Informatica World 2024 | Graeme Thompson

Clips
News
More from Informatica World 2024

Graeme Thompson

Senior Vice President and CIO

Informatica

Navigating the AI data maze: Insights from Informatica’s CIO

When thinking about artificial intelligence, the old phrase, “You are what you eat,” comes to mind, and what AI eats is data.Understanding where the data comes from and how it will impact the model is critical, according to Graeme Thompson (pictured, right), chief information officer of Informatica Inc.Thompson (right), CIO at Informatica, discusses how the data storage landscape has changed in the face of AI, with theCUBE hosts, Rebecca Knight (middle) and Rob Strechay (left)

play_circle_outline Importance of data quality and preparation for successful AI implementation

play_circle_outline Challenges of privacy and security in the AI era

play_circle_outline Need for cataloging and lineage to ensure data trustworthiness

play_circle_outline Considerations for enabling innovation while managing risk

Info

Graeme Thompson | Informatica

Graeme Thompson

Senior Vice President and CIO Informatica

Rebecca Knight

>> Hello everyone, and welcome to theCUBE's live coverage of Informatica World. We are kicking off three days of wall-to-wall coverage. I'm your host alongside my co-host and analyst, Rob Strechay. Rob, back in the chair right where you belong, right next to me at another amazing show.

>> Yeah, and we're going to be talking a lot about Informatica and how really AI is having its moment, but that the data and how you actually bring the data to AI is really important and how you bring AI to the data as well.

Rebecca Knight

>> And who better to talk about that than Graeme Thompson? He is the SVP and chief information officer at Informatica. Thank you so much for joining us, Graeme.

Graeme Thompson

>> Great to be here.

Rebecca Knight

>> So, you are, as I said, CIO of Informatica. And you're a tech veteran, have been there for about eight years before that at Oracle. I'd love you to just maybe shift into a reflective mode and talk a little bit about what we can learn from past technology shifts that can help prepare data leaders as we enter into this new era of gen AI and AI.

Graeme Thompson

>> Yeah, I mean, I think it's always good to reflect on what's happened before, because often, if we do the same thing, you're going to get the same outcome. So, I think most recently it's really informative, I think, to look at our analytics journeys. So, 11 years ago or so when Tableau came on the scene, everyone went nuts on buying these visual analytics tools and putting it on top of what at the time was horrible data, and expecting a really great outcome. So, no shock, we were all disappointed. If we hadn't made the investment to make sure that the data underlying the analytics solution was of high-quality, was trusted, was defined in such a way that it could be used cross-functionally, our analytics expectations were not going to be met. The same thing is absolutely true with AI, except it's going to be even worse because you're creating content based on this information, you're probably going to be at some point sending that content out to your customers or using it internally. The stakes are just so much higher that if we make the same mistake, it is going going to be worse.

>> Yeah, I mean I think that when we look at it, and again, we follow Informatica, so we get to know you guys pretty well, but from a perspective of when you start to look at how Informatica is really dealing with the data, the metadata, helping understand here's lineage and things of that nature, you're on the inside, you're owning the inside, looking out for this as well. How does that really change your thoughts when you're seeing people, "Hey, we want to use an LLM. We want to use these things," how does that really come together? And what are some of the challenges that you see that others are... you may be ahead of them by thinking about these that they should understand?

Graeme Thompson

>> Yeah, in some cases, I think we are. We have the benefit of talking to thousands of customers who are going through this journey. And it's so exciting because it's so new. Many of them are here this week, so I'm sure that it wouldn't be difficult to find people that would agree with you that AI is only going to be successful if you manage the data. That's why these people are here at our conference this week. But think about all the other people that aren't here and all the companies that haven't even started to think of this, they're the ones that I would worry about if I was a customer or a stakeholder in those companies if they're not thinking about how to get their data ready.

Rebecca Knight

>> And so, in thinking about those people who are not here, and I know you're not in marketing, you're the CIO, but how are you getting your word out and making sure that they are, in fact, worried about the right things?

Graeme Thompson

>> Yeah, so I think your marketing team has done a fantastic job simplifying the message. It's very common for technology companies to overdo it and have the complexity in building the product show up in some of the marketing. But the way that they've done it for the last year or so where data and AI come to life, I think that is a really powerful message, and it makes people at least ask, "Well, what does that mean for me? And what should I be doing that I'm not doing that's going to be in the way of our company's AI aspirations?"

>> Yeah, I mean I think we see it as well as that it's been challenging for those organizations and a lot of them have some trepidation. And as the CIO, you're chartered to protect Informatica and protect the information and all of that. How do you see privacy and security in an AI era really challenging organizations?

Graeme Thompson

>> Yeah, so I think it's going to be the deal-breaker or the deal-maker for many companies. So, if you think about a simple use case where you just want to put an industry available LLM on top of some data and create some content or interrogate the content in an interesting way, your legal team's not going to be comfortable letting you do that unless you can demonstrate that you know for sure where the data came from, who has access to the data, if it's customer data, has the customer provided consent? And that is going to take a lot of hard work to prove to yourself that you can do that, and then demonstrate it to your legal organization that what you want to do is safe, responsible, and isn't going to get the company into trouble. So, for the companies that have done the hard work to understand the lineage of the data through a catalog, to understand how the data's accessed using a data access management solution. And then, to make sure that the data is of high-quality. Those companies are going to go way faster and ultimately end up ahead of the companies that can't do that.

Rebecca Knight

>> So, where are we now? I mean obviously with the birth of ChatGPT in November of 2022, it showed so many people the potential of LLMs. And really said, "Okay, this is what we can do with this." Can you talk a little bit about that moment in time and how much it captured the imagination of normal people as well as high-flying technology executives and really where we are now too?

Graeme Thompson

>> Yeah. And it was so interesting that what was a new technology breakthrough for many people... I mean obviously folks have been working on this for decades, but in November it became something that was a common discussion around the dinner table. And people that have no interest in tech were now experimenting with ChatGPT. It was fascinating and it is so exciting. But I think we have to caution ourselves and think that, okay, ChatGPT is built on top of the internet. You could argue that the internet is a reasonable proxy for the general knowledge of the world. It's got pop culture, it's got literature, it's got music, it's got news, fake news, real news. So, there's no surprise that it can write a birthday poem or a wedding toast or a cold call email to a CIO that they can ignore. It's no shock that it can do that. But if you reflect on how that is going to translate to an enterprise use case, my data's not on the internet. Google hasn't been organizing it for 20 years. People haven't been tagging it, so that you can find it. People aren't organizing it in such a way that it can be optimized through search and found. Most company's enterprise data is a mess. So, it is as far from ChatGPT on the internet as you can possibly imagine. And I think the thing that's really different this time round is when sometimes in the past when new technologies come about, even the internet, the internet was useful without a browser. The content was there, there wasn't much of it, but you couldn't find it or do anything with it. IoT was useless until 5G and real cellular connectivity came along. The cloud took a while to get traction because people were worried about security. But as a consumer of those technologies, I could sit back and wait for the vendors to solve those problems, and all I had to do was to be ready when the tech was ready. This is not that. We have to do the hard work to build the trusted data platform, so that when the tech is ready, we are an enabler and not an inhibitor. CIOs and data people often end up having to choose between being an enabler or being in the way, and this is one of those moments.

>> Right. Yeah. I think that to me is one of the big things because you see what happened with the Canadian airline that we won't call out. But when they go in there and it gives the wrong answer and now they're forced to give a discount or do something of that nature, hallucinations and other problems lead to a lack of trust and things of that nature. How do you talk to that? Because again, it talks to a bit about the sources of data and where it comes from.

Graeme Thompson

>> Yeah, even now, you can ask an LLM to confine what it looks to be something specific. So, you can even feed it a document and say, "I'm going to ask you questions about this document." So, that's fairly simple and doable, and it removes a lot of the risk of hallucination. But imagine five years from now when you can have a piece of content created and just like a well-annotated college paper, you can highlight a passage and it'll tell you where the data came from, so that you knew it was from real news, not fake news. It was from a medical journal, not just someone with a blog. That's going to be possible through cataloging and lineage and properly organizing the data, and that's the part that I think is so exciting.

>> Yeah, I agree. Because I think in that instance with the Canadian airline, one of the things that was funny about it is what if you do want offers to be in there, but you need to pull them back and that lineage aspect has to be there? For sure. Definitely.

Rebecca Knight

>> I wanted to ask about something you'd said about making the decision to be either an enabler or be in the way. And I'm thinking about the personality type of a CIO, because you obviously have to be the ones driving innovation, and you've got to be creative and forward-thinking, but I would also think you should have a certain measuredness about you and be thinking about the risks because you are worried about lineage and security and privacy and all those things. So, what are you hearing from your peers about this tightrope that you need to walk about making sure that you are enabling and not getting in the way, but maybe sometimes being in the way at the right time to say, "Hold up. What about the lineage?"

Graeme Thompson

>> So, honestly, that's the part that's fun. Sometimes you have to be the adult in the room and not get carried away with the promise of a new thing that just came out. But if you don't innovate, you're going to leave your company behind, and that sounds very trite. But think about just the internet on its own. Nobody's saying in their company, "We're going to increase productivity by using the internet," because it's available to everyone equally. The Copilots, the things that are inside your applications, they're going to be available to everybody that's not going to be differentiating. So, if we are going to be an enabler, we have to really lean in hard and figure out how am I going to make this available for the stakeholders of my company? How am I going to make it safe? How am I going to make sure that the data's used responsibly? How am I going to make sure I can trust the outcome? I think it's a really exciting opportunity for CIOs and other data leaders to really lean in and provide that enablement capability. There'll be a lot that focus on the risk avoidance part, and those folks are going to be in the way, just like they were in the way when the transition to cloud came up or other big shifts that we've seen people take advantage of. And this is when great companies become better and companies that aren't great sometimes get hurt, and that's the part that makes it so exciting.

>> Yeah, I think that to me brings up the whole risk of doing it or not doing it is pretty great. And either way... And in fact, we have a partner, ETR, and they collected data. And when you started to look through the data they collected, you got to the Fortune 100. Of the Fortune 100, they had 41 of them that responded. Of that, 49% said that data privacy and security was their biggest, I guess you could say, inhibitor to getting to production and legal and compliance, or legal and compliance that hasn't yet happened. I mean, you have the AI Act and you have some other things. What do you see as advice to those organizations that are trying to overcome that to get to production?

Graeme Thompson

>> Yeah. So, I think centralizing all the good ideas... And we've created a COE, so a center of excellence that has community members from all around the company. So, that's number one. So, that you've got a single place where the ideas can be vetted. And it helps avoid duplication. It helps figure out, "Hey, your three use cases are kind of the same. Let's just build it once." It helps with all of that. But also, I would advise people to have their legal and privacy teams directly represented on that COE, so that they can provide good guidance on what's likely to be allowed and not allowed, but also so they have the benefit of seeing the use cases and the business challenges firsthand. Because if your job is to avoid risk, and every idea you see is just more risk, your go-to answer's going to be no. But if you can see both sides and you can see the risk and the benefit, then most people can be reasonable and figure out, "Okay, what is a safe way to take advantage of this potential benefit?" As opposed to, "What's the safest thing I can do to avoid all the risk?" So, I think involving all those stakeholders, your finance team also, because it can get expensive, so you want the finance team to be involved too. So, it may be tempting to exclude some of the oversight functions, but it's not going to ultimately help you go faster. So, I would err on the side of including the traditional oversight functions.

Rebecca Knight

>> Last question. We're just getting underway here at Informatica World. What are you most looking forward to this week?

Graeme Thompson

>> It's always the same. It's always hearing from our customers. We have some of the most important companies on the planet using our technology to do great things, but I am so excited to hear what they are doing now, what they're thinking about doing next, and where we can help with AI and the data underneath. I'm sure it's going to be a great few days.

Rebecca Knight

>> A great note to end on. Graeme, thank you so much for coming on theCUBE.

Graeme Thompson

>> Thank you. It's great to be here.

Rebecca Knight

>> I'm Rebecca Knight for Rob Strechay. Stay tuned for more of theCUBE's live coverage of Informatica World. You are watching theCUBE, the leader in enterprise tech and analysis.

Graeme Thompson | Informatica

Rebecca Knight

>> And who better to talk about that than Graeme Thompson? He is the SVP and chief information officer at Informatica. Thank you so much for joining us, Graeme.

Graeme Thompson

>> Great to be here.

Rebecca Knight

Graeme Thompson

Rebecca Knight

Graeme Thompson

Rebecca Knight

Graeme Thompson

Rebecca Knight

Graeme Thompson

Rebecca Knight

>> Last question. We're just getting underway here at Informatica World. What are you most looking forward to this week?

Graeme Thompson

Rebecca Knight

>> A great note to end on. Graeme, thank you so much for coming on theCUBE.

Graeme Thompson

>> Thank you. It's great to be here.

Rebecca Knight

>> I'm Rebecca Knight for Rob Strechay. Stay tuned for more of theCUBE's live coverage of Informatica World. You are watching theCUBE, the leader in enterprise tech and analysis.