Name: The Data Integrity & AI Forum
Uploaded: 2026-03-19T16:31:00.037Z
Duration: 55 min 52 s
Description: The Data Integrity & AI Forum

Transcript for "The Data Integrity & AI Forum": Hello, everyone. Thanks for logging in early. We're gonna give everybody just a few more moments, to get to get settled, and we will get started here shortly. All right. Why don't we go ahead and get this show on the road? So welcome, everyone, and thank you for joining the data integrity and AI forum. Recently, whether I'm talking with customers or meeting attendees at at conferences or even speaking with some of my colleagues here at at Precisely, one one thing has really stood out, that's, you know, overall AI ambition across organizations and, customers and friends of ours in the in the industry has never has never been higher. But the 2026 data integrity and AI ready readiness report, which is a recent study conducted by Precisely and Drexel University's LeBeau College of Business, presents a a pretty complex, picture when you when you think about, that ambition perhaps, you know, against some of the the confidence that that folks are are feeling and the and the realities that they have. You know, many many leaders are expressing confidence in their AI readiness, and those same organizations that say that they're really ready for for AI or for agentic AI. They're citing infrastructure and skills and data integrity as barriers, to overall progress, and the result is this growing gap between AI excitement and enterprise, execution. So today's session is gonna be designed as a roundtable discussion. It's not a traditional panel. So we have several primary participants that I'll welcome on stage, shortly, and we're gonna have a few different topics that they're gonna help, help us kick off. But the goal is to be able to bring perspectives from across the audience as well. So if you'd like to contribute to the conversation, you can raise your hand, and we'll invite you on stage to to join the discussion. But before we begin, let's introduce our participants. If you can come up on on stage, and then we'll also take a a a quick, poll as well to to get, things going. So for today's, conversation, I'm joined by Raven Jones, chief information officer at Seaspire Andrew Brust, founder and CEO of Blue Badge Insights and Dave Schuman, who's the chief data officer here at Precisely. And I re really appreciate all you for for being here and look forward to to the conversation with. you all today. Great to be here. So, Yep. yeah, so let's go and and, go ahead and get an idea of the of the industries that are represented, here here today. So we have a quick poll that you'll be able to, to respond to. We're just trying to get an idea the the industries that were are represented, and that'll kinda help us guide some of the the conversation that we'll have, have here today. Awesome. So we see some of those coming in. Very cool. Financial services, telecommunications, high-tech. Fantastic. And one of the other things that I'm sort of interested in understanding a little bit from from the audience is, you know, maybe you could share a little bit about your biggest challenge in advancing AI initiatives within within your organization. We have a second poll that we'd like to to offer as well. Makes sense. Absolutely. That data quality and trust is one that we get quite often as well as the governance and and risk management. Antonio, I already wanted to vote for multiples in this this poll. guess before we get into our first topic today, I mean, for for you all here, the participants, any initial reaction to to some of the the votes that we see coming in through the through the first poll? Well, I'm glad to see the results for data quality and trust being such a high percentage because I I think that means the audience is focused on on substance and reality and some of the nuance involved in getting AI to be successful. Yeah. I wouldn't say it's surprising. I mean, I think these are common traits across everybody's trying to get AI stood up at, you know, record pace, and these are the common obstacles that we're all trying to tackle. And, Tony, I almost wanna see this in a right choice voting in here because I. think each of them has a a a you know, we all had to choose one to answer with, but I think we would stack those in in various different ways. And I would probably say that, you know, talent and skills gap would have come as a second or third behind the data quality and trust. That makes that makes a ton of sense. So, why don't we go kinda jump into our into our our first topic, for today? And I actually want to to bring up a a quick stat that that'll go along with with this with this topic. So one of the findings in the report that really stood out to me was about 87% of data and analytics leaders that had responded to, the survey, had said that their their organizations are AI ready. Yet roughly 40 to 43% of those same exact folks said that infrastructure skills and data readiness, some of those things that we had just just mentioned in our reaction to the poll, were still major blockers. So there's clearly a bit of a confidence, you know, to to reality gap between between the two. So I really wanna understand from from you all's perspective, and then, again, from the audience, once once we hear from our participants, I'd like you to come up as well. Are organizations, sort of overestimating readiness, or are we defining AI ready differently than we used to? Sort of what's behind some of this tension on that we're experiencing? And I I'd really like to start here with, with. Andrew. Good. I have strong opinions here. But I I think, gosh, the sort of dual reality, repeats itself quite a lot. I think, first of all, if we ask a headline kind of binary yes, no question, the the pressure in the environment is such that it's hard for people to say no because that that looks like they're cynical about AI, and there's so much pressure to be optimistic about it that, saying no is tough. It's it's, it's politically untenable in a way, and it's sort of a loyalty test. But there's another side to that that's maybe a little less cynical, and that is that, it's sort of the other edge of the sword, which is that enthusiasm is is great and it's genuine, and it it naturally can. overshadow, the nuanced approach that's involved in considering the underlying challenges. In other words, when something is new, relatively new, and AI still generative AI certainly still is, just getting it going feels like, that's that's really the big, that's really the big threshold. And often we have to get past that threshold to then start thinking of more of the nuance. But, ultimately, I think, folks tend to look at AI as you know, we can use the term game changer, I guess, or I'd rather say that it's it has transformative powers in the minds of a lot of people. And that somehow. that makes it transcend the more pragmatic challenges, on the one hand and the realizing the realization that addressing these challenges makes AI outcomes far better on the other. So I think it I think it's a maturity issue ultimately. And as the industry gets. further into AI, I think we'll we'll we'll we'll be able to get more nuance quite mastered in the way we look at things. No, thanks, Andrew. And, Raven, any thoughts you'd like to add there? Yeah. I think I agree with Andrew's assessment of it as far as the you know, if you don't have the right right amount of enthusiasm, then we will find the right amount of enthusiasm to put into it. But at the same time, there is a lot to be excited about. I I do think that some of it is a definition drift, that you sort of mentioned. There's you know, and that comes as we, start to really see what AI is capable of. And I think everybody will agree if you think back to what you were thinking about a year ago with AI or what it could do, it's very different, than what you're thinking about today. And as we push into, you know, where the goalposts are moving on some of this stuff, I think that's also a part of what's driving, that, you know, hey. I was ready. Am I ready now, under the new sort of, terminology and stuff there? You know, that's to me where, you know, we started out and maybe it was that, yeah, I have the data. I'm ready for that. I have the capability on, you know, managing that data. But, you know, do you have the governance? Do you have the security? Do you have the training? Are you capable of deploying all of these things and managing it at scale for your organization to achieve the outcomes that you're looking for? So if your definition is that's, you know, the definition of ready. versus, like, the definition of, like, to start, I was like, maybe that was the definition of being starting to be ready versus, you know, actually ready to complete the race in some ways. So I think that definition drift is real, and that's a lot of what contributes to where somebody may have felt ready last week, but, you. know, this week it may look different. Yeah. There's there's almost like a little bit of a difference between perhaps being ready for experimentation versus being ready for more of an enterprise. use case. Any anything you'd like to add to that, Dave? Yeah. And, Tony, what what I Andrew and Braywood are describing is something that I call altitude confusion. So organizations are evaluating the the readiness at the platform level. Do we have the infrastructure provision? Do we have subscriptions to the appropriate LLMs? But the real test of readiness lives one floor down from that at the operating model level. So lineage, metadata, life cycle management, governance, all those things that Raven was just bringing up here, that's where the cracks are. And so the scaling litmus test I use is pretty simple. Do you have production SLAs on your models? Do you have defined owners for model drift? Do you have KPI dashboards tied to model outcomes? If the answer to those is no, then confidence is exceeding reality, and that's the gap that will cost you. So I'm really keen to understand from the audience, would anybody like to come up and add some of their thoughts to the conversation? I see there was an interesting comment from from Jose in the in the in the chat around being able to find the correct business case that represents real ROI for for AI initiatives with the within the organization. That would it would it'd certainly be helpful to to kinda continue to hear from from the audience around around your thoughts around AI and AgenTic readiness within your organization and perhaps maybe the differences between, you know, what you see from an executive leadership perspective versus maybe those that are are closest, to the data. Certainly no pressure, but what what we'd love to hear from from folks in the in the audience as well. And then, Tony, while while folks are finding out where to raise their hand, I'll put something tag onto this as a thought. We talk about AI readiness, and AI readiness isn't experimentation. It's about repeatability. So experimentation is easy. Repeatability requires data quality, measurements, governance processes, monitoring, cross functional accountability. And most organizations have one or two of those. Very few have all four. And I see we're joined up here now by Jose. Do you have a point of view on this? Your microphone is muted. I would say the bottom left might unlock you there. Okay. Perhaps we can come back to to Jose here in in a in a bit, with some additional thoughts. We'll, we'll we have another hand raise here. Is it Levin? Can anybody hear Levin? I yet. cannot. Levin, you might be on on mute. It's the difference between enthusiasm and the more nuanced of that. Alright. Well, Levin, why don't if you have some some additional comments that you'd like to add into the into the chat, we can start to incorporate that into into the conversation. But I appreciate you having the having the courage to come up and and and and offer some of your your thoughts there. So we're we're we're gonna go ahead and actually we're gonna tee up the next the next poll so that we can we can get some of that additional feedback. And what we'd really like to understand here is, like, how you would describe your organization's approach to governance for, for AI and AgenTic, AI. That's gonna be sort of the next topic that we'll get into is some of the results that we saw in the research regarding AgenTic AI and and AI governance. Okay. So, yeah, exploring what governance for AgenTic AI should should look like. Definite definitely in that discovery phase. I think that's honest. I would say that the vast majority of folks that I'm talking to right now are in that phase two. Oh, that's a great question, Wyvan. I think we can definitely work that into this next part of the conversation. So, thank you for sharing that. Okay. I think we have we have some some solid results on there. Before we kinda get into the second topic, any any immediate thoughts around the the results of that of that last, poll? I think it's good that people are considering. I think that's I don't know. If if lots and lots of people said they have it in place, I'd be a little worried about the integrity of the results. I think I think the platforms haven't caught up yet really to the need quite yet in terms of AI go I don't. mean data governance. And they're both important. And, gosh, eventually, I hope we'll stop thinking about them separately because they're really a continuum. But right now, we have a new frontier, so we have a new set of governance requirements, I think. Yeah, I would say that it's a lot of it seems connected to that same thing of the definition changing over time as well. I think many would probably say that they've got, you know, at least some aspect of the governance setup, but that's very different than, you know, what Dave mentioned of being able to do that at scale and repeatably. And that's where we've got to get to in order to actually, you know, be able to, successfully, realize that value on a continuous basis. And, Rimena, I I would stack on top of that consistently. So. if you're, you know, doing and experimenting, you're working in Copilot. You have one set of governance tools. You're doing it in Snowflake. You've got another set of governance tools. You're over there on Anthropic. You've got a third set. And they're all coming at it from very different angles and very, different levels of maturity even of the tooling to be able to provide governance on this. So, you know, for example, we push a new model out there. We wanna be able to monitor the inputs that are coming in against the model and ensure that the questions that are being asked are questions that we were prepared to be able to serve with that set of tooling And consistent even monitor all of that is different from our tool set. Right. That's awesome. Another finding from the research, and you'll see that's what I have here up on the on the screen right now, it really stood out, is the role of governance in in building overall trust. We saw that in some of the polling results at the at the beginning. It's just about 71% of organizations that we surveyed, you know, the governance reported programs reporting high trust in their data compared to much lower levels without those governance programs in in place. So as AI becomes more embedded in decision making, how are organizations evolving that that, that governance to support trusted AI? And I'd like to layer on, Lavin's question here as well. Is it a is it a problem of governance? You know, boards are finding it hard to govern the investment, and that turns out to be on the next topic as well. So maybe we could just we could talk a little bit about the the data governance, the AI governance, as well as, you know, I I suppose the governance of the of the investments that are that are being made in these in these programs as well. And, Dave, I'm gonna come back to you, here to sort of kick off the the next topic. Antonio, I I wanna frame how we think about governance because. too often I hear it described as as a break on innovation or the the iron fist of governance. And I think that framing is is both wrong and dangerous. Governance shouldn't be viewed as friction. It is traction. And so traction is the ability of that frictional force to be used to create motion, to be able to create velocity off of that. So the report was pretty unambiguous. 71% of organizations with formal governance programs report high trust in data. And if you drop governance, that number falls to 50%, and that's not. a marginal difference. That's the difference between an AI initiative that can scale and one that stalls out at the pilot stage. But what's interesting for me is how the winning organizations got there. The pattern that outperforms everything else isn't creating a separate AI governance program. It's extending data governance to include AI. So 40% of the high performing orgs did exactly that. The ones who run up a separate AI governance function, they created fragmentation. And so that gives you conflicting definitions of what is trust. It offers two different teams solving the same problem with different tools. And so the reason governance driving trust is simple. It it operations operationalize accountability. It tells you where your economy came from, who touched it, whether it's PII, what quality score is. And without that, you're not doing AI. You're doing very expensive guesswork. So here's what I tell my team. Governance has to produce automation, not doc. Policies alone don't scale AI. Automated controls, quality checks embedded in pipelines, access rule, lineage tracking, privacy enforcing, baked into CICD and ML ops. Those are what actually move the needle. And so when governance is code, it accelerates, innovates, and rather than slowing it down. And that's the mind shift I think we need to make. Andrew, any any reaction to that or additional thoughts. you wanna. add? I I'm a little worried. I'm just gonna be so in line with Dave that it'll it'll sound almost, you know, almost automatic. But I guess governance itself, the word kind of has a branding problem. I think I think people their backs tense up. All all they hear is kinda like, okay. This is gonna be a list of things that I can't do now. But governance doesn't have to be the land of no. It should really eliminate the trust barriers that have blocked people from saying yes to AI kinda in a full throated way. It inspires confidence. It drives adoption and better results. And, I guess, terms of, data governance specifically, I think AI does two things. It makes it more critical, as we've seen in terms of, things like hallucinations and mitigating those hallucinate. Mhmm. But, again, let's not just think of the downside or or the or the intimidating part. Let's think of the upside. It makes the payoff for data governance much more apparent. So I'm optimistic about it because I think this is gonna help governance efforts overall get a higher priority and more political support. I do think we have to be mindful about the new things that AI brings in, especially agents. Don't forget that agents are identities, Right. and agents. are acting on behalf of people who have identities. Mhmm. They both have, different sets of permissions, and, sometimes it needs to be a blend of the two, and that requires some sophistication. So, let's get serious about data governance. And as Dave was saying, let's extrapolate that and extend that into AI. Yeah. Do tend to think about this. At least this is how I've been thinking about these, autonomous agents as just being extensions of the teams that we already that we already have and should be treated, as such within our our governance, programs or just new new folks that have been added to the, added to the team. Raven, any anything that, you wanna add before we open it up to the to the audience? Yeah. I would just say a very similar sort of line of thinking of, you know, trusted AI comes from trusted data. Right? So if you if you don't have that chain of value going from, you know, up the chain, you've got to solve that. Otherwise, you're gonna be impeded in being able to move at scale. Alright? We're not gonna be able to do certain things because we can't trust the data to allow you to do it. I think the piece that Andrew mentioned of that gets lost in some of that is whereas our data governance discussions often were about, like, how we democratize the data and get it out to more people to see it. With agents, you're talking about taking action on the data. So it takes it to another level, and it also both in power and in, you know, responsibility that goes with that because your the ultimate goal there is to whereas, you know, in the past, it was somewhat of my data quality and I think what's this was somewhat mentioned earlier. My data quality shows up, in the fact that my insights aren't great if, you know, I have data quality problems. Well, now it's gonna show up in, like, actual bad actions that are taken against the wrong data on your on your processes. So, I think that's a piece of it that we've got to wrap our head around. And I love the idea of an of your data governance up into the agentic space where the actions are taking place on that. So one, I'd like to, you know, get some other folks from the audience. Maybe we can give this an another attempt, to to get somebody up on on stage to to chat chat with us. In the meantime, while while someone, while we get somebody else up here, there was a a question in the q and a that I I'd like to to throw out there as well. And this was from JD, asking, what were some of the biggest misconceptions you encountered regarding your data being AI ready for repeatability as David had had mentioned? So I'll throw that to the group. Anybody that has some has some thoughts, I'd love to love to hear them. Oh, I think it's that it's the one that I was just mentioning of the fact that we're going to move from insight to action. I think there's a gap between what I would allow as far as a data quality component. Allow I say allow, but more like what we have to drive out of our data before I'm gonna allow it to be actioned on versus. looking at it for insights of information. And I think that is a big gap that if people haven't realized it yet, that's what's coming is the fact that you're gonna take action independent of having you know, yes, you'll have guardrails. Yes. There'll be human in the loop, but the goal is to take additional action against that and do it quickly. So I think that's a big piece. Raymond, I wanna one interesting thing in I go back to this misconception thought in here is we've had a maniacal focus on data quality. Like, we produce over 400 data products for our organization that really drive all of our insights that the the companies have been using up to prior to AI. What I heard from some of the early drive the early model results was the inability to to differentiate between the model not performing well and the underlying data not being accurate. And so I heard it was all lumped in, throw the baby out with the bathwater kind of a a lumping, saying the data that is in the data warehouse didn't produce the right results for me. And so that data was still the highly highly precise and accurate data that we've been managing the past decade. But now we put an interpretation layer on top of that that we don't cover in that same level of quality. And so the context that's being returned by these models and some of these are folks building their own agents out in Copilot, throwing it on top of the data, and then saying, oh, the data's wrong. You came back in there. Like, no. Your model governance is wrong. Like, Yeah. you haven't looked at the results that are coming back. And so I think that misconception, back to JD's point in here, is that you sort of are gonna cast all of the insights and all of the work that you produced in producing quality outcomes by the way that the user is interfacing with the data. And so it was accurate a few days ago in dashboards, but now that data is not accurate because you haven't trained or tested or managed the model correctly. Managing the inference side. of it is very important. And being able to to baseline your models against, like, how well is it doing against our data. Yeah. I agree. And constantly we're looking for relevance, coherence, correctness in our models, but those are the ones that are in our data fabric that IT is managing. I have an entire developer community out there that is sort of, you know, the citizen developer Mhmm. in Copilot, and they're just hammering away. And then they're like, I built a model on top of the data, and it was wrong. Like but where did was it the model wrong? Was the data wrong? It lump it all together. Right. I'm glad you mentioned dashboards, Dave, because I I do think, you know, before and and Raven mentioned action. Right? The aesthetics of dashboards are are are impressive, and they're they're easy to hide behind, and they kind of pave over, important details. But when it comes down to action, things get real crisp. If, it's hard to fudge the numbers when the action just comes out wrong. And, Yep. AI just has that rather refreshing ability to okay. You know, if you've got if you've got issues that have been long standing with your data quality and you've been able to kinda compensate for it previously, this is this. is gonna get a lot more clear what what level and what standard you're up to. Yep. So, you know, I appreciate that we had a couple people before try to raise their hands. If you want to give it another go, I know that Alex had thrown some additional instructions into the chat. You know, just ensure when you do accept the the pop up message after you raise raise your hand on the right side of your screen. This way, you'll be able to be able to to to to speak. So if there is anybody else that wants to to jump back on and and raise their hand and and jump into the conversation, please feel free. We can we can move on to the to the next topic in the in the meantime, but please, if if you would like to join the conversation, we'd love to love to have you. So I'm gonna go ahead and I'm actually going to throw up onto the the screen another stat to just sort of tee up the the next part of our our conversation. So one of the other things that the research shows is that 51% of of data leader leaders say that data quality is their top data integrity priority, which really highlights, you know, how how long standing some of these data issues are becoming, you know, becoming more urgent as organizations are scaling their AI. Think we were getting into this a little bit into that last part of the last part of the discussion. So it was a a nice transition into this into this next topic. You know? But it also points to something kinda deeper that many organizations are dealing with years of accumulated data quality debt, you know, issues that have been, tolerated in traditional analytics environments. I think we are actually just just pointing at this. Right? There's a maybe a a tolerance with traditional analytics or traditional business intelligence, but, you know, much harder to ignore when you're powering AI systems. So as these AI initiatives scale, how are organizations starting to to confront that that debt? And, Raven, we'll we'll we'll turn to you, turn to you first. Yeah. It it's definitely, a real problem. I think the you think of it like we were talking before, if it's like a it's an escalating value chain. If you've got problems at the data layer that you don't trust it, then that's gonna be a problem, obviously, at the action layer and at the, Sure. you know, Yeah. all of those as well. I think a lot of it is going to come down to, what was mentioned by Dave of basically being able to have, out of our data governance, evolve this into the realm of action and agentic AI take, you know, type of capabilities. I like what Andrew mentioned about that this is sort of like the this is the the cool way to look at it, that maybe in the past, you didn't get as much traction when you stopped at the dashboard. But now, this really highlights it, and you're gonna be able to show, well, look, we have data quality of enough, you know, high enough percentage to take these actions. We're confident in taking these actions because we have this process in place. And I think the inference component of this, and that's what that is is really the canonical models, all the stuff that the teams put in place to basically get everybody on the same page so we get the same kinds. of answers when we ask the, you know, similar questions. All of that's gonna come back into being in vogue again because now you've got an action that pays it off in the agentic space, I think, is gonna really start to show. So I I think it's a it's a fun time to be in data because now you can doubt. you can do some things with it and well, you could do stuff before, but now you can do it in a more automated fashion. And it does show off and pay off the investment, from the you know, from doing that because you'll be able to move further up that value chain with more confidence than someone that's, man, I don't even trust that what I've. got here on a foundational level is sufficient to basically give me basic insights on what I need to do, much less have a automated agent take action on that on top of that. That would be a nightmare for, you know, for anybody that's involved with compliance and type of legal type stuff as well. Now, I mean, to your point, we've been able to do some pretty incredible things with data for a very long time. For me, I'd see it just kinda opening the or broadening the tent, to a a larger audience. Now that does bring in other challenges, as well, you know, AI and data literacy, making sure that those folks that are sort of being brought into the the process not only have the the high integrity data that they need, but they also know how to go about using that. So I'd to kinda simply, it is a very exciting time to be to be working with data and working in this in this space. Oh, yeah. And I I think that's gonna be one of the evolving roles of IT is. going to be to provide that trusted access and trusted action against you know? So I can have some citizen developers, but they're working against canonical models and understanding and stuff. Yeah. That's you know, whereas maybe I had to put people into some of those areas and try to go do it for you before. Now I'm I'm basically handling the model and the trusted aspect of it and the security aspects and such you know, setting you up to be able to run Right. what we're, you know, what we're all about. So I think it's also a change in what IT looks like in the future for enablement too. Dave, how are you seeing organizations or ourselves, you know, starting to confront some of that that that data quality debt that the, research, surfaced? Well, I I think it's changing the urgency of the data quality debt. It's no longer a technical problem. It's a risk problem. And we've been carrying this debt for years. CTOs, CIOs optimized for delivery velocity and came up with a like, we'll we'll we'll fix the quality later. So quality was framed as an IT cost center. It didn't have KPI accountability, and it never really competed for budget against the next shiny initiative. And. so what happened is the debt just accumulated. The organizations I see getting ahead of this are treating data quality not as a cleanup task, but as a control. They're investing in data observability, automated alerting, anomaly detection, scheme adrift monitoring. The data quality is a dynamic continuous capability rather than a static gate that you run through before you complete a project. And so that's a shift. It's from quality as an event to quality as an operating condition. Andrew, any any closing thoughts before we. open it up to the audience? I I look at AI kind of almost cut in a cartoonish way as a big magnifying glass and a big spotlight. So when it comes to data quality, you know, flaws forget the the dashboard and how it might obfuscate certain things. But if if if thoughtful humans were working with the data and had domain expertise with it and there was a flaw in the data, those those humans could could, kind of improvisationally course correct and compensate for. it. And AI is more automated. It will it will try to get the best answer, but it may have completely different heuristics for that than domain expert humans would. And it's gonna put a lot of transparency around quality issues. So the the debt we're talking about, the bill the bill is definitely coming due, and AI is, of course, Yeah. in function. And, ultimately, I think that's good because Dave Dave's right. Right? This is this is gonna make it not just look like a a matter of elegance and crafts craftsmanship, but something that's really important to risk, and manageability and control. Andrew, one of the interesting things is when when I was looking at some of the slides that are being produced throughout the organization, and I was thinking, alright. How is AI going to take over some of this analytical function? I started looking towards the bottom of the slides, and that's where the footnotes exist. Mhmm. And the footnotes are the exact that's a that's a symptom of exactly the thing that you're talking about there where, you know, like, oh, we we spackled over that crack with a footnote. that said, oh, by the way, this only accounts for years, you know, '23 through '25. You can't really do '26 or '24 with this data. Whatever whatever those are in the footnotes, that's that's where the cracks are. And you have to think to yourself, well, how is a model going to be able to to understand all the footnotes that exist in our data? And how is it gonna have the emotional drive to feel the need to cite those footnotes? You know? It's they're intelligent, Yeah. but they're not human, And they don't think that they, the models, don't really succumb to the same the same pressures that we do. And some of those pressures are positive pressures that keep the business healthy. So, you know, we'll figure it out, but it's it's important to realize what the, you know, what the deltas are. Is there anyone in the audience that has any thoughts on how organizations are starting to confront that data quality debt? I kind of think about it as a really long kind of pharmacy or drugstore receipt, you know, that we've all seen. Like, that's what it looks like now. I'm really curious to to hear from from the audience on on this topic. And it's safe space, so all thoughts are are are accepted. No brave souls? Totally fair. Okay. Hey, Jay. So, doc. Hey, Jay. Up here. oh, we do have JD up here. Out of. the q and a. In Out of the out. of the q and a and into real life. We're gonna find out fundamental if we have a platform problem hear. me? I think I I lost my connection moments here. You're. good. Great. We got you back, Andrew. Yeah. Right. Alright, JD. Can we get you off mute? That is that is our big question. And in the same way, if you don't do have data quality, you can't get your data off mute. Okay. That'll be something we can take a look at moving forward. But one of the thingsoh, have aJose, jump back into the chat. So from my point of view, data quality is one of these existential questions, type of who the the who's first the the the chicken or the or the egg? Yes. I it is something that is, or just fundamental and foundational to to all of this, not just to the to the AI itself, but really to the the success of organizations that are focused on being data driven. And I would I would argue that if you're being, you know, competitive in the marketplace right now, that's exactly what you're attempting, attempting to do. So it is it is really, foundational and, primordial in in in that way, Jose. Alright. I'm going to, I'm gonna throw, one one more, stat up onto the screen as we get into our final topic for today. And one of the other things that came up in the research is the is sort of the gap between expectations and measurements. You know, we have about 71% of leaders say that AI initiatives are aligning with business goals, but only 31% have metrics tied to those business KPIs, which I've always found really fascinating from the the results of the research. And at the same time, 32% expect positive ROI from their AI initiatives within within that, that first year. So the question really becomes, like, how do organizations actually prove, AI value? So from your perspective, you know, what separates the organizations that are successfully translating AI initiatives into measurable business outcomes from those that are struggling to really show impact, quantify that impact, be able to share that with, with their stakeholders? So, Andrew, I'm gonna I'm gonna come back to you here for for our final topic, Okay. Great. for today. So I think this gets back to kind of the magic factor again. Right? That that there's there's a there's an appeal of AI that it's so transformative that it it makes us think it it it changes the rules around precision and and the metrics that you measured. So, and the the power of seeing that alleged magic kind of divorces us from more down to earth considerations like actually, managing what you measure and therefore measuring it. I I think once. the the new car smell of AI's power wears off. Right? Organizations are gonna get much more attuned to the quality of the results, and the importance of measurement is gonna become more real again. And that's that's when the KPIs are gonna be important, especially the financial ones that you mentioned. Yeah. And, ultimately, I said this before, I think it's really a maturity model. Right? The the orgs that measure as well as impress are the ones that will be the more more mature, and and I believe that translates into them being more successful. So that that's gonna be the incentive. Right? We're gonna measure because we're wanna gonna we're going to want to be more successful, and that's what will get us there. And we're having to prove the prove that value to these stakeholders that are that are, you know, investing in these in these technologies and ensuring that, like, we're gonna be able to continue to to use them, continue to innovate, that there is some measurable ROI that's coming out the the other side. I mean, Ray Raven, turn to you. What what what are you seeing and and experiencing, for yourself? I definitely think it's a it is a maturity model. It is you know, the if you're not already involved in that model of looking at you know, making that value chain connection of moving up data, this the, you know, the inference, all of these things, like, already you need to be catching up to that quickly because that's how you make it work, and that's how you get to the value is that you invest at the at the foundational level. But then you also take use cases where all of that is sort of, that you can deploy up and down that full scale. I mean, up the value chain. Meaning that, hey. I you know, I know my data in this area is strong. I can go after it. There's customer value here. The. connection between the two is is where I can go for, you know, deploy use case around AI. One of the things that we are doing is is we're focused on a sort of a customer zero type approach. And what that is is, basically, our internal teams, we're all figuring it out at the same time that all of you guys are. and trying to figure out how to roll it out for our organization. And we're vetting out use cases internally. and then taking those to our customers with the value proposition already sort of baked in of, look. Here's what we were able to do. Yeah. Here's why we think we were able to do it. You know? Here's some of the the principles of the data quality and all of that. and bake that into a value proposition for them, basically, by being able say, look. We were able to do this under these conditions with this. We think we can do the same thing, in your in your area. And I think that's gonna be partnerships and especially around, like, with precisely around data quality and around some of those enablers around the scaffolding of this is gonna be important, because we're all trying to solve the same problems. Nobody has zero data problems. Nobody's got zero. security concerns. Yeah. And we're all, like you know, it's like a I mean, this race has been going on for a while, but it's like something happened in November that's like, no. That wasn't the race. We'll start the race again, and now we're gonna run it twice as fast. Yes. So that's where my advice would be is to focus in on partnerships along those ways, digging into the hard stuff because that's what unlocks the value ultimately is you gotta be able to do it trusted. Otherwise, you're not gonna trust the outcomes that you're gonna get from the, you know, genetic capabilities, and it's just gonna be a at at best, you're gonna get an augmentation of certain things, but you're not gonna be able to trust it to do things on its own. So you just you won't be able to get there, and that's that's gonna inhibit the value. I've never, in any of the road races that I've ever run, I haven't seen a new starting line halfway through. It's that's that's that's what it feels like. that's a that's a scary, prop proposition. Dave, thoughts thoughts from you, and then, what then we actually we have a question in the in the chat from from JD that we can get to, but really quick on, on the on the value conversation, from your perspective. Love to love to hear you. Well, I think Raven and Andrew really laid out the landscape pretty well. But what separates organizations that actually realize value, I think it comes down to three things. First, the organizations define success before they start, and it's not in these tech volume metrics like model accuracy or query volume, but it's in business outcomes. It's cost avoided, revenue influence, labor hours reclaimed. Second, I think they resist the temptation to keep things in pilot. Pilots are safe. Production is where value. is created, and that's where data quality, governance, and integration work either holds together or falls apart. But the third, and this goes back to what Raven was just focusing on there, they build an integrated data integrity operating model. This is not a program. It's not a platform, but it's a repeatable operating model that ties together quality, governance, context, observability, skills, business alignment. And so I think those combined together really are where organizations are realizing value, but they're doing it by defining the success criteria, getting it into the real world. Like, I can't tell you how much shifts from when you've been working in the pilot phase where everybody's aligned to what the outcomes are to where you get that out into production and you find somebody that comes at your model from 90 degrees from where you were thinking, and you get to see exactly how did you respond to those questions, or those prompts or those ability to take action off that. And what did you how did you handle the those outcomes? Then third, that operating model. So I'm gonna I'll ask this question from JD before we kind of close out. And if you all do have any other questions that you wanna throw into the chat, we can we can take those. Apologies again for some of the, the the, the challenges we we've had on thus far. But if you do have any other questions, we'd love to to address those here in the chat. But from JD, my question is regarding data inferences, how are you managing these inferences as inferences are the bridge between insights and actions? Accurate insights drive model accuracy. Does inference come from data dictionaries, business rules, etcetera? Anybody have any any thoughts on on JD's question? Yes. It comes from all of the above. And any of your canonical queries, you know, whatever your your accepted canonical view of the world is, oftentimes that is, you know, surfaced up from your data teams. That's where the inferences come from that then become training for the model to give you better outcomes. If you don't, then you're getting inferences based off of whatever was, you know, worked through the model over time. It doesn't understand your context. So it's. it's up to you to bring the context to it. And and I think that's one of the areas where it's just another link in that value chain. But, yeah, it's it's all of the above. Whatever you can put in there that brings value that is, that explains your business context is important. Yeah. And we're I wanna poke back at that context because in in sort of our historical methods back to dashboards for a second, that context is built in by the dashboard design. And so the end users, you know, whether they're engaging with that or picking various different filter elements, that's applying the contact context and has a an embedded semantic model. As we're moving away from that, you know, we need to ensure that the semantic model that the AI agent is using has the the full context of our business. And so that's where things like the catalog, the data dictionaries, the business rules bridge through to agentic is through the lens of that semantic model. Yeah. That's where your that's where a lot of your value of your business resides right. now The thing about large that semantic models, model. of course, is that they're very eager. They're not they're not gonna stay quiet if they think that their dumb answer is gonna embarrass them. They they see it as it's part of the spec to come up with the best answer they can. So there there will be all kinds of unbounded inference if the data is not there. And as Dave mentioned, if the if the context and the semantic model and grounding are right there. And we have to we have to go beyond the semantics that came from the BI era of having a dimensional model. That's very important, by the way. It's just that now we're moving to a superset where it's not just metrics and and categories that we drill the metrics down by. Now it's facts, it's relationships, and it's also things that are, true now in a very narrow temporal window as opposed to things that, you know, have been true over a historical period. So, what's what's good about the dimensional models is that they're a lot better than nothing, and so they've been adding a lot of value. But we we have to be ambitious and go beyond. that. The kind of metadata context and the semantics that you all have been talking about are obviously a major pivotal part of the equation. There's also the location. based context, sort of the the context that represents the physical world in which our businesses actually operate. That's also another important part of being able to ground AI results in in reality and and how they're gonna be applied in our in our day to day lives. So it's another thing I I would sort of impart on on the audience today is, obviously, the metadata context, huge, something that we need to have a major focus on. But there's also the location based context around our businesses that should also be taken into into consideration, That's what's almost counter. So I'm I'm. going Right? We look at semantics as something that have has been true historically, and now it's some semantics are real time. Mhmm. That's that's a really good point. Okay. So really, really appreciate you all joining me on today. So and thank you to our primary participants, Raven, Andrew, Dave, and everyone in the audience who contributed, to the discussion. I really appreciate the engagement in the in the in the chat here today. What we've heard throughout the the conversation is that, you know, while, enthusiasm for for AI, remains strong, organizations are still working through the kind of the practical realities of of data, and AI readiness, governance, data quality, operationalizing these things, and and and being able to measure measure value. So as as AI capabilities kinda continue to evolve, particularly with the autonomous and the agentic systems that we discussed, the integrity, the accessibility, that metadata and location based context that we just discussed, you know, of all of the data that's powering those systems will become even more critical. So I know we've referenced these statistics throughout today's discussion, but if you'd like to explore these insights further, we we certainly encourage you to review the the 2026 state of of data integrity, and AI readiness, report. You'll see that here on the screen right now and and a link down to the the right hand side where you can find that at @ precisely.com/labowreport. In the report, we dive deep into many of the themes that we discussed here in in today's conversation. So thank you all again for joining us. I I really appreciate your time.