example. We had a team that, we run our own cloud warehouse, and it costs us money on AWS.

And our team noticed that every time we reloaded, there was a huge spike in costs. And so they went to the bottom of it. It turned out some technical solution behind the scenes, but they were able to identify it because of all the data coming together and found that they could save 90% on the cost of that reload.

Which, correct me if I'm wrong, team out there, but I believe that it was about $300,000 a year, which is pretty hefty savings. So that's one example. We are doing things with the UN on the climate crisis.

Some are good, some are bad, some are no longer in part. Well, I honestly expected there to be something involving furry animals. Come on.

I'm trying to think if we've done anything with furry animals. I was walking around the show floor, and there were some stuffies, but I need to work on my furry animal examples. Especially after your introduction.

Yeah. There's got to be like some cool agriculture. Next time you bring me back, I'll come back with that one.

Right. We'll keep that. Yeah.

So, I know that FinOps is probably the second-most sexy application of data and analytics. Yeah, talk to us about something-- Like I said, we literally have NHS Ambulance Service predicting on times, rescuing people's lives. Talk to me about something else as well.

Yeah, what else you got? So a big one that we're seeing is predictive maintenance. Yeah.

We're seeing that on factory floors. If you are a car company and you are building out cars, then you're looking at predictive maintenance in two different ways. You're looking at it, one, you want to make sure that on the factory floor, all of your machines are working, and that you're able to get in there proactively and fix a spot along the work floor before a piece of machinery goes down and halts the entire supply chain there, production floor.

On the flip side, let's say you are a car rental company, and you want to know how often your vehicles need to be brought in for maintenance. And so one of the big use cases that we're seeing is predictive maintenance. If you're that rental company, you want to drive that car or that van all the way to the very end of where it's going to go, and then get it fixed.

And so we're helping organizations do that as well. That's an interesting approach. I think people have been doing that for a long time.

Yep. " So everybody's looking at the elevator, and boom, there it fails then. So I think it's definitely a use case, a very important use case where things break and help you out.

There's so many. There are so many. We talked about a few on stage this morning, UPS, Ingersoll Rand.

We've got a really great story that's going to come out this week from Engage Together about human trafficking and the work that they're doing to identify people in need and ensure what are all the factors that you need to play with. Play is probably not the right word, but all the factors that you need to leverage to detect and reduce the incidences of human trafficking, especially in the US. Yeah, I think the thing that I love about coming to Qlik Connect is, number one, because there are so many customers here and because they're so eager to tell their stories.

You talked about hoping that they would share their data, hoping that they would share their stories. It's hard to get them to stop, honestly. They love to talk about the ways that they're using data and the ways that they're helping their organizations to do some pretty incredible things.

And so that's always a lot of fun. Last year at Qlik Connect, everybody was talking about AI. This year, again, they're all talking about AI.

I think that there's a real need to have an understanding of how to make it real, how to make it practical, how to use it, and how not to use it, and those are the kind of conversations that I'm hearing here as well. Yeah, I totally agree. People sit on piles of data, and they don't really know what to do with it.

And I think with organizations like Qlik, to use your words, is they make it real, right? People can feed it with data, see results, improve, and in most cases, people come up with new use cases. The challenge is to get started, come up with one use case, and then reuse that data and improve to build many more use cases.

And then I think the other thing I think is also very important is trust. It's very difficult to trust a large language model, so to speak, if even the large language model providers themselves don't really know how they got there. And I think this is a really exciting time to work in data.

I've worked in data for 10 years, and I feel like data is kind of like this, oh, we do it, we talk about it, we know that it's important, but we don't really care. But now AI has become so critical and pivotal in everybody's lives. Everyone you talk to, regardless of whether they work in data or not, is talking about AI.

But it's been out long enough that people understand that AI hallucinates. AI can't always be trusted. It might say the paper plane came in the 2010s.

And so data suddenly has come up again as this really important thing. And so now everybody on the data side is at a place where if we can demonstrate trust, then everybody is focused on good data and really practical applications of AI. So I don't know.

I just think it's such an exciting time. There's so many functions and industries. It's everybody.

I'm a nerd. I love working in data. Well, that's a great place to wrap this one up.

Thank you so much for joining us. Thank you everyone for listening. Before we go, let's briefly hear where folks can connect with you and continue this conversation.

Yes, absolutely. com, and Qlik is spelled Q-L-I-K. Might not be intuitive, but that's how it's spelled.

It's really cool. com. We've got so much out there, and if you really want to catch up on Qlik Connect, I'm sure there will be a recording, so you can kind of see everything that we've announced live, including so many exciting customer stories.

And also, please feel free to reach out to me on LinkedIn. It's Sidney, S-I-D-N-E-Y. Don't get that wrong like Steven did.

Yeah. Hey. com.

And of course, you'll find Frederik and I as well at our forthcoming AI Field Day event- Right ... in May, so check out the Tech Field Day website to learn more about that. " If you enjoyed this discussion, please do subscribe.

You'll find us on YouTube as well as in your favorite podcast application. This podcast was brought to you by the analysts and experts from the Futurum Group, and of course, we are here at Qlik Connect on a sponsored appearance here at Qlik Connect as well. ai, the "Utilizing AI" YouTube channel, or the Techstrong TV app.

Thanks for listening, and we will catch you next week.