Demystifying Data Intelligence: What it is and Why it Matters in 2026

Hello everyone. Thanks for joining today's call on demystifying data intelligence, what it is and why it matters. I would like to introduce our speakers, Emma McGrattan, CTO of Actian, and Matt Aslett, Research Director of ISG.

Before we begin, I would like to mention again that this call is being recorded and you'll receive a copy of the recording. Additionally, we welcome your questions and we'll try to answer as many as possible on the call. With that, I will turn it over to Emma.

Great, thanks Danielle. So Matt, thanks for joining us today. So for those not familiar with ISG, can you tell us a bit about ISG and about the buyer's guides?

Yeah, sure. Happy to, and yeah, thanks for having us on the webinar. It's a pleasure to be here.

Um, yeah, for ISG is a industry analyst. Um, in particular, one thing perhaps people don't know is we actually influence over $200 billion of enterprise technology spending. So quite significant influence.

A lot of that is the core business is focused on service and technology providers and engaging with enterprises advising, you know, the enterprises on their service and technology provider consumption. So the software research part of the business that obviously I'm part of comes under research and was actually acquired into ISG well, a couple two and a half years ago now. So it's expanded into the software side, but the advisory, benchmarking, governance and, and service provider part of the business is where the core sort of business is done.

So our software research covers two main areas. It's really sort of the underlying IT and technology platforms, analytics and data AI, digital business, digital technology, and that's part of the team that I'm part of. And then the other side is all the applications that run on top of that.

So customer experience, HCM, Office of Finance, Office of Revenue, and we've got a, you know, our colleagues in the team that addressed that side of the software research of the software business, as well. So in terms of the buyer guides, obviously we're talking, we will be talking more about the data intelligence buyer's guide, but our buyer's guides are really something that's been going on since through history of the software research business, so coming up, actually, just over 20 years. Um, and it's this, you know, methodology that we've, you know, tweaked and obviously hopefully improved over those years.

And we started, so we used to be known as the Value Index Reports. Value index is still the methodology, but we rebranded those as buyer's guides about three years ago. Essentially what we're doing there as, you know, analysts and research directors, we're putting ourselves in the shoes of someone who's buying technology or looking to buy particular software in a particular category.

In this case obviously data intelligence. And so we say well what would somebody who's looking for that software need to know? And so we obviously draw up a list of providers that we think are relevant that would be assessed as part of, as sort of a R-F-I-R-P process.

We draw up a list of all the functionality that we think should be addressed. And that includes not just the core capabilities of, in this case, data intelligence, but also things like reliability and availability, user, you know, the user experience. And those are things that apply across, well you know, all the categories of buyer's guides that we do.

And then in addition, we look at kind of the customer experience. So, you know, if you are working with a provider, you know, what kind of assistance do you have in terms of understanding the potential value or measuring the potential value of that investment? What's their onboarding process like?

How do they, you know, what's their support and services capabilities like? So, you know, we assess all of those things. So essentially what we do is we send out the information to the providers.

Some of them choose to participate and send information back. Some don't. Either way, we assess either the responses or, you know, whatever publicly available information we have, just like, you know, an enterprise would do if they were looking at providers they were looking to invest in, and we run through the questions, we give them a grade from A++ down, to where each question has a score.

You apply the grade and, you know, you end up anyway with this matrix here, which is, you know, obviously the, the buyer's guide quadrant.

And obviously the goal for a provider, I suppose, needs to be in that top right, as you'd imagine so, which we call exemplary. And those are basically the providers that score the highest in terms of both the product experience and the customer experience. So yeah, that's it in a nutshell.

I'm sure that we can discuss more as we go through here, but that's essentially what we doing. And we produce, we produced nearly a hundred of these, I think last year in various you know, software segments. Wow, okay.

So interesting. Thank you. So you're talking about this to a lot of people and I think top of mind for many people on the data side is data prep for AI.

So what are you hearing from them when it comes to data prep for AI? And, and then kind of a follow on from that is how do you define AI ready data? Right?

Yeah. So it's funny, we had, uh, we've been running for the last couple of years, ISG AI events. And the first one we had actually was in the UK and I was at that event, and it was quite funny because there was a lot of enthusiasm, obviously, around AI, everyone talking about, you know, their AI projects and getting running with their AI projects.

And there was an element, I actually, there was an element of being sort of that guy who's going, well, yeah, but don't forget about the data. Don't forget about data quality. Don't forget about data governance.

And to be fair, you know, it wasn't completely ignored by others, but it was what is sort of kind of fascinating in like the year since then, it was coming up to our event, this, you know, towards the end of last year, it was completely different. What everybody was talking about was the need to sort out their data management, their data governance in order to accelerate AI projects. And, and I think, you know, there's been a lot of focus on that MIT study and failure rates.

And I, you know, I take that with a bit bit of a pinch of salt because obviously we're in an innovation stage, so there's projects, some of them are supposed to fail, right? So you can fail fast and move on to the next thing. But also what we definitely have seen is, you know, you see here the number one challenge that we see in, in this, this is from our data market lens study, number one data challenge is getting that data usable for AI.

And I think we've seen a lot of enterprises have realized in order to go from sort of a focused, domain specific project, proof of concept into something that's more substantial, particularly if it's across the enterprise, then they need to have their data, as you said, AI ready and, you know, there's what we, the way I think about that is in terms of being AI ready is, you know, it needs to be clean and it needs to be well organized. Um, and it needs to obviously be compliant with whatever regulations are relevant to that project. But also it needs to be, you know, kind of, it's aligned to the specific business goals and objectives and KPIs of that project, right?

And I think, you know, that that sometimes is kind of overlooked, like, well we've got all this data, it's in a, you know, whatever, a lakehouse, whatever environment it is, and we, we can now make use of it for lots of different AI projects. And that is potentially true, but there is this, you know, need to understand the goals and the requirements and the KPIs, and then apply that to the data that is being accumulated and prepped and processed for that specific project. So yeah, there's lots of complexity involved in that.

Obviously. One other thing we see is, you know, having good quality clean, well processed data has always been an enterprise requirement. So, you know, all this is in some ways is doing is shining a light on the problems that enterprises already knew they had, but have kind of maybe, you know, let, well, let's forget about that for because these are complex projects and problems.

But I think if you are going to move ahead with a substantial AI initiative, then you have to address these challenges. Maybe not beforehand, but as you go through that process. Yeah, good point.

I mean, so we talk to customers all the time about data quality, and it's like, you know, why is there a bigger lens over the whole data quality landscape today? And it's AI moving at machine speed, right? With analytics, you have a human in the loop, and they were like ya those numbers aren't right.

Right. I know I did a big deal last week. It's not reflected in the numbers here.

So, you know, let's rerun the report, whereas machines aren't gonna be doing that for us. So yeah, it's a really interesting time and data. Yeah.

Yeah. And, obviously, you know, as you suggested the, at least the potential for agents to be making automated decisions really highlights that because as you say, you no longer have the person with the domain knowledge to be able to spot, hang on that, that doesn't look right or that's, you know, there's some, difference there between the whatever dashboard I was looking at yesterday and what I'm looking at today. Right.

Uh, yeah.

So it absolutely needs to be in place to, enable those kind of initiatives. Awesome. So we obviously participated in your data intelligence buyers guide.

But for those who have joined us today that aren't familiar with the term data intelligence, can you shed a bit of light on that? What is data intelligence? Yeah, happy to do that.

I think, you know, to begin with, I think we just to illustrate that we, why we did this buyers guide, and why we think data intelligence is important, you know, we do see, we believe a significant proportion, you know, three quarters of enterprise will be engaged in what we term data intelligence initiatives, um, through, the end of next year. And what we, the reason we said that is about understanding how, when, and why data is used in the organization and by who. It's one of those things that I think, you know, organizations you think ought to have a handle on, but very few do.

Um, you know, in particular the reason for that, (I didn't build that out) Um, you know, there's multiple participants in relation to any kind of data project. And, you know, to simplify, we think about that in terms of data producers and data consumers, and they've got different requirements or they've got different sort perspectives on, on the data. And historically, you know, all of these things were being addressed within the organizations, but historically with multiple tools that, you know, the data administration, the IT teams had to, you know, keep working together and integrate and keep in sync.

And I think what we've seen is data intelligence as a sort of a product category, at least in the way we look at it, kind of emerged from data catalogs and the evolution of data governance as platforms that can potentially serve that role of a single platform for addressing those multiple requirements. So we define data intelligence as software that provides specifically a holistic view of data, you know, production and consumption that enables data administrators to understand and manage the use of data in BI and AI initiatives. And it that enables them to then accelerate strategic democratize data democratization initiatives to provide, you know, analysts and business users with governed, self-service access to data across an enterprise.

So there's a lot in there, there's a lot to unpack, but that's the way we view the market. And, you know, it is a term that I think we've seen a lot of, you know, several providers have used over the years. And for a while actually I was kind of skeptical about it being a category because providers were coming from different places on the, you know, this map here.

But actually I think we have seen the market coalesce and it, it is become clearer that what capabilities are required to address, you know, all of these requirements. So yeah, hence why obviously we started using it because we identified, yeah, not only was it a category, but there was a very real need for it, you know, as a complement to investments in more traditional sort of products for data management, data integration, data quality and data governance. Obviously, none of those things go away in and of themselves, but there's this broader data intelligence category.

Yeah. I mean, we've seen a shift kind of from that modern data stack where it points solutions best of breed for various things, right? And people now anticipate that the intelligence platform should deliver all of it, right?

And whether we build this out through partnerships or we build it ourselves, the expectation is that it's all integrated, right? That if you're using something like a data product, that that data contract behind the product has observability on it, it's making sure that they can trust the outcomes. That problems arise, you've got the ability to kind of hit a kill switch and say, okay, we need to fix the data pipelines, you know, until we do that we can't continue with this initiative.

But we, we definitely see from the Actian side that people love best of breed, but they kind of want it all to work together. And the easiest way of doing that is kind of coming to a single vendor with some good partnerships. Yeah, there's definitely, you know, it's funny how sort of the pendulum swings back and forth and, you know, you mentioned the modern data stack and I think, you know, there was lots of good reasons why, you know, obviously those products, a lot of those products emerged and they were adopted.

But I think, you know, we saw that as, I used the phrase of the modern data smorgasbord. Because it was like, hey, here's like hundreds of things that you can choose from. But it's up to you to get them all to work together.

That's good, right? Yeah. So yeah, I think, you know, obviously, all enterprises are concerned about lock in and putting all their eggs in one basket.

But definitely I think the pendulum has swung towards, you know, unification and, you know, having all these, most of these capabilities provided by a single provider, particularly because they all do work together and they are all inter reliant. And if you are stitching that together yourself and sort of, you know, with duct tape and hope, then, you know, at some point, particularly large scale, that's going to cause you problems. Yeah.

Particularly machine speed too, right? Yes, absolutely. Yeah.

Alright, so we've kind of defined data intelligence now for the audience. So in terms of functional requirements for data intelligence, how is ISG thinking about that? Yeah, so I mean, I think you can see obviously some of the key capabilities here in terms of obviously, you know, that governance layer and, obviously security is going to be part of this, absolutely.

You know, data quality, and you mentioned data observability, and perhaps we can get onto this, but you know, how those, you know, to some extent overlap. And, and I think part of what the reason why we've seen this drive is some of the existing tools, the traditional tools, as you were kind of alluded to in your question earlier, we're sort of applying capabilities after the fact. Once you've built the desk dashboard, run it through data quality, ensure that it, that's, that it's, you know, it's of a high enough standard.

Whereas data observability is very much about tracking the pipeline and ensuring, well, trying to ensure that those data quality challenges don't occur in the first place. And obviously being a lot more adaptive, data lineage and understanding usage, you know, whether they just come from who's touched it, what they've done with it, how it's been, how it's been used is obviously important. And then from the consumer side, obviously we saw a lot of focus on self-service, data discovery of, you know, again, as I said, sort of emerging from that catalog space related to quality observability is this issue of trust.

And you mentioned data contract is really important here in terms of understanding now, not just understanding, but having this agreement about what the quality of that product will be and how it'll be delivered. And understanding the nature of that data because it's one thing to, you know, be able to search for self-service, you know, on a self-service basis, search for data. It's another thing to actually understand what is that data, what does that, and particularly in relation to, you know, data definitions within organizations, to understand what all that that data means and what the business, you know, applications and objectives related to that data are.

So here I just pulled out, so this is related to the buyer's guide. And so I mentioned earlier we, you know, we grade from A++ down to F and so what we, with some of these things we've picked out what are the proportion of providers that are scoring A- or above, so you know, the highest level. And you can see some, you know, obviously some of the key capabilities that we see in this space.

Obviously metadata extraction is pretty, you know, it's pretty much, well, I was gonna say table stakes, obviously there's a lot of differentiation in how you deliver that. But you know, the fact that that is a requirement is key natural language discovery for, you know, search base access to data is another, you know, pretty widely adopted. And then obviously the, you know, an emerging space.

This sort of the knowledge graph capabilities that, actually tie in. You mentioned data as a product ties in very well to data as a product as well in terms of, you know, providing that representation of the data and the relationships that are related to that data and the definitions that provide that common understanding of the data. So, you know, less well adopted by the providers we're assessing at this stage.

But obviously, you know, we do see that's a, it's a big area of focus for investment by either new providers in this space or some of the older ones that are, you know, trying to catch up with some of the innovation that's happened in the last few years. So obviously data intelligence is definitely helping companies deploy their AI initiatives through to production. But there's also a role for AI in the delivery of data intelligence itself, right?

So using AI for things like glossaries and natural language interfaces to the data and so on. So, do you wanna talk a little bit about that? Yeah, no, absolutely.

And, you know, we do see, I mean, obviously all software provided, well, almost all software providers are investing in AI capabilities. If they're not, they absolutely should be. Yeah.

And they're getting that. Um, but yeah, as you said, data intelligence in particular, sort of data management capabilities are really ripe for, you know, acceleration using data, using Gen AI. And so yeah, we believe that almost all data intelligence software providers definitely will be delivering support for those kind of capabilities, you know, throughout, uh, throughout next year.

Um, yeah, as you mentioned, some of the key areas, identifying, classifying, maintaining, you know, those maps of relationships, creating those knowledge graphs, automating the generation of data descriptions, summarizing data, you know, obviously as always with Gen AI, you know, you have to be wary of making sure that the output of the, that's generated is correct and accurate. So, you know, there are obviously those caveats, but you know, there is certainly, you know, significant areas where a lot of acceleration can be done in terms of reducing the amount of, you know, really critically important, but also mundane and time consuming, right? Yes.

Mind numbing me tedious is building out a glossary, right? Yeah, yeah. Absolute, but hugely important.

Yeah! And that has to be done. So obviously anything that can accelerate that.

I think, you know, in this space, you know, obviously there, we are genuinely looking at tasks that could be accelerated to enable those people to then focus on higher value tasks rather than replacing them. And obviously clearly if you, as you said there, if you're doing, you know, generating data quality checks, if you're generating summaries of data, you know, things like that, that has to be, you know, checked by, that has to be human on the loop to check that and um, and qualify that so that the data can be trusted then by those that are finding it. But yeah, I mean, and again, you know, so these again, are percentages of providers that are scoring A- or above, so it, it gives you an indication, even though as I said, you know, we anticipate that all providers of data intelligence software will be investing in these areas over the next couple of years.

Some are further ahead than others. Um, and, uh, so there's a lot of investment ongoing in this space. Um, uh, but it's definitely an area where I think, you know, we see agents as well can deliver additional value in relation to automating some of these processes.

Yeah, I mean, it's exactly what we're doing, right, and no surprise. Um, so yeah. Okay, great.

So now, yeah, we talked about earlier when you were defining data intelligence, right? We had the, the data producers and the data consumers. And you know what, I've seen a lot of successes in organizations where they're using data products where they've defined a data site owner, uh, I'm sorry, yeah, a data engineering side owner of the data product, right, and a business side owner.

And that helps with consumption because the two teams have agreed on what it is that they're going to produce and what it looks like and what's quality like, and and so on. And that kind of helps with the, with adoption. So how are you seeing that?

Are you seeing this? Sorry, go ahead. I mean, this is definitely one of those areas where there's, you know, huge, you know, room for improvement.

And it's one of the, I think it's one of those areas where our organization, many organizations have wanted to improve that they've known this has been a challenge that they, to get these two groups of people within their organization to collaborate in order to improve the delivery of, you know, data and analytics initiatives. Oh God, I'm not gonna go into the reasons why that there, there are many reasons why that is difficult and each organization is different. Each organization has different challenges with that.

But I think, yeah, that's one of the reasons why we've seen a lot of focus on data products. You know, it's not, um, you know, it's not going to so solve everything overnight. You have to adapt your business processes and your, you know, your collaboration processes additionally to take advantage of the technology.

But as you just described there, it focuses attention on defining, you know, the data, the data inputs, the outputs, the assumptions, and the goals, and the KPIs. And I was especially, as you said, if you are building data contracts into that process, which we see, you know, most organizations, well, most providers are in the process of delivering those capabilities to enable organizations to do that with data intelligence software. That really provides the focal point and that focus of agreement within those two parties, that gives, you know, the users the confidence that they can take advantage of that, data that they find, they do understand what it's for, what, you know, what the the definitions are, and they can then accelerate their use of those data products.

Yeah. So I was on a conference recently and somebody came up to me after I'd done my keynote and they were asking, do I have to wait until I fix all of my quality issues before I implement governance? I'm like, oh God, you're never gonna fix all your quality issues.

Right? There's gonna be new sources of data. coming in all the time.

It's like, don't wait. Right? But I have to point them to this concept of data contracts and data products and say, okay, let's identify one small project, show success with that, right?

You don't have to boil the ocean, just like narrow the aperture down to one app, right? And let's look at quality for that one app. Let's deliver a product and let's deliver success.

And that's how I love to think of data contracts and data products is kind of giving you the opportunity to show success because everybody is like, I think diluted by how quickly we're gonna deliver production grade AI into the enterprise. And they're seeing so much success with, you know, the use of ChatGPT in their personal lives and making their emails look a lot smarter and so on, that they're expecting that the business experience is going to be similar and it's going to take us some time to get there. So identifying where you can get some quick wins, I think is important.

Yeah, yeah. No, definitely. And I think another aspect of that in relation to data products is the, you know, the main based focus of data products.

And so you talked there about, you know, quick win or, you know, defining the scope within a particular business department or domain and identifying the KPIs around that. Obviously with the goal of making the output or the outcome available to others within the organization, which is part of the whole point of the, you know, the data, product approach. But it does provide that focus, which I think is, you know, is one of the reasons why, you know, everyone can have really good intentions about improving, as I said, sort of collaboration between data producers and consumers across the organization.

If you try and do it across the entire organization, that's a huge project. If you can do it at a project, on a project at a time focused on different domains, then yeah, it alleviates the process at the very least. So, I mean, talking of domains, are you seeing a lot of federated governance being deployed where each department kind of has some level of autonomy around governance for their particular domain and bringing in, you know, data nerds and, and data passionate people across the various domains in the enterprise?

Yeah. Data passionate people rather than data nerd. Let's say data curious is probably a better term for it, right?

Okay. Yeah. Yes.

Um, what I think is interesting, so let's look at it, so with the sort of data product concept obviously emerged with data mesh. And I've seen, we've seen l less adoption of data mesh as a whole. But I think the data products concept and the domain based delivery of that, and as you said, federated governance has then obviously related to that.

Right? You can, obviously you can adopt data pro. I think, well.

We see different organizations approaching it differently. Not everyone is going, 'Hey, we're all going. We we're going all in data mesh.

We're all in federated governance.' It depends, obviously we've done some interesting research recent last year around organizational approaches to data, you know, whether that is distributed, the IT responsibility. Is that distributed across different business units? Is it with centralized team?

Is it a, you know, a mixed model? And of course, every, really every organization is going to be slightly different. So I think lots of organizations are figuring out what approach works for them.

But broadly speaking, yes, I think we see more of this, of the federated approach because there are some really good advantages that we've seen from that particularly in relation to, improving collaboration and accelerating the delivery of projects. But you know. Every organization, as I say, is different, and therefore their success with that is going to be predicated on many other factors rather than just adoption of technology and deciding, right, this is what we're going to do.

You've gotta actually, yeah, it's easier said than done. Let's put it that way. Yeah, very true.

Alright, so we had a few questions submitted in advance of the webinar, and we've had a few questions submitted at why we're online. So I think the first question here and that is, 'Has AI intensified the value of data intelligence over the past 12 months?' And I mean, from my perspective, absolutely.

Right? Not only has it increased the value of data intelligence, but it also has exposed the cost of not having it. Right?

We talked earlier about how if traditional BI, didn't have to be perfect, right? You had a human in the loop that could catch glaring errors, but for the most part, the answers you got were good enough to make decisions for the business. But with AI, and, you know, the speed at which agents and LMS are consuming data, you can't always have that human in the loop, right, to make sure that the data is of sufficient quality.

So you have to use observability, data quality, and so on. So yes, to me, data intelligence has shifted from a nice to have to, to you absolutely have to have it. But Matt, from your vantage point at ISG, are you seeing organizations moving from viewing data intelligence as a kind of a modernization initiative when it comes to data management, to treating it as an absolutely foundational infrastructure for AI?

No, definitely. I think, you know, as I said, we've seen, many organizations, not necessarily, again, I go back to that MIT study, it's not like they're abandoning AI projects, but I think that, you know, you can do your proof of concept. And we see companies doing this.

Do the proof of concept, prove the value of AI initiatives or the small scale, but also recognizing if you're going to, if they're going to take this out across the organization, if they're gonna move from, you know, a couple of projects to 10, you know, tens of projects, then they need to have a base level of confidence in not just, you know, some of their data, but all of their data. And the data in investment in data intelligence and the capabilities we were talking about earlier is, you know, potentially a way of, of delivering that. Because just as you said there, driving some of those processes that were previously done after, you know, the creation of dashboards and reports and or even AI projects to the to front load those in terms of observability and defining, you know, things like KPIs and, data quality expectations upfront.

Um, because as you said, in terms of sort of the automation and the, you know, the increased use of agents and the fact that there is, you know, some of the these capabilities need to be guaranteed. Uh, so they can have confidence to adopt those. But also in terms of, you know, if you are looking at large scale AI initiatives, then you are looking, you know, depends obviously on the project, but you can assume data coming from multiple different data sources, multiple different, you know, what might have been silos.

And so you can't do that on a piecemeal project by project basis. You need to have a base layer of confidence in the majority of your data. Obviously, you're not gonna use all your data every time, but yes.

And so therefore, a lot more focus on data intelligence as a way of potentially achieving that. So kind of a follow up of that, Matt, where are firms underestimating the operational risk of deploying AI without a governed metadata and semantic layer? Oh, numerous.

This is a surprise Question, right? Yes. Well, let's see.

I mean, you know, I think it's interesting, the concept of risk. I think, you know, one of the, well, there are multiple reasons why there's a lot more focus. You think about, there's a lot more C level driven AI projects and there's huge expectations, you know, realistic or not about what can be achieved using AI.

So, you know, the risk of, you know, that's more of an internal risk of failing to deliver what the c, you know, the board expects you deliver is huge. Obviously there's enormous, potentially, you know, regulatory risks and legal risks. There's huge potential ethical challenges.

And, um, that's all, I'm missing the word, but, you know, in terms of, you know, the reputational risk if you, if you know, projects are not delivered, well, one, if they're not delivered successfully, but two, could be they're delivered successfully but you haven't got the guardrails in place to control what the AI is doing. You expose like a massive bias. Yeah, yeah, Yeah, exactly.

So, yeah. The, you know, the risks are many and numerous and absolutely, you know, those are all things that are potentially, have been existing risks in relation to projects, but the level of it is absolutely highlighted, you know, is increased, for each of them. And, you know, and especially I think what's a little bit different, you combine all those together when you're looking at, you know, at particular agentic and generative AI initiative, and yeah.

So multiple parts to be governed. And that and hence why we've seen a lot greater focus data focus on, you know, governance, not just data governance, but AI governance governing the models and all that stuff. And the combination of those two.

Yeah, absolutely has to be addressed at the same time. Okeydoke. So our next question in advance, 'How do you balance data quality work with speed, especially if you're in an early stage environment?' And it's a great question, right?

Because early stage environments often think that governance is going to slow them down, right? But then the reality is a lack of governance is, is really going to slow you down later. And I think, you know, when it comes to, you know, governance and quality, that 1 -10-100 rule still applies, right?

So I love to give the example of, I moved to LA straight from graduating college in Ireland, and I went to get my driver's license and I knew that in America they talk about weight in pounds, so I filled in my weight in pounds, but I assumed that if weight was pounds, the height was inches, right? So I'm five foot three and I put in 63, right? And I got a license in the post, and it had me as six foot three, right?

I'm like, oh, that's kind of amusing. So I go to the clubs, I show my license, I was six foot three, didn't, didn't fix it. And then when I got to, um, I got a speeding ticket, I was driving out to Las Vegas, got a speeding ticket, and I think I could have talked my way out of it with the Irish charm and whatever, except that when you looked at the disparity between the height listed on the license and me in person, I got a ticket, and then I had to get a lawyer, and it just became a really expensive thing.

So I am living prove to that one 1-10-100 rule is something you need to look at, right? Fix these quality issues upfront. Alright, so we've been thinking about the fact that like if you are early state and you don't want to get bogged down with governance, and you're thinking governance really as an enabler for innovation, right?

And not something that's really going to be too burdensome. How do you think about minimum viable governance and how do you put that in place so that it could scale? It, I mean, yeah, as you said, it is a really good question because it is a real, you know, it is a key challenge for organizations and, you know, yes, I mean, completely agree with you that, you know, governance is seen as a delay, but obviously bad, you know, bad governance and not failing to address governance is actually going to slow down more.

But, you know, so there is that balance. I do think, you know, as I said, where we've seen this shift from data quality being something that's applied after the fact, you know, after the data pipeline has reduced the report, then you run, you know, there's still a role for that. There, obviously, clearly you want to be assessing the quality of the output.

But, you know, with data observability has shifted that towards, you know, the beginning of the process. And again, going back to data contracts, I think where you can, you can, you know, really fairly rapidly define what the expectations are for a project right from the get go. And so, you know, then obviously there's more work to be done in terms of delivering on those expectations, but you can really quite quickly set a level through, you know, data product and data contracts of expectation.

And, I think, you know, there's lots of really good, you know, approaches, you know, these days as well in terms of not just traditional approaches to quality, but just, you know, rules and data quality checks, but actually AI driven, you know, assessments. And also, you know, I know we've talked before about like trust scores, like, you know, from a human perspective, the user's perspective, like, you know, is this data of high quality? You know, yes, it may be of high quality, all the data's correct.

Do I actually trust, am sure about this, you know, this data, are we sure about the source? Do we, do we really know where it's come from and where it's been in between? And so the ability for, I think, you know, the investment in, you know, data lineage and data observability, you know, yes, there is an upfront investment there and, and, you know, data contract capabilities.

But that I think, you know, we do see that some of those capabilities can absolutely provide, so then the guardrails that can then enable you to go faster. Um, and obviously, you know, it doesn't have to be a all in enormous sort of data, traditional data governance project from, from the start. But yeah, it is a really good question.

It is a real key challenge. Yeah. I mean, there's a real risk of over-engineering it, and then everybody tries to find ways around it, right?

So, yeah. Right. Yes, absolutely.

Yeah. All right, next question. Oh, data contracts question.

How do data contracts and federated catalogs ensure long term data product quality? This is a great question, right? Because data products degrade pretty much as soon as they're launched, right?

Uh, unless you have very clear ownership, and I talked earlier about like, we recommend when there's a data product that you've got somebody on the data engineering team that owns it and somebody on the business side that owns it. So something goes wrong, you know who to call and you know, who owns this. And I think having that ownership really helps.

But, um, federated catalogs, right? Helping with discoverability, helping with federated governance, right? Lineage I think is incredibly important.

Policy enforcement's important. Understanding context. I mean, to us, context has never been more important.

So getting a semantic there right, is incredibly important. But to me, catalogs are a vehicle for operationalizing enforcement and, uh, it kind of helps with ensuring that your data products don't wither on the vine, right? So, what guidance do you give to companies that are looking to deploy data products, whether with or without data contracts and to make sure that yeah, it's an initiative that keeps moving forward and it's not kind of a catalog of dead unknown products?

Yeah, I mean, again, this is a really good question. Uh, I, and as you've sort of addressed there, obviously the challenge is to make the data products approach sort of fundamental to how you, operate as a business and actually operationalize some of the key, you know, processes that go to delivery of those data products. And obviously maintaining their freshness, ensuring that they, you know, it's not just a one and done and, you know, they are continually updated.

Because I think this is, this is another part we haven't mentioned yet, that is a key aspect of this is that, you know, data is continually produced and these, you know, products should be continually being updated. And that should be, the information in relation to that should be available to consumers as they're searching products. Or even better, they should be alerted to that, you know, there's a new version of this data product or, you know, obviously where, where it, you know, makes sense to do so, and where they can sign up for that.

But, you know, that's about sort of, yeah, operationalizing and making it those ways of approaching data projects part of the way that you operate as a business, as you say, both in the business side and the technology side. And that it does require, you know, well, depending on the size and the age of the business, it potentially requires a real cultural change. And so, you know, and so what I mentioned earlier, this, you know, this is to certain potentially sort of easier, easier said than done, but the other thing I think is, you know, we think about, you know, data products, you know, part of what the reason, you know, we they're called data products, is they should be the, the outcome or product thinking.

Yes, they should be designed to be a product that the consumer wants first. And, you know, and has defined, you know, the expectations as we've said in terms of the quality and the value, you know, the nature of the data and the quality of the data. But also, you know, we think about how we consume and purchase, you know, products today, you know, the ability to say, you know, yes, this is a good product, no, this is a bad product.

Yes, this, you know, give a review, five stars, whatever. Yeah. I think, you know, we see in, in some cases, not all providers are building these into their, you know, their data intelligence layers into their data products sort of consumption capabilities.

Because that is part of, you know, it's not, obviously not the only thing, but it's part of keeping that fresh and ensuring that, you know, if you as the consumer can go in and, you know, yeah, the data may be however old, but you can see that somebody used it yesterday and they approved it and like, yes, this is still relevant. And yeah. You know, and that's a business user rather than the technology user, you know?

Oh, sorry. You know, the technology team. You know, that's part of it, but it's, yeah.

So it's about operationalizing all of those capabilities Yeah. And thinking of that product mindset, right? So it has a development lifecycle, right?

And it has a data product manager, right? So I think in more mature organizations, we're seeing that as a, as a role that almost guarantees success, right? If somebody owns it, that their job is product management for data products, I think that can help too.

Is that what you're seeing in more mature organizations? Yeah, no, definitely. I mean, it, you know, as I said, it does require cultural change.

So obviously organizations that have been doing this for longer. I mean, how long has sort the data mesh been? Right?

You know, it's a good few years now, and obviously companies that were early to that yeah. Have, you know, experimented. They, you know, obviously I'm not saying every, every case has been, uh, you know, successful for me off, but they found their way to what the processes that work for them and implementing those.

So yeah, more mature in terms of the adoption of, uh, of data products. Definitely. Yeah.

All right. Our next question, this is an interesting one. I'm kind of not sure how to answer this one.

Uh, what oversight is required for agentic AI networks versus single agents? So to me, this really is, you know, governance has to evolve from just providing the model oversight. Now you're looking at a whole workflow, right?

because you've got those chain of agents and agent to agent interactions, and you've got agents that might be modifying context dynamically. This one is a, it's a huge orchestration of governance project, right? It's, I'm, uh, I'm not sure how to answer that one.

How about you, Matt? What oversight required? Yes, oversight is required.

Um, Yeah, I mean, obviously we're at the very early stages of, what, well, the phrase I think was was agentic AI networks, but I think, yeah, multi-agent applications or however you wanna phrase it. You know, I think with, we are seeing a lot of interest in, in agentic AI obviously, and the development of agents. I think most of those at the moment are, you know, single, maybe not single tasks, but they're focused on very specific domain, very specific set of tasks.

They're really about automating perhaps existing processes, even if they're multiple task processes. Whereas I think the question, you know, the question is getting to actually, you know, reaching across perhaps multiple applications. Because obviously the other thing as we see is, you know, each of the application, big application providers is delivering their own agents and agent capabilities.

And so most organizations are looking at a future where they're going to have to, well, you can assume they will be requiring agents that operate across different software providers, across different, you know, reach out to, external, you know, cloud services as well. And so the, I mean, this is such a huge topic. Yeah, I mean, the, the governance is absolutely required for that.

I think in terms of, and we are trying to, at the moment, we're in the process of our putting together our next buyer's guide, which is on AI and data platforms, and obviously within the AI platform space, we're looking particularly there at AI governance and AI operations, and the ability to govern, you know, all of those moving parts and monitor and be aware of all those moving parts within an agentic, multi-agent process is, you know, is currently, I would say, you know, a really significant challenge for software providers and enterprises alike. But it's where clearly everyone is moving towards. So a little bit, so what this space, but yes, I mean, yeah, it's a huge, huge challenge to, you know, I think we can agree that that's where it, where the market is moving.

Yeah. How fast we'll get there to, to your point earlier, how fast we'll get there, I'm not sure, because I think there's a lot of really significant challenges like that, that we're really only at the beginning of, of addressing, yeah. Yeah.

Agreed.

All right, so our last question for today, and it's a super simple one. I'm joking. It's huge, right?

How do you set up a data governance strategy, right? So how do you make sure that as you define your data governance strategy, that it doesn't just become a series of compliments pages, but actually becomes a framework that gets operationalized across engineering and business? Yeah.

Nice. Easy one. Yeah, right.

I mean, I mean, we could do a whole, we could do a whole other webinar on that, right? But we could, I mean, I think that the thing that, where do you, I guess where do you, how do you set it? Where do you start, I suppose is the point here?

If we're assuming that we're talking about an organization that doesn't have one. Correct. I think you need to start with, obviously, why are you implementing a daily government strategy?

And there'll be multiple aspects of that. There'll be regulatory aspects, presumably, there'll be, as I said, you know, earlier, there'll be risk related aspects, there'll be ethical. So, you know, but all of those components, there'll be potentially business acceleration of, you know, requirements from the, you know, from the business, from the board.

So I think a key thing would be actually not to, if you're starting from scratch, not to dive straight in with what data we do, we have, what is it for? How do we give them that? But actually, what are your goals?

What and how are you gonna measure whether you fulfill those goals? Um, and maybe I'll just stop there for now, because I think that's where I start. And then you can go on.

The why, right? Yeah, Yeah, yeah, yeah. We have, uh, one of my colleagues here, she'll ask why five times, right?

So it's like, why, but, but why, but why, right? And, and until you really understand, why are you doing it right? That that is probably a project that you shouldn't start.

And, and one thing that's really surprising to me is that sometimes, you know, we'll go into a, uh, a governance, what we think is an opportunity, and sometimes at the end of the POC, they'll decide that they actually want to do nothing because they see just how big an undertaking it is. Right? And unless they can answer that question, you know, why are we doing this, maybe they don't do that.

Yeah, yeah, Yeah. And because I mean, obviously, as you said earlier, I mean, governance, you know, clearly is an ongoing process. You, you cannot be like, okay, we'll get here and then we'll finished.

But at at least if you can define why you are doing it, you can then understand the metrics by which you will measure how well you are doing as you go on that ongoing never ending process. So thanks so much for the conversation. I really enjoyed it.

I hope our audience enjoyed it also. It was very interactive. Loved that part of it.

Um, I'd love to invite our audience to visit actian.com. You can download the, the buyer's guide, So we've got on our homepage today, the ISG Buyers Guide, and then a product tour for our data intelligence platform. But Matt, really appreciate the time that you took to explain this topic and then talk us through it today.

So thanks so much and have a great rest of the day. No, thank you. Great.

Yeah, thanks. Take care. Cheers.