Data Lineage for AI Governance | Philip Dutton, Solidatus

David Sweenor 0:00 Hello and welcome to the Gartner data analytics conference in Orlando, Florida. I’m here with Philip Dutton. He is the CEO and founder of solidatus. Philip, how’s the show going?

Philip Dutton 0:10 David, thank you very much for talking with us. Show’s going fantastic. Actually, got off to a really, really busy start, which is fantastic. And it looks like they’re hitting record numbers. So also great.

David Sweenor 0:21 Yeah, I saw your booth earlier. It was hopping. For those who don’t know or aren’t familiar with solid data, so I’ll give you the 30/62 overview.

Philip Dutton 0:28 Certainly. So solid data is we are the data lineage company, so we our superpower is data lineage, but really it’s data lineage for a purpose, and that’s really generally focused on data governance for an organization. Now, we grew up in financial services, and so our background really is in a highly regulated industry, and so the risk versus reward is kind of where we kind of help organizations kind of determine how to use their data effectively.

David Sweenor 0:54 You mentioned data governance and purpose. So this was lineage has historically been used for sort of compliance and regulations. The theme of this conference, one of them is AI. Turns out to be pretty useful for AI, doesn’t it?

Philip Dutton 1:09 Yeah, it really does. And I think you know for us, originally, the focus was, was really on helping organizations understand themselves, so they could change better. And really that change is the natural evolution into AI. So this is a whole new paradigm that we’re now focused on, but the fundamentals, the concepts that we’re dealing with, are exactly the same. And really that’s where solidatus is kind of so powerful in that we don’t need to do any modifications to the software in order for it to take on these new concepts and to help govern them. And of course, we’re going to also launching our new AI capabilities in the tool, which is, again, another kind of use of AI, really effective use of AI in a product.

David Sweenor 1:50 So that would be the AI lineage assistant, right? It is, indeed, yes. And so what, how would people use the AI lineage assistant, and why would they use that versus whatever they’re doing today?

Philip Dutton 2:00 Yeah, that’s that’s a great question. David, and I think for me, the first thing, of course, is we’ve now overlaid a natural language interface, so allowing customers from the C suite down to the very, very technical to have natural conversations, so no longer needing really any solid as expertise at all to interact with their data. But it goes further than that, because now that we’ve allowed kind of more people to access the information, we want them to be able to do that more quickly. So do work in the tool more quickly. So if you’re a data governance professional, if you’re someone who works in compliance or risk, you know you can now access, interact with your data, and we’re seeing between 10 and 100x kind of acceleration. Wow. And the last piece really is then, well, that’s great that we can interact with kind of and do what we already could do, but faster. So we need less people spending less time. But actually, what it’s also done is increase the number of capabilities that are in the tool, so things which were previously outside of, let’s say, kind of the lineage, metadata, purview that we existed, have now come into focus. And so that’s really exciting for me. Because the great thing about solidatus is you put it in the hands of smart people, and they come up with really fantastic new use cases. And so these new capabilities I see as a huge unlock for a lot of new activity.

David Sweenor 3:17 Okay, you know, one of the things I know people are talking about here, and especially a lot of the vendors on the floor, there’s actually two dimensions to this. One is trust, and meaning trust, can I trust the AI that’s in the software? The second dimension is really trust. How can lineage be used to help improve the trust of AI? So let’s start with the first one. You know, is it safe? Because people are worried about that, yeah.

Philip Dutton 3:43 So, I mean, I think you know, for for the the AI within solid Avis, is product 100% safe, you know, we allow you to bring your own LLM, and so that means that you’re plugging into your control environment. It’s not data that is coming to us and then going through other third parties. This is all well contained. And then from there, you know, the the real challenge with the llms is hallucinations, right? David, when they start to hallucinate, you don’t know whether you can trust the response or not. And so with solidatises, we’ve actually built in hallucination protection. So the idea is that if the LLM returns a response that isn’t backed off against metadata that exists within our platform, that we catch it and we say no, that obviously can’t be correct, go back and have another go at it, and to try and have it kind of be forced into ensuring that the data that it responds is based off and is grounded in real data, which is a big difference from kind of just allowing it to be, kind of generating anything, and having that go straight to the user.

David Sweenor 4:45 Sure, it’s like an extra layer of safety. And still, financial services companies, they’re pretty conservative, so kind of always want human in the loop, or it’s their choice, I suppose, depending on the use case.

Philip Dutton 4:56 Yeah, no, exactly. Humanoid is critical, and that’s one of, again, the. Features of solid edX has always leaned into heavily, is really allowing people to have the authority and the responsibility for what goes on in the product. So you talked about kind of the other piece of the trust, sure, and I think for me, if we can’t see it, if we can’t understand it, how do we trust it? Humans are fairly kind of simple creatures at the end of the day, right? And if you can see it, if you can see that it matches to, you know, your reality of what you have in your organization, then you’re like, Okay, that looks correct. That’s great. I’m like, I’m happy to go forward with where you can’t see it, then the Well, can I trust this? Or I can’t trust it? You know? Then the questions start to ask. And I think actually we’re doing a presentation tomorrow at 335 on Data Trust with London stock exchange. So we are one of the foundations for their data trust program, because lineage really enables you to have that visual representation of what’s going on, what data is interacting, etc,

David Sweenor 6:01 you know, what’s interesting? When I was looking at the demo, this is very, very visual tool. And when I see lineage from other vendors, I don’t see these beautiful maps, like, why do you why do we need this? Isn’t, I mean, I would, is it automated? You know?

Philip Dutton 6:18 Yeah, no, I think that’s, you know, for me, one of the foundational things is people are very, you know, Vision driven, like, if you give someone a whole lot of text, it takes them a while to process, you know, the old adage of, you know, a picture tells 1000 words. That’s exactly what solidatus was designed to really pull through, is we want people to interact with their data quickly to make decisions quickly. We don’t want them to have to spend hours weeks poring over metadata to understand a simple concept. And so the visualization is really critical, and because of that, you get much more of an affinity, kind of, across the organization, for people to interact with it, which is helpful, because the needle jumps out of the haystack in complex organizations where it doesn’t in others. I love

David Sweenor 7:05 that, because I’m a visual we’re all visual creatures, I suppose. But I think maybe another dimension of it as well as AI can’t automate everything, so you need a way to manually do this, and maybe visually is the best way to do that.

Philip Dutton 7:19 Yeah, no, I think that’s the thing is everyone kind of really wants to jump to 100% automation. And kind of, I go to the kind of the the adage of, like, the first time Henry Ford kind of built a production car assembly line. That wasn’t the first time he built a car, right? So he already knew what kind of good looked like. It it was the automating of what good looked like in terms of the building process, and so jumping to 100% normally ends up kind of lots and lots of rework. So what we want to do is is incrementally get there, but get there in a really kind of well thought out and structured way. Because when you do it that way, you get an outcome that you expect if you just go and try and automate everything, what you naturally find is, well, I didn’t think of that. We can’t automate this piece. So what do I do? I’m stuck now, right? So solidarity fills that gap in terms of the you know, I’m stuck. Well, if I’m stuck, let’s put a placeholder in there, because we don’t need to have everything kind of done to field level. Sure, at the start, at the end, we might want to get there, but at the start, like we’re just drawing out the, you know, the vision for the future, right?

David Sweenor 8:27 So this data lineage is just data lineage. Tour, can help with AI lineage and things like that.

Philip Dutton 8:33 No, I think, I think, you know, as I said, kind of conceptually, whether it’s an AI that’s utilizing data, or whether it’s, you know, a peer, a BI dashboard, or whether it’s, you know, another system that is pulling data like conceptually, they’re all consuming some data, right? And that data has particular purposes that it’s allowed to be used for. There’s certain obligations that come with it. So you know, you can use this data for this purpose. You can store it for this long. You know, you can share it across these different lines of the business. All of these things kind of naturally come with the AI, as they say, in what they do, kind of any data product that an organization has created. And so I think the great thing is, is this means that you don’t have to change your operating model for AI governance. You can use the same operating model that you’ve been using, which the organization knows, and it takes them a long time to get to know it and to feel comfortable with it. So this really gives you, like, a nice accelerator.

David Sweenor 9:29 Yeah, that’s a huge value. So something that was built for compliance is now incredibly useful for AI. So I think that that is amazing. So you’ve talked to a number of people at the booth, what’s on their minds? What is sort of the key question

Philip Dutton 9:41 you’ve gotten, yeah, I think, you know, there’s the question of sustainability. So, you know, the automation piece, like, how do I not have to do this manually? I think we’re all sick of having to do things manually, because it’s the challenge comes, I think, is we do it partially, and then we stop, then we do it partially, then we stop. And so we’re always in that manual. Phase. So this is, like, the last mile, yeah, exactly. And so what we want to try and help them is to, you know, it’s, it’s like when, I guess, Google mapped out the earth, like there was a big heavy lift, and they did it, but once they did it, then it was a small, incremental changes. And so that really means that you move into a different type of operating model. And that’s what we’re trying to get customers to get to, is whether the heavy lift comes from full automation or partial automation or fully manual, it doesn’t matter, as long as we get that heavy lift done once, and we keep it evergreen, then we’re in a process where the organization can maintain it at a low friction and a low cost. Okay?

David Sweenor 10:36 And maybe one last key topic I’ve heard. I don’t know if you’ve heard the term agents. It’s all over the place to have good agent and hygiene and all that. You need a good foundation. Is that accurate? Yeah.

Philip Dutton 10:48 And I think kind of the a lot of what people are talking about, our agents aren’t really agents in the true sense of the word. So, you know, when we talk about, like, agentic AI, like, what does that actually mean? It means that the the AI can go off and autonomously create its own plans and execute on those plans and make its own decisions to reach the outcome. A lot of the agents that people are talking about are more chatbots. So request response. I say something you respond. I say something you respond. They don’t have the kind of the full capabilities that allows it to take a request, build a plan about how I’m going to execute that request, to achieve the outcome, and then to go about those and to correct itself along the way, if it goes well that didn’t work. Like what’s my alternative? And so with our agent, again, built in to try and really be able to process complex steps. So you know some of the steps you’ll see kind of it might run, 20, 3050, different steps. Wow. To get to the the outcome that you’ve requested, and in that, you get far more bang for your buck. Because, you know, now, when I was having a conversation, now I’m saying, do my work for me, and it’s doing my work for me, and I’m checking and

David Sweenor 12:00 balancing it though. Yes, I like that. So I know it’s hard to predict the future, but what’s on the horizon for solidatus? Where do you see it headed? I know technology is changing so fast. What we know today, the landscape is definitely different tomorrow, but kind of, where do you see the future of solidace headed?

Philip Dutton 12:16 Yeah, I mean, I think, you know, when I think about how we thought about the problem at the start and kind of how the kind of the industry has evolved. I think, you know, we really set out to help organizations capture the intellectual property in the organization and make it exposable to the organization. Now, when we were thinking about that, that was very much a human driven thing, right? We wanted to get it out of Bill and Bob’s heads and into some repository that was easy to interact with so anyone in the organization could access information. So we didn’t have, you know, legacy stuff because of, you know, a lock in, because no one understands that technology anymore. And so when we think kind of the rolling forward, that context is now what everyone is talking about as the great unlock for AI, right? It gives the organization the ability to understand itself, but at a far more like deeper level than kind of at where llms are currently operating, right, where people operate, where they understand the different products that are traded, the different lines of business, the reasons why a decision was made, one way what technology was was used to support it, because of it may have been a political reason. It may not even been a technology reason, right? But at least that context is important to the decision that we make for the next decision. And so that’s really, I think, where solid eight is really sets ourselves apart is we can intersect vast sets of very, very different data that naturally doesn’t live well together anywhere else, right? And I think that, for me, is going to, I guess the greatest unlock is the flexibility to allow an organization to model itself based off what it needs to understand for it to go forward. And that’s, I think, what the future is, is continuing to doing that, but doing that no longer just as a human interaction now at a, you know, an agentic kind of orchestrated level approach.

David Sweenor 14:18 All right, so with that booth 929, you guys have a talk tomorrow with lseg.

Philip Dutton 14:24 Yes, that’s right. So London Stock Exchange and ourselves are talking on kind of the work we’ve been doing with them, on their data trust program. We’re also going to kind of take that and extend it a little bit further from just, you know, a typical internal lineage use case and metadata to well, how do we make this value add for them externally, so interacting with their customers and their clients, and so that’s a really powerful kind of presentation that we’ll give. And then we’re back to the booth to do some launch activity around the agent. We’ve got some live interactions really showing some of the power, some of the new capabilities, which. You know, solid data is not known for but now kind of is providing to customers,

David Sweenor 15:04 all right. Well, I’m super excited. Don’t miss that session tomorrow. So coming to you live from the Gartner data and analytics conference 2026 in Orlando, Florida. Cheers. Thanks for.