Denodo and The AI Trust Gap: The Enterprise Data Crisis Behind AI Adoption

[00:00:04] - [Speaker 0]
What happens when AI agents start making decisions inside live business operations, but the data beneath cannot be trusted? Well, today, I'm delighted to be joined by Terri Dorsey. She's the senior data architect and evangelist for North America at Denodo. And we're gonna discuss the AI trust gap and why data architecture may actually decide which organizations succeed with Agentic AI. And Denodo's new AI trust gap report based on a global survey of over 850 executives reveals a growing disconnect between AI ambition and data resin and data readiness.

[00:00:50] - [Speaker 0]
Because, yes, businesses are investing heavily in AI, but many still struggle with real time access, shared business meaning, provenance, governance, and access control. So today, Toby's gonna talk about why Agentic AI turns old data problems into operational trust problems, and we'll talk about live data, the right data, semantic consistency, explainability, guardrails, and why many enterprises are still trying to build AI on fragmented architectures that are designed for a very different era. And yet, we'll also explore how logical data management and data virtualization can help organizations support AI without creating more silos, duplicated data, and governance headaches. What I'm trying to say is we got a lot to get through. So enough from me.

[00:01:44] - [Speaker 0]
Let me introduce you to my guest now. So a massive warm welcome to the show, Teri. Can you tell everyone listening a little about who you are and what you do?

[00:01:56] - [Speaker 1]
I'm I'm Teri Dorsey. I am a senior day architect and the North America evangelist at Venodo. And and throughout my career, I worked across industries in disciplines such as enterprise architecture, business intelligence, enterprise integration, and operationalizing AI at scale. And over the past decade, I've focused heavily on researching artificial intelligence, both traditional and generative AI. And I recently completed a PhD in data science specializing in cognitive semantic modeling for natural language.

[00:02:29] - [Speaker 1]
And at Denodo, what we do is we help organizations deliver trusted, governed, and real time access to distributed enterprise data through logical data management and semantic abstraction. And what really shapes my perspective is a focus on meaning and abstraction in architecture. Many organizations today are data rich, but meaning poor, in my estimation. And as AI becomes more autonomous, consistent enterprise meaning becomes critical for trust, governance, and operational scale.

[00:03:07] - [Speaker 0]
You have an incredibly cool job title and a cool backstory too. You were working in AI before everyone was talking about AI, before everyone, the hype took off, etcetera. And, of course, this year, it's moved on from generative AI. Everyone's now talking about agentic AI, but many business leaders will still see it as another AI buzzword. So from your perspective here, what actually changes when AI moves from just supporting decisions to actually taking action inside a live business operation?

[00:03:39] - [Speaker 1]
Well, the shift to agentic AI and and particularly autonomous agents is conceptually not new conceptually. Like, so organizations have pursued automation for decades through workflows, rules engines, RPA, and analytics driven decisioning. But what fundamentally is different now is the mechanism of execution. We're moving towards systems that operate through language and semantic interpretation rather than, you know, what used to be, like, narrowly defined, predefined procedural logic. And and that dramatically changes the stakes because language introduces flexibility and contextual reason along with ambiguity.

[00:04:24] - [Speaker 1]
And historically, humans, they humans were able to act as the interpretation layer inside organizations because people could reconcile inconsistencies between systems, and they could compensate for fragmented architectures. The problem is that many of the foundational issues we saw across analytics and reporting, BI, and integration as well were never really solved. They're they weren't fully solved. Organizations accumulated what I call semantic sprawl, and this is duplicated business logic, inconsistent definitions, fragmented governance, and disconnected interpretations of enterprise concepts or core principles. Now when we look at AgenTic AI, it's inheriting these very same foundations.

[00:05:11] - [Speaker 1]
The difference is that instead of generating inconsistent reports, these systems are increasingly expected to autonomously interpret context and then take action. So what was once a reporting problem becomes an operational trust problem. That's why the digital report that we're gonna talk about a little later focuses on this the idea of live data, the right data, and guardrails. That's why that's so important. These are not just technical AI requirements.

[00:05:43] - [Speaker 1]
They're architectural trust requirements and then very critical for enterprise automation at scale.

[00:05:50] - [Speaker 0]
And the report that you alluded to there was the report that I was incredibly excited to get you on here and talk about today because it it's called the AI trust gap, and it highlights a major disconnect between what organizations need for production AI and what they can realistically deliver today. So as someone that is right in the heart of this space, I'd love to know more about what that research reveals that surprised you the most because I suspect you've seen a lot of things over years. And were were there any anything that surprised you in there?

[00:06:21] - [Speaker 1]
I mean, honestly, very little in the report surprised me because it you know, but what it did was it largely confirmed what I have observed over many years, many, many years, and what I consistently hear from enterprise leaders today. You know, what the report did was validate one of my concerns that have existed for years but are now becoming impossible to ignore as AI starts to move towards this autonomous execution and reasoning. You know, for example, the report found that 66% of organizations believe that AI data must be real time or near real time to be trustworthy, to to be trustworthy. And this while organizations are simultaneously pulling and storing data from hundreds or even, some cases, thousands of distributed systems. That reinforces to me something many of us have experienced firsthand.

[00:07:13] - [Speaker 1]
Right? Organizations are trying to operationalize AI on top of architectures built for historical analytics, localized implementation decisions, the fragmentation of ownership, and this duplication of business logic. And the issue is not that enterprises lack the technology or even the talented people. The deeper issue is that businesses business meaning being distributed across applications, pipelines, teams, projects, and rather than managed consistently at the enterprise level. So AI is not creating these foundational problems.

[00:07:50] - [Speaker 1]
It is exposing and amplifying them. And to me, the report confirms that trustworthy agentic AI is ultimately an enterprise architecture and meaning management challenge as much as it is an AI challenge.

[00:08:04] - [Speaker 0]
And the report also talks about three foundations that are required for trustworthy agentic AI. One is live data, two is the right data, and three is guardrails. So let's start with live data. Why is relying on copied or stale data such a a major risk when AI is making those decisions in real time? I'd love to hammer home a few points today that hopefully listeners can take away from, but tell me more about live data and the importance of that.

[00:08:32] - [Speaker 1]
So I'm gonna talk about two reasons why it's important. But first, I wanna establish that there are absolutely legitimate cases for copying and consolidate consolidating data. I mean, historical analytics, trending you know, determining trending, large scale modeling, model training, and certain performance optimization scenarios. These can all benefit from some forms of replication, but the issue is not that copying is wrong. And and and here is so this kind of ties into the first reason.

[00:09:04] - [Speaker 1]
Right? The issue is treating replication as a default architectural approach for operational AI, where many use cases, particularly for operational AI related to generative AI, this whole, know, framework kind of thing. Many of these use cases depend on live continuously changing business context. And so many emerging AI use cases, and frankly, if we think about it, many non AI operational use cases as well require access to data as it exists within live transactional and operational systems. So imagine a customer service chatbot that cannot access you know, any payment issues or active service interruptions in real time.

[00:09:51] - [Speaker 1]
That chat chatbot may still provide some value, but its ability to meaningfully assist the customer becomes significantly limited without live operational awareness. And the second reason or the second reason is is an architectural one. Excuse me. Historically, organizations design systems around anticipated questions and predefined use cases. But AI changes the interaction model because the next question is often unknown in advance.

[00:10:21] - [Speaker 1]
That's where combining live and historical information becomes powerful. Imagine presenting current operational state, historical context together through a single governed semantic object. So the current customer activity combined with prior interactions, trends, preferences, and risk indicators, that enables much richer and more adaptive AI experiences because the system is no longer constrained to only the questions architects anticipated ahead of time. This was a this was a major, challenge, you know, even if we think about, you know, be a business intelligence and and reporting, you know you know, five to ten five to ten years ago. This is one reason I strongly advocate logical architecture approaches.

[00:11:05] - [Speaker 1]
So because instead of creating another physical copy of the enterprise for new use cases, we create governed semantic extraction layer that can dynamically connect live historical and contextual information in real time while maintaining consistent meaning and governance.

[00:11:23] - [Speaker 0]
And further looking at that list of foundations there, the second is, that you emphasize is the importance of the right data, not just more data. So can you tell me a little bit more about why shared business meaning, context, and semantics, all these things are collectively becoming so important, especially as AI moves closer to execution.

[00:11:45] - [Speaker 1]
And and and, Neil, this is probably one of the most underestimated issues in enterprise AI today. So most organizations assume the problem is simply access to data. In reality, the bigger challenge is inconsistent meaning across systems, teams, and operational domains. An AI system can retrieve perfectly accurate data and still make the wrong decision if terms like customer risk, active, and financial exposure mean different things across the enterprise. You know?

[00:12:15] - [Speaker 1]
As I said before, humans can compensate for that ambiguity naturally because they understand organizational context. And AI systems do not without, you know, without some help. Right? That's why semantics are becoming foundational. I often describe, semantics as the future control plane for enterprise AI.

[00:12:37] - [Speaker 1]
Shared meaning allows AI systems, analytics, governance processes, and business operations to interpret data consistently regardless of where the data physically resides. But at the same time, this does not mean that every system or use case must interpret information identically in every situation. Context still matters. The goal and if we also, you know, think of this in line with, you know, the the the the things I brought up for around, you know, ambiguity and and and questions being unknown at that time, The goal is not rigid semantic uniformity either. The goal is controlled semantic flexibility built on shared enterprise foundations.

[00:13:22] - [Speaker 1]
Organizations need stable enterprise concepts, relationships, governance, and lineage while still allowing contextual interpretation and specification at the edge where people are actually consuming or where AI or autonomous agents are consuming. It's it's kinda like a balancing act act. And without shared semantic foundations, every team creates its own isolated interpretation logic and trust fragments across the enterprise. I mean, we've seen it happening over the years. But without semantic flexibility, architectures become too rigid to support evolving operational and AI use cases, which is the other end of what we've seen.

[00:14:02] - [Speaker 1]
So without that balance, agentic AI becomes extremely fragile because the systems cannot reliably distinguish authoritative enterprise truth from conflicting representations of truth.

[00:14:15] - [Speaker 0]
And traditionally, provenance is another area that many leaders have have overlooked over time. So tell me more about understanding where data came from, how it was transformed, and whether it can be trusted. It has also become such a big issue once AI starts acting autonomously.

[00:14:32] - [Speaker 1]
You know, you know, frankly, I mean, so if we think about, you know, the whole we started with the big a with big data and the whole AI analytics explosion and so forth. You know, providence has always mattered, especially in regulated industries where there are compliance, audit, and accountability requirements around how information is sourced, how it's transformed, and how it's used. But it becomes even more critical once AI systems begin participating in, you know, these operational types of workflows. At its core, providence matters whenever someone asks a simple question, you know, why? Why was this recommendation made?

[00:15:09] - [Speaker 1]
Why was this customer flagged? Why was this decision approved or denied? And this is this idea of having data to back it up. Right? And so, you know, the whole thrust or or or or or future or drive around business intelligence and data driven is to be able to do that, to be able to answer those types of questions, and provenance is part of that.

[00:15:32] - [Speaker 1]
The challenge is that many organizations already struggle with Provenance today because over time, they've accumulated enormous sprawl across these analytic environments. The semantic the semantic views, pipelines, extracts, and localized business logics, there's so much out there. And that create these things have created redundancy and consistencies, undocumented dependencies, and much complexity, within the organization. And as a result, tracing lineage and understanding the true source and meaning of information has become increasingly difficult. Even small underlying changes can ripple unpredictably across downstream systems.

[00:16:14] - [Speaker 1]
AI systems inherit those same those same challenges, except now, the decisions and actions may occur autonomously and at and at machine speed, which becomes problematic. And this is, you know, the fear that many organizations have. And one of the reasons I strongly advocate prioritizing enterprise core meaning is that it simplifies a significant portion of provenance challenges. Shared semantic foundations make lineage, governance, and explainability far more coherent across the enterprise. This is also where logical data management becomes extremely important.

[00:16:51] - [Speaker 1]
Rather than creating disconnected copies of localized semantic implementations for every use case, logical data management helps centralize semantic meaning, governance, lineage, and policy enforcement logically across distributed systems while still supporting contextual flexibility. And that balance between shared enterprise meaning and control semantic flexibility is critical for explainability, operational trust, governance, validation, and the the ability to scale AI, long term.

[00:17:28] - [Speaker 0]
And going back to those three foundations for trustworthy agentic AI, we've already spoken about live data, then the right data, and the third one is guardrails, which sounds perfectly like just common sense to me, but many organizations, of course, still bolt governance on after the fact. So why do access controls, explainability and governance? Why do these need to be designed into the architecture from day one? Because they they do seem to be missed, don't they?

[00:17:56] - [Speaker 1]
Oh, and to that's typically because governance after implementation almost always becomes inconsistent governance. Yeah. So most most enterprises evolved through decentralized projects and localized implementation spend decisions. Right? So things built for explicit purposes or specific use.

[00:18:17] - [Speaker 1]
And in many of these cases, if you were to compare them, the security models would be different. Policies would be different. Definitions would be different, and the access patterns are different. And AI would amplify those inconsistencies because agents interact across systems at the speed of machines. When I was working in business intelligence, so we would get requests in for finance, say, you know, or finance for building some reporting around, say, you know, gross sales, and then sales would give us another request around gross sales.

[00:18:50] - [Speaker 1]
And sales security model would be different than finances security model. And if you happen to be have access to both of those, you could literally get different information based off of the the vehicle or the the the the the the mechanism you chose in order to to to to to acquire that information. And and this and the the this becomes, you know, really challenging across organization, and the report highlights this idea that 67% of organizations struggle with AI security and access controls. And if we think about this path of development, it should not be that surprising. It doesn't surprise me because most organizations still govern this way primarily at the project and platform level rather than at the semantic and policy level.

[00:19:38] - [Speaker 1]
And I believe the future architecture for AI governance must centralize meaning, policy interpretation, and access enforcement logically even if the underlying systems remain physically distributed. Otherwise, organizations will spend years building fragmented AI governance overlays on top of already fragmented architectures. So, you know, it it seems obvious. It seems, you know, pretty straightforward. But, you know, in practice, it's sometimes very challenging the way projects are executed and the way organizations are, well, basically organized or sore or or resourced.

[00:20:16] - [Speaker 0]
And you're someone that's worked across business intelligence, health care, manufacturing, utilities, and defense. Of course, if you put all this into a large LLM and and pull out insights from all your experience there, are you seeing certain industries moving faster with AgenTiKi because they understand that operational trust better, or or are the same mistakes happening everywhere? What are you seeing and hearing?

[00:20:41] - [Speaker 1]
Right. So, you know, I, you know, I I more recently, and, you know, after I joined the organization, I've talked to significantly more, practitioners, at at varying levels of of various types of organizations. And the same poor architectural issues exist across virtually every industry. What differs are the consequences of failure. You know, some reputational damage only, some financial loss, regulatory penalties, you know, operational disruption, or even mission failure in some instances.

[00:21:12] - [Speaker 1]
But what I consistently see is that organizations have highly intelligent people working within architectures that evolve through what I mentioned before, these localized projects and applications and departmental priorities rather than through a shared enterprise meeting. And over time, local organizations created global and consistent across data governance and operational processes. I mentioned before humans historically compensate for this, but as AI systems move toward autonomous execution, those same architectural weaknesses become amplified. And that's why AI is forcing organizations to rethink architecture more fundamentally, shifting from application centric thinking towards enterprise capability and meaning centric design. The organizations that succeed will be the ones that establish stable from semantic foundations and reusable operational capabilities that can scale, scale across evolving AI use cases, technologies, and some of the use cases of the future.

[00:22:15] - [Speaker 0]
And Denodo has also long championed logical data management and data virtualization. So how does that approach help organizations better support agentic.ai without creating more silos, duplication, and governance headaches? Because, again, it's a of a balancing act sometimes. Tell me all about how you do that with Denodo.

[00:22:38] - [Speaker 1]
Right. Yeah. So yeah. For sure. And and and, you know, my you know, most of my experience has been in industry in in those very industries that we mentioned earlier.

[00:22:46] - [Speaker 1]
And, you know, I I've kind of seen this, you know, this this this execution of silos and and this duplication within organizations. And one of the most important architectural realizations today is that the enterprise is already distributed and will main distribute it. And even the efforts we've taken to move to cloud or to move our information into lake houses or, you know, if you think thirty years ago, warehouses, The you know, only a portion of that moved and only a lot of times based off of project funding, and and many people are left in the cold to try to find information themselves. Right? So the answer is not and has not been to centralize every piece of data physically.

[00:23:28] - [Speaker 1]
That often creates more duplication, more latency, more governance complexity, and more what I call semantic drift. And frankly, in many, in many cases, unnecessary cost and complexity compared to the actual usage and value of those projects or those things that we build. Logical data management changes the model by creating this government abstraction layer across distributed systems. But what I find powerful and have found in practice myself is that this if that this approach is that it separates business meaning from physical implementation details. And that's incredibly important for AI because it stabilizes meaning even as the systems and platforms and operational environments environments evolve or change out underneath.

[00:24:20] - [Speaker 1]
It empowers a very purposeful management. I mean, you manage purposely managed cost, resources, usage, and risk of how data is actually delivered and consumed. And and that's very powerful in practice as well. It's not just theory. I mean, I often describe this as moving from system centric architecture to meaning centric architecture.

[00:24:43] - [Speaker 1]
That's where enterprise architecture ultimately has to evolve if organizations want to just to to scale trustworthy AI.

[00:24:54] - [Speaker 0]
And we will have many people listening that feel that very real pressure to move fast with AI because, hey, their competitors are doing exactly the same. So how do you help leaders balance that speed and innovation equally with the discipline needed to avoid fragile, ungoverned AI systems? Again, another big balancing act, but tell me more about how Denodo helped get that balance, that that that tricky balance right.

[00:25:19] - [Speaker 1]
Right. Very good. And I like so so I I try to reframe the conversation away from move fast versus govern because that framing, you know, I feel it's misleading to a lot of organizations. Right? So to me, the real question is whether organizations are building reusable architectural foundations, or are they repeatedly creating isolated AI implementations that increase long term complexity, kind of the similar way that many organizations approach, analytics.

[00:25:51] - [Speaker 1]
I remember back in 2017 when I was you know you know, this was before MLOps. It was before, you know, we had many of the new stuff that we have today before LLMs. Right? And and I had to work on a a AI project. And, actually, it was one of the best experiences, you know, of my of my career.

[00:26:11] - [Speaker 1]
But it it wasn't like I had an option to say no. You know? And one of the things that I was very cognizant to keep in mind is this understanding that organizations, how they're structured, is very volatile. And so I can build something that I'm gonna have to keep rebuilding, or I can build something that is going to be stable. And not only can I build it, I can leverage it for other activities within the organization?

[00:26:42] - [Speaker 1]
You know, right now, many organizations are unintentionally accumulating this what I, you know, what I call semantic technical debt. You know, all of this business, this duplicated logic, these inconsistent definitions, they end up costing organizations many, many millions of dollars in some instances with rework, with rewrite, with with resourcing and and support. Right? And when you think about putting AI on top of that, you can just imagine how that debt will compound rapidly, you know, in these types of the Ingentic environment. Suppose we took it and we created the same sprawl with with this Ingentic environments as we have with analytics.

[00:27:23] - [Speaker 1]
So my guidance is usually this, accelerate innovation, but stabilize meaning, build shared semantic foundations, shared governance models, and reusable architectural capabilities that allow AI teams and other teams within the organization as well to move quickly without recreating enterprise inconsistency every time. That's kind of the approach that I took. And when we implemented in 2018, we, you know, quickly after that went through a cloud migration, you know, and, you know, ERP migration and other systems moving in and out, people moving out. But we did not have to refactor what we had done, and we were able to keep enhancing with little to no interruption or disruption to to the to the product that we have built. And I think this speaks to how I think organizations have to start looking about looking about how you actually deliver information and and get get back to some of these architectural principles so that you can be successful at scaling.

[00:28:28] - [Speaker 0]
And I always try and give everybody listening some valuable takeaways. So as we look ahead into the months and years ahead, what should CIOs, CTOs, and data leaders, what should they all be prioritizing right now if they are serious about closing that AI trust gap and begin building systems that really support Agentic AI enterprise scale rather than becoming just another pilot project stuck in pilot purgatory, which we've we've both seen? Any advice there that you'd leave everyone with?

[00:28:56] - [Speaker 1]
I think leaders need to recognize that the long term differentiator in AI is not just gonna be about models or better models. It will be about enterprise architecture and not just the components of enterprise architecture where you're documenting, you know, x, y, z, or whatever. I'm talking about active enterprise architecture where we are envisioning or designing systems that are resilient and stable. You know, right now, many organizations are approaching AI the same way they approach analytics over the last twenty years. I think I said that before.

[00:29:34] - [Speaker 1]
That everybody's solving things independently, developers embedding business logic. And even now, I mean, and actually it's continued through the years, many application developers have the onus you know, I remember, you know, and back in the BI days and application development days, we have the onus of risk and securing things, which probably is not the best place from it for it from an architectural, particularly enterprise architectural perspective. That approach does not scale for agentic AI. It doesn't scale for people, let alone agentic AI. It creates the same sprawl we already see across these these environments that we have within our organizations.

[00:30:15] - [Speaker 1]
So the first priority is establishing enterprise level semantic consistency around core business capabilities, entities, relationships. Second, provenance and explainability should emerge naturally from these shared semantic cement these foundations, governance, lineage, and auditing. All of these things should evolve naturally from the the architectures that we put in place to support those. I'm not through, you know, the the old disconnected after the fact tooling. And at the same time, organizations still need semantic flexibility because context matters.

[00:30:49] - [Speaker 1]
So the goal is controlled semantic flexibility built on shared enterprise foundations. Number three, organizations need logical governance and access models that operate consistently across these distributed environments without forcing this constant duplication of data, policies, and business logic. And finally, architecture must evolve toward enterprise capability based design rather than project based or even tool based implementation. And that shift creates reusable operational foundations for security, monitoring, governance, scalability, supportability, all of the good things, that that fulfill these future extensibility across these evolving AI use cases and technologies. And to support that kind of architecture at enterprise scale, this is where, you know, I I I I truly believe that logical data man, it becomes a foundational capability because it allows these organizations to centralize meaning, governance, lineage, all these things, and and enforce policies logically while still supporting this distribution of systems that, you know, these many live operational access and this contextual semantic flexibility that I spoke of earlier.

[00:32:11] - [Speaker 1]
And the organizations that succeed with AgenTic AI will not simply be the ones that deploy agents the fastest. They'll be the ones that create these stable enterprise meaning and reusable architecture capabilities underneath those agents and also power the enterprise in other areas as well.

[00:32:29] - [Speaker 0]
Wow. I think that is a thought provoking moment to end on, and I can't thank you enough for coming on here today and starting this conversation. And it doesn't have to stop here. For anybody listening that feel inspired, maybe we've set off a few light bulb moments, and they wanna continue that conversation or just dig a little bit deeper on everything we talked about today, connect with you, your team, learn more about Donodo. Anywhere in particular you'd like me to point everyone listening?

[00:32:54] - [Speaker 1]
You can get information on our site at the at, donodo.com.

[00:32:59] - [Speaker 0]
Awesome. Well, I will include a link to donodo.com. I'll also, add a link to your LinkedIn along with the link to the AI trust gap report that we referenced today. Don't anyone listening interested, go check that out. Learn more about it.

[00:33:13] - [Speaker 0]
And feedback to me or Terry, let me know your thoughts on on the problems that you're having or the successes that you're having. Love to hear from you. But more than anything, Terry, thank you for sitting down and starting this wonderful conversation today. So many big takeaways, but thank you for your time.

[00:33:29] - [Speaker 1]
Oh, it was great to be here. Thank you for having me.

[00:33:32] - [Speaker 0]
So a huge thank you to Teri for joining me today and helping unpack why trusted data has become such an important part of that enterprise AI conversation. And Teri's point that AI is not creating many of these data problems, it's merely exposing and amplifying issues that have been building inside organizations for years. So if business logic, definitions, provenance, and governance are already fragmented, then it stands to reason that Agentic AI can turn that that fragmentation into operational risk. That makes data architecture a business issue, especially when AI systems are moving from answering questions to taking action. So for me, the message for every CIO, CTO, and data leader is clear.

[00:34:20] - [Speaker 0]
Moving fast with AI matters, but speed without shared meaning, live data, and governance right from day one could create fragile systems that are difficult to trust or scale. So I'll include links to Denodo, Terry's LinkedIn, that AI trust gap report in the show notes. As always, I wanna hear your thoughts. Is your organization closing that AI trust gap we've talked about, or is data fragmentation still holding back your AI ambitions? And if you pop over to techtalksnetwork.com, you'll find over 4,000 interviews, many different podcasts talking about all of this, and I want you to be part of that conversation.

[00:35:01] - [Speaker 0]
So check those out. Get back to me. Send me an audio message or a DM, and we'll keep this conversation going. But that's it for today. So thank you for listening as always, and I'll speak to you all again very soon.

[00:35:13] - [Speaker 0]
Bye for now.