The Data Value Chain: Key to Successful AI

Michael Krigsman: Successful AI requires the right data and an understanding of the AI value chain and the data value chain. Here to explain these topics is Mike Capone, the CEO of Qlik. 

Mike Capone: People know us as a leading provider of data, integration, and analytics tools, and that's certainly true. But we like to think of ourselves as having a much broader mission to help organizations transform data, the data they have access to, into actionable insights that can help them grow their business. 

We honestly believe that data, analytics, and AI should underpin every single decision that you make, every single operational decision, and every single strategic decision. Our capabilities are what bring that to life for our customers. 

Michael Krigsman: Mike, you were a chief information officer in your past life. How does that affect your work at Qlik? 

Mike Capone: Two things. I really understood what it's like to be on that side of the equation, being somebody trying to deliver value for my company through technology, and dealing with providers, and partners who are helping me do that. That experience working at ADP, a really large data processing company – in fact, that's the name – taught me how data can be such a strategic asset to a company and how harnessing that information can provide strategic advantage, as it did back when I was at ADP. Of course, it also taught me how hard it was to get the results. 

Michael Krigsman: Mike, when you talk about strategic decision-making through data, that is a crucial part of the broader AI value chain. Tell us what that is and what are the components. 

Mike Capone: AI value chain is just a name for the end-to-end process to capture, to curate, to catalog, prove the veracity of data, and get it ready for AI and analytics. Then analyze it and apply modern AI capabilities. It's that sequence of events that comprises the AI value chain.

Data is the crucial part of the AI value chain. Everything starts there. 

I think the problem with all the hype going on right now is everybody wants to just jump to the end. It's like, "Okay, I can plug in ChatGPT and all of my problems are solved." Right? 

We just know that it's not true. It all starts with data and harnessing that data and getting it into an analytics and AI-ready format, which is really hard work. 

When you think about the data component of it and the value chain around data, for example, it starts with data capture, the ability to harness data from disparate systems everywhere, whether you own them or not. But certainly, ERP systems, cloud systems, mainframes, wherever that data may be, you have to harness that. We call it capture data.

More and more, it has to be real-time. Yesterday's data, last hour's data, is just not good enough anymore. You need up-to-date information.

Then you need to work with that data to improve the quality of it, to make sure you feel really good about the veracity, the quality of it. You have to capture the lineage of the data. You also want to remember where that data came from as you move forward in the AI value chain.

Then, of course, you move on to the ability to catalog it, curate it. Then finally, analyze it, get insights, and apply modern AI capabilities to it. Then last but not least, as we move further up the value chain, once you've done that is then do something with it, right? 

There are a lot of C-level executives who tell me, "Even when we get the answer, we don't know what to do with it." So, how do you take all the insights you get and then automate that and force a decision inside your company, force an action? That comprises the entire value chain.

Michael Krigsman: Why is the AI value chain so important for business and technology leaders who really understand?

Mike Capone: It is the most fundamental question, as we move into the next era, which is really going to be AI-driven, the ability to understand what drives value in AI. Values derive from the underlying data, so making sure your AI value chain is robust and delivers the quality and veracity of the data at the velocity in which you need it is fundamental to everything that comes after that in the world of AI.

Michael Krigsman: Mike, this concept of the AI value chain is so fundamentally important. To what extent do technology leaders and business leaders really have a grasp of that value chain inside their own organizations?

Mike Capone: If you asked me that a year ago, I probably would have given you a very different answer because everybody was caught up in the hype of AI and just doing something. Fast forward to now, and people are being much more thoughtful. 

The good news is, most of the C-level executives I'm talking to (everywhere in every industry and every sector) are talking about the AI value chain. They're understanding more and more that we need to do the work to build this robust foundation, this value chain, to be ready for our organizations to truly embrace and gain the benefits of AI.

Michael Krigsman: There is an increasing organizational maturity in the market around understanding this AI value chain.

Mike Capone: There is absolutely a maturity happening right now. In some industries, it's happening faster than others. But I think now that we're a year into this cycle, people are being much more thoughtful and saying, "Look. I need to take a step back, and I really need to focus on the foundation," and that foundation is the AI value chain that we've been talking about.

Michael Krigsman: Mike you emphasize the importance of data. There is also a data value chain. What is that and how does that fit in with AI and the AI value chain?

Mike Capone: They're inextricably linked but they're not exactly the same thing. The data value chain, which is the ability to harness data, understand where all of your data is in real time, that underpins a lot. It underpins your AI strategy, for sure.

But it also underpins your application modernization strategy. It often very much underpins your day-to-day transaction processing. So, that's really, really important and foundational, but you need to do both. You do need to focus on the data and then focus on your AI strategy.

Michael Krigsman: Let me ask the proverbial question. Which comes first, the AI value chain or the data value chain?

Mike Capone: There's no question the data value chain comes first. Look, neither one of these are new. I was working on AI when I was in college some ungodly amount of years ago. This is not new.

What's new, Michael, is the capabilities we have, the unlimited computing power we have now through cloud computing, through the hyperscalers. What's new is the advances we made in algorithms and technology.

Now we're able to do more things but the foundation has always been data. Now we're just layering AI on top. 

Michael Krigsman: When it comes to the data value chain, can you give us a high level view of the strategy to really understand the value chain and implement that data value chain in our own organizations?

Mike Capone: Everyone wants to start with the data itself, and that's absolutely the wrong thing to do. You really need to understand the fundamentals of your business and what are the drivers of your business.

What strategic and operational decisions do you really need to make and, more importantly, you need to improve? Where is the leverage in your business and what decisions can you make better informed by data and artificial intelligence? That is an exercise unto itself. 

Once you do that, then you can work your way back and understand, okay, what data needs to unpin the strategic decision making, this operational decision making? Then how do I harness that data, get it AI-ready, and get it to a state where it can apply these modern capabilities and get accurate results? 

The reality is, once you decide what those decisions are, making sure that you have curated, quality data underneath it is going to be critical because otherwise, you'll end up making bad decisions based on bad foundations.

Michael Krigsman: Can you elaborate on that? You said curated and quality data. 

Mike Capone: We've all seen what's going on in the market right now. They have all different terms for it now (right): hallucinations, model drift. These are bad outcomes you get because you didn't take the time upfront to understand the underlying data in your AI landscape.

Really, what we tell customers to focus on is making sure that that foundation is 100% correct and reliable. 

Michael Krigsman: You talk with so many folks across different industries. What are the challenges, the patterns that you see that present obstacles to really having a clear understanding of the right kind of data and ensuring that that data has a sufficient level of quality?

Mike Capone: Everybody knows that the data that goes into the operational systems isn't always of the highest quality, right? That needs to be addressed, and that's a pretty heavy lift very often.

The second thing, honestly, is cost. People are starting to get the first anniversary of bills from their cloud data lake providers. Suddenly, they're realizing, "How do I build a data landscape that isn't going to bankrupt my company?" 

That's really a data fabric discussion where you talk about, "Okay, what data do I need to move? What data can I leave in place or take parts of? What is the frequency of movement, et cetera?" Really, getting our arms around that is going to be super critical.

Then the last piece obviously is, "What do I do with this information? And what are my risks? What if I apply this technology and it produces some bad outcome that puts my company in jeopardy? How do I manage and mitigate that risk? How do I weigh it, measure it?"

Those are all the conversations that are going on right now. And if you're a CIO, it's a hard place to be, right? But never have you been more important to your organization than right now in terms of making sure we get this right. 

Michael Krigsman: Mike, you've given a broad overview of the kind of challenges that organizations face with looking at their data, curating the data, and going through the quality process. Where do companies get stuck? Where are the pitfalls?

Mike Capone: Not having an underpinning strategy, right? Every business unit, every person in the company kind of has their own data and analytics strategy. The data is not governing, harnessed. There has to be centralized governance. 

That doesn't mean that IT is command and control and holding onto everything. On the contrary. You have to democratize the data. But you have to build a governance structure that can ensure the veracity, the quality of the data across the enterprise. 

The pitfall is if you don't convince people, if you don't help them understand why that is necessary, you're really going to struggle to win the hearts and minds of people. 

Michael Krigsman: Is there a linkage between the kinds of governance structures that organizations put in place and their underlying data infrastructure? How do these intersect?

Mike Capone: The technology that you deploy needs to support your data governance strategy, the capabilities inside of the technology you need to deploy, the change data capture, the data quality mechanisms that you're using, the tools you're using to curate and catalog your data and transform your data. They have to have governance capabilities built into them. It's super important, and it's something that we build into all of our products all along the chain. I think that's absolutely critical.

If you're just moving data around without any quality checks or any governance, in the long run, you get into a lot of trouble because you're basically going to wind up with a free-for-all all of data that nobody is controlling the quality of. 

Michael Krigsman: Mike, you're describing a set of technology capabilities and a set of what we might call organizational capabilities. Both need to come together in order to develop an effective data strategy. What have you seen at your customers when it comes to this issue?

Mike Capone: There's a whole group of constituents that are terrified of what AI is going to do to their function inside of a company. Worry is never good when you're trying to implement change.

That's why it's super critical to have the cultural aspects and the cultural dimensions of a data strategy along with the technology. In many cases, the technology is easier because you can implement the technology.

What's not easy is getting people to buy into, "Okay, we're going to harness this data. There are a series of disciplines and processes we're going to need to implement in governance that everybody has to come along with so we can ensure that our foundation of data is correct. Then when we deploy AI to help you be more effective in your jobs, we know that the data can be trusted and the output from these kinds of modern tools, these modern AI tools, is going to be accurate and correct."

Michael Krigsman: You've mentioned this term "trust" several times. Why is that so foundationally important?

Mike Capone: The reality is AI has huge potential. It's also fraught with risks. But if we deploy AI correctly and we get people to trust AI by first getting data that people could trust (that's validated and ready), and then showing them when you do that heavy lifting around the data and then you apply the tools, the outcomes are much better. Then you absolutely can earn people's trust.

Then there's a second component of this, which is you also have to get people to be data literate. We spend a lot of time at Qlik talking about this concept of data literacy, and this is the ability to get people comfortable with working with data because working with data today and tomorrow is going to be as important to the new economy as reading and writing was when we came out of the Industrial Revolution.

Reading, understanding, questioning data and outputs of data, that skill has to be taught as well. Only then can you start to build a culture where people can trust data because they know how to ask questions.

Michael Krigsman: You raise a very interesting point. We typically think of data as being this highly technology driven topic with infrastructure and data lakes and so forth. But you've just made a direct linkage to overall corporate brand reputation and this highly technical data subject.

Mike Capone: Bad practices in this area can really lead to bad outcomes. The technology is advancing so fast that you cannot possibly legislate this stuff. So, it's up to people in the private sector (both providers like Qlik but also executives inside of companies deploying AI) to forge the trail. 

We have built a council inside of Qlik of outside industry experts to advise us on responsible AI, how you can deploy it, and not run into some of the ethical, moral, and reputational things that you just described, Michael. It's really, really important. 

In fact, I was out at the Whitehouse Advisors a couple of weeks ago, and they were asking the same thing. "How should we engage in the private sector to govern AI, to talk about AI, to be responsible about AI?" It's something we're all wrestling with. But what I do know for sure is those of us in positions to have influence have a pretty awesome responsibility right now.

Michael Krigsman: Mike, you mentioned the Qlik AI Council. Tell us about that. That's really interesting. 

Mike Capone: We wanted to make sure that we were not being overly insular about our deployment of AI, both internally but, more importantly, with our customers. So, we have an executive advisory board made up of really important customers, which helps us quite a bit. But we wanted to go further.

We solicited a series of experts and built what we call our AI council. Those are folks from the industry, from academia, from various parts of the world. There are four really talented individuals who advise Qlik on our AI strategy and how we make sure that we're building responsible AI, moral AI, as we move forward on our journey.

Michael Krigsman: Mike, you're describing AI at so many different levels: the AI value chain, the data value chain, the technology infrastructure, responsible and ethical use of AI. How should CIOs manage this? 

Mike Capone: They really need to invest the time and take a step back. That was the thing that had me unnerved a year ago was everybody was scrambling. 

By the way, vendors weren't helping (right) because vendors were making claims about what they could do, what they couldn't do. There was a lot of press releases about this capability, that capability. So, it created this frenzy. Then, of course, boards were also not helping (right) because they were putting pressure. 

The good news, everybody has taken a breath. I think that phase of overhype is over. And now people are saying, "All right. Let me be thoughtful about this." And that's the key. You've got to be thoughtful about it.

We're all running 100 miles an hour. Take a step back. Take time to understand and learn. Continuous learning is a huge part of being a successful executive. Not enough of us do it. 

My technique is to solicit, like, our AI council or solicit expert opinions. Try to just talk to as many people as I can and, sure, read as much as I can, et cetera. 

What I found is peer networks are great because everybody is going through the same thing. Everybody is apprehensive about how to get their arms around this. And so, the ability to take the time to do that is going to be really critical for anybody's success, but particularly CIOs.

Michael Krigsman: Mike, any thoughts on how CIOs should balance the technology dimensions of all of this versus the organizational, the cultural aspects?

Mike Capone: In the beginning, I think you have to overweight the cultural aspects of it, too. Any technology deployment is going to fail if the organization is not ready for it. 

You can't just jam technology down the throats of a business. I think having that cohesive strategy that, like I said, the data strategy and the AI strategy is not a CIO strategy; it's a company strategy. 

It's going to impact every decision, every process that you make inside of a company. It should if you're doing it right, and it's going to be a big part of the future. 

Getting that executive buy-in and then the layers of the organization onboard almost has to be overweighed in the beginning. Then once you get that foundation, you can run as fast as you can to make sure the technology is in place.

In the interim, you need to be building proof of concepts. You need to be experimenting and doing all the things you would expect a good CIO to be doing. What you don't want to do is rush ahead with the technology and ignore what is going to be a massive change management and cultural impact.

Michael Krigsman: On the surface, AI deployment looks like a technology problem fundamentally. But what you're saying is that, in reality, underneath, it's actually a business challenge that is expressed through various technologies and implementation processes and so forth.

Mike Capone: That's right. There are really two dimensions of the business process. The first is what's possible. 

What parts of our business can really be empowered, enabled by AI beyond the normal sort of decision-making? But what can we do to make our company more effective? Then, of course, how do you build back from that?

But the second is the moral aspect to it, too. Just because we can doesn't mean we should. 

How do we use customer data? How do we use technologies like facial recognition? Companies have gotten in front of some of these things with great technology, but there was really no governance.

Michael Krigsman: You're advocating a highly strategic approach that relies on deep understanding of the business goals, gaining consensus across the organization, and a clear understanding of the data and how that data will ultimately be used and what will the outcomes be. 

Mike Capone: I'm not advocating a "Mother, may I" approach where you say, like, "Hey, we want to do AI. Everybody good? Are we up for it?"

The CEO has to be very direct, "We're going on this journey. What I need is everybody to participate in the strategy and the execution to make sure that we do things correctly both for us, for our customers, as well as for the organization itself." That's the way to do this. 

It requires a top-stop condition to go down the journey, and everybody can't have veto power, but everybody can at least be heard in this process. I think that's the kind of governance framework you need to build.

Michael Krigsman: The governance framework is based on a broad strategy. But then you drill into the data strategy, which really must form the underpinning of any successful AI effort.

Mike Capone: That's absolutely true. AI is just a given. It's a given in everything we do. But it's like, "Okay, now where do we get the most leverage, and are we sure that the road we're going down is one that we think A) will make us successful but B) will make us feel morally okay in terms of how we're going about it."

We have 40,000 customers. We have a much larger obligation, which is, how do we help our customers get to those decision points for themselves and provide the governance framework, tools, and the capabilities to make them as successful as they can be with the parameters that they want to apply to this?

Michael Krigsman: Mike, AI is changing so rapidly. Where do you see things headed over the course of this next year?

Mike Capone: We're going to continue to make sure. We're going to continue to get a little bit more grounded. We're going to have to slow down before we go faster again.

The good news is everything that you and I are talking about, Michael, right now is necessary no matter what. You don't need to guess what the next cool AI tool is going to be or how LLMs are going to evolve to know that you've got to get your data house in order, that you have to build the data value chain that will ultimately underpin all of this. 

It's not just structured data anymore. It's unstructured data. It's all of the capabilities that need to be built.

I think the good news is you can make lots of progress and you can also keep a close eye on what's happening in the market and watch as things unfold because the capabilities are going to continue to morph and grow. But you don't have to jump on every shiny object that comes out these days. If you wait long enough, it'll change.

Michael Krigsman: Tell us how Qlik supports the data strategies you've been describing.

Mike Capone: Qlik supports all of our customers. We're really the only ones who have an end-to-end solution for both the data value chain and the AI value chain (when you think about it), everything from change data capture at high velocity and real-time, all the way through data quality, data transformation, curation, and catalog. 

We've always been known for our powerful analytics platform (right), the Qlik Sense, which has been around for 30 years. Then we have modern AI capabilities built into our platform. 

We also interact with every possible computing solution out there, every cloud provider, et cetera. We're agnostic when it comes to who we work with, and I think that's what companies want. 

Then the last thing we bring is we bring automation capabilities, so it's great to get the answer. It's even better to actually do something with it. Once you know what to do, go do it, and you can do that through automation, which is something we provide to our customers. 

That end-to-end vision, we started building five years ago. I've been getting ready for this moment for five years when we saw this coming. 

"It's about the data, stupid," was what we were saying back then. That's what we try to build coming into this year. We really feel like we got it.

Michael Krigsman: Successful AI is all about the data. 

Mike Capone: I think that's 100% right. Without the data, nothing else matters.

Michael Krigsman: With that, Mike Capone, CEO of Qlik, thank you so much for taking time to speak with us.

Mike Capone: Oh, Michael, it's a great pleasure. It's great to be talking to you again.