Latest / The Adults In The AI Room / Earley AI Podcast - Episode 82: Data as the Fourth Pillar: Aligning AI Strategy with Real Business Outcomes

Transcript

Earley AI Podcast - Episode 82: Data as the Fourth Pillar: Aligning AI Strategy with Real Business Outcomes · The Adults In The AI Room

• 0:01
  SPEAKER_00: This is the Early AI Podcast hosted by Seth Early as detailed deep into the passions, expertise, and experiences of thought leaders and practitioners to talk about what's possible with artificial intelligence.
• 0:14
  SPEAKER_00: The Early AI podcast is sponsored by Early Information Science for a digital transformation journey with design and deployment of innovative technology solutions as well as Better.
• 0:26
  SPEAKER_00: Now enjoy the show.
• 0:42
  SPEAKER_01: Today, we are focusing on the role data as the foundation for every successful AI initiative.
• 0:48
  SPEAKER_01: Joining me are two guests who spent their careers helping organizations align data strategy and real business outcomes.
• 0:54
  SPEAKER_01: I'm pleased to welcome Sujet Dutta and Sinnath Raja Kapol, authors of Data as the Fourth Pillar.
• 1:00
  SPEAKER_01: Did I uh reasonably get your names pronounced somewhat correctly?
• 1:06
  SPEAKER_01: Together they bring deep experience guiding leaders through realities of AI adoption from preparing data and governance to scaling AI responsibly in complex environments.
• 1:15
  SPEAKER_01: Sujay, Sid, welcome to the show.
• 1:17
  SPEAKER_01: Thank you.
• 1:18
  SPEAKER_01: Thank you for having us.
• 1:20
  SPEAKER_01: Thank you for having us.
• 1:20
  SPEAKER_02: And uh, where are you guys based again?
• 1:23
  SPEAKER_02: I am in Sweden, Stockholm.
• 1:26
  SPEAKER_02: Okay.
• 1:26
  SPEAKER_03: And I am in the Netherlands, so we are both in Europe.
• 1:29
  SPEAKER_03: Yeah.
• 1:30
  SPEAKER_01: Okay, I'm in Boston and there's a ton of snow outside right now.
• 1:33
  SPEAKER_01: So right in the middle of the winter.
• 1:36
  SPEAKER_01: So let's start off with misconceptions about AI.
• 1:39
  SPEAKER_01: That's always a good place to start.
• 1:41
  SPEAKER_01: What are the things that you see in your world, especially among leaders trying to decide where to invest and how to get started?
• 1:47
  SPEAKER_01: Who'd like to take that first?
• 1:50
  SPEAKER_02: Yeah, if I can, uh I would say AI is being misconstrued as just a technology.
• 1:58
  SPEAKER_02: However, in my view, AI impacts every aspect of your business, the people, the processes that need to be updated and changed, and then the technology, and also last but not the least, are you prepared for the data as well?
• 2:14
  SPEAKER_02: So there's a holistic angle to it, and it's just not a technology play.
• 2:18
  SPEAKER_02: Technology is just one of the factors.
• 2:20
  SPEAKER_02: So do you want to answer that?
• 2:21
  SPEAKER_03: Yeah.
• 2:21
  SPEAKER_03: The first thing I'd say is, you know, uh AI has taken a lot of buzz, especially in the last two, three years, since the advent of uh generative AI.
• 2:30
  SPEAKER_03: So still, I think a lot of business leaders see generative AI or forms of it as we as we also use in our daily lives, things like you know, Chat GPT or Gemini and so on.
• 2:40
  SPEAKER_03: But consider this the entire spectrum of AI.
• 2:43
  SPEAKER_03: But as we all know, I mean the spectrum of AI is really wide, right?
• 2:47
  SPEAKER_03: And every solution, first of all, every business problem doesn't need an AI solution.
• 2:52
  SPEAKER_03: So you need to first, as a board and a C-suite leader, understand which are the key areas where AI can make the difference.
• 2:59
  SPEAKER_03: But a second nuance to that is every AI solution doesn't require a GI solution.
• 3:04
  SPEAKER_03: Gen AI is very good at specific kinds of uh things, as we know, things like you know, generating content in terms of text or audio or speech, which may be very useful for certain jobs, for marketing, for summarization, for generating uh new kinds of content, but may not be very much.
• 3:22
  SPEAKER_03: So, yeah, I think to add to Sujai's point, I feel there are uh a lot of uh confusions on AI.
• 3:28
  SPEAKER_03: First of all, uh every uh business problem does not require an AI solution.
• 3:33
  SPEAKER_03: Because AI may not be the silver bullet for every business problem.
• 3:37
  SPEAKER_03: So it's important for business leaders to understand which problems require AI solutions.
• 3:43
  SPEAKER_03: And within that, if you double-click on that, every AI solution doesn't require a generative AI solution because generative AI is good for certain kinds of use cases, things like generating text or summarization of text, but may not be good for other kinds of uh situations like predicting a model uh for a sales forecast, for example, where a machine learning model might be much more cost effective as well as more accurate, right?
• 4:06
  SPEAKER_03: So it's understanding a bit the nuances when we say AI, what exactly is possible?
• 4:12
  SPEAKER_01: Yeah, yeah.
• 4:14
  SPEAKER_01: Uh and in terms of how they decide where to invest and how they get started.
• 4:20
  SPEAKER_01: I mean, again, lots of organizations are doing multiple pilots and multiple initiatives.
• 4:26
  SPEAKER_01: How do they step back and kind of put a clear evaluation framework in place to look at those efforts and to determine where to focus and prioritize uh their their investments?
• 4:39
  SPEAKER_03: Yeah, so if I may start off, the first thing is to understand what you need to do as an enterprise with AI, right?
• 4:46
  SPEAKER_03: So that's first a really a board and C level decision.
• 4:49
  SPEAKER_03: Once you have a good understanding of what is possible, what is not possible, what you really want to achieve, it's first making sure whether your people, your process, and your data is ready for that.
• 5:01
  SPEAKER_03: Because a lot of enterprises are diving straight into AI and then later finding out, you know, they're not ready either from a people standpoint, meaning the people who use this day in, day out, or the people who build this day and day out do not know how to do this most effectively.
• 5:15
  SPEAKER_03: And then there's that data angle as well, right?
• 5:17
  SPEAKER_03: Whether your data is ready for for AI or not, because especially with the generative AI, they are basically commodity black box models.
• 5:25
  SPEAKER_03: They just spit out next tokens or words in simpler forms, right?
• 5:29
  SPEAKER_03: Unless you provided that right context, and that right context is both data, it's both metadata, it's both information of what your enterprise tone and language is, what it can do, what it cannot do, all that information is first required for your AI model to succeed.
• 5:46
  SPEAKER_03: So for leaders, I would say is first also gauging where you are from a data maturity before looking at your AI, you know, ambitions.
• 5:53
  SPEAKER_02: To lean onto that, I would say, Seth uh said as well, uh, there's a practicality of the matter as well.
• 5:58
  SPEAKER_02: Boards and CEOs are hammered to show outcomes.
• 6:02
  SPEAKER_02: What are they doing in AI?
• 6:04
  SPEAKER_02: So some level of balancing act is also required.
• 6:07
  SPEAKER_02: That's why, Seth, we see a lot of MVPs and trials being done.
• 6:11
  SPEAKER_02: So experimentation will happen, that's the nature of how innovation happens.
• 6:16
  SPEAKER_02: You will experiment, you will learn on the way, you will fall, learn from mistakes, and then the journey is never an upward curve.
• 6:24
  SPEAKER_02: There'll be ups and downs the road.
• 6:26
  SPEAKER_02: So beginning the journey is also important.
• 6:29
  SPEAKER_02: And as you go along the journey, then you bring in that realization.
• 6:33
  SPEAKER_02: But I think it's also first looking back, you understand where you are.
• 6:37
  SPEAKER_02: If you are a legacy company, a large multinational conglomerate, uh, the matter is very different from a startup or a medium-sized company.
• 6:46
  SPEAKER_02: So it's not a one-size-fits-all approach.
• 6:48
  SPEAKER_02: Every company has to decide for themselves how they and also the capital and the resources they have at hand, right?
• 6:55
  SPEAKER_02: That also plays a lot.
• 6:56
  SPEAKER_01: So in your book, you talk about the challenges of getting stuck in pilot mode.
• 7:00
  SPEAKER_01: What patterns do you see in organizations that struggle to move beyond experimentation?
• 7:06
  SPEAKER_01: You just talked about the importance of experiment experimentation.
• 7:08
  SPEAKER_01: Yeah.
• 7:09
  SPEAKER_01: What differentiates those that successfully scale those experiments into the business?
• 7:16
  SPEAKER_02: I would say when companies plan experimentation as part of the journey.
• 7:21
  SPEAKER_02: So in the book, we talked about that maturity journey, and that's where leaders, uh, the their role becomes important because as you are higher up the organization chain, you have that visibility, line of sight into the organization's strategic goals, and then you plan a journey of how AI is going to impact your two-year, three-year, five-year goal, right?
• 7:43
  SPEAKER_02: And so having that journey view is critical, and then you experiment with the view of the journey.
• 7:49
  SPEAKER_02: It's like if you go out for a hiking, you plan your day, this is what I'm going to do, but then you execute on a shorter time period basis what I am going to do, three months, six months time period.
• 8:01
  SPEAKER_02: So that journey, the telescopic view is very critical to have, I would say.
• 8:06
  SPEAKER_01: So having that North Star in mind, but then trying to see how to get there incrementally, learning from your pilots, learning what the success factors are, curating the data and the content, understanding those use cases.
• 8:19
  SPEAKER_01: So those are all very important.
• 8:21
  SPEAKER_01: And um, the clarity of the challenges, the clarity of the initiatives, the clarity of ownership, how do you work with leaders and leadership teams to define the right starting points and really identify those measures that are going to reflect business value?
• 8:38
  SPEAKER_01: Again, we start the organizations start at a higher level with less granularity, and it's more ambiguous.
• 8:45
  SPEAKER_01: You know, if you say you want to, you know, increase revenue or you want to improve customer service or whatever, that's broad and ambiguous, right?
• 8:52
  SPEAKER_01: But that has to be broken down into very specific objectives and specific parts of the process.
• 8:58
  SPEAKER_01: I like to say that you know, you're not going to, you know, replace a person or a role with AI, but you're going to facilitate a step in the process, right?
• 9:09
  SPEAKER_01: You're going to find a place within by understanding those very detailed steps.
• 9:14
  SPEAKER_01: So to me, that's having increasing levels of granularity from those ambiguous goals, those board-level goals, down to the really specific details.
• 9:24
  SPEAKER_01: How do you see organizations accomplishing that and beginning to the one hand, you know, a lot of times executive will wave their arms or check the box or say, yes, we're doing AI and we're using it for customer experience or customer engagement.
• 9:40
  SPEAKER_01: But then they lose track of the nuances of that and they don't, they're not in touch with the details of how to get there.
• 9:47
  SPEAKER_01: And then many times what gets reported up from the front lines is motherhood and apple pie.
• 9:52
  SPEAKER_01: Oh, yeah, everything's great.
• 9:53
  SPEAKER_01: You know, we're moving along.
• 9:55
  SPEAKER_01: I remember years ago when I wrote an article about Verizon and the customer experience, and I had a terrible, terrible, terrible customer experience.
• 10:05
  SPEAKER_01: And there were six major areas of failure.
• 10:08
  SPEAKER_01: And I had six recommendations for improvement.
• 10:12
  SPEAKER_01: But what you saw from the teams, what you saw from PR press releases, you know, the leadership, everybody said everything was great and they were achieving all of these goals and all of these objectives and motherhood and apple pie.
• 10:26
  SPEAKER_01: Yet I, as a customer, had just the worst possible customer experience.
• 10:31
  SPEAKER_01: I'm actually going through another situation right now with them where they can't find an order.
• 10:35
  SPEAKER_01: Last time it was a two-for-one deal that they never honored, but despite months and months of trying to get them to do that.
• 10:41
  SPEAKER_01: Now we just ordered another phone with another similar deal, and now they can't find the order.
• 10:46
  SPEAKER_01: They can't find the phone.
• 10:47
  SPEAKER_01: And it's like, okay, so how well is your digital transformation going when the customer on the ground doesn't experience it?
• 10:53
  SPEAKER_01: Now, that's not to meant to beat up on Verizon, but uh that's just my personal customer experience.
• 10:59
  SPEAKER_01: But I think the same thing happens with lots of organizations.
• 11:02
  SPEAKER_01: So, how do you go from the motherhood and apple pie and arm waving and the board level checking the box to the real uh detail?
• 11:10
  SPEAKER_03: Yeah, I think you made some very good points, and you're right, it's not very only Verizon, but we see this in many large enterprises having this problem.
• 11:17
  SPEAKER_03: So, if we if I can break this down into a couple of patterns, I'm seeing, uh, and then Sujet, please add in.
• 11:23
  SPEAKER_03: One pattern I see is looking at AI as some kind of a separate thing, a little bit what Suj alluded to, rather than embedding it as part of your entire business breakdown and process.
• 11:34
  SPEAKER_03: Typically, large enterprises already have an operating model for your people, for your processes, for your technology, right?
• 11:40
  SPEAKER_03: So when you have a big business ambition, for example, you're gonna open up a new plant, let's say in Mexico, just to make it up.
• 11:47
  SPEAKER_03: You already have a process of how I'm going to set up my people structure.
• 11:51
  SPEAKER_03: What are the processes based on the specific laws in Mexico that I'm gonna follow?
• 11:56
  SPEAKER_03: What kind of technologies I will need to run this plant?
• 11:59
  SPEAKER_03: Why not embed the same extension of these processes for AI as well as for data as well, to add on to things?
• 12:06
  SPEAKER_03: I think that is a good way to start then, rather than seeing, hey, let me just hand it over to either my IT organization to figure this AI piece out, and then they will figure it out, and there'll be some magic results, as you just said, right?
• 12:18
  SPEAKER_03: So that's kind of one pattern we see.
• 12:21
  SPEAKER_01: The second pattern we see, especially when your data estate is not beginning with whatever the initiative is, and weaving into that initiative the capabilities around AI, identifying the use cases that will require better access to information, faster access to information, new information, regardless of what the mechanism is.
• 12:42
  SPEAKER_01: It could be RAG, it could be retrieval documented generation, it could be uh generative AI summarization, could be it could be machine learning and predictive analytics, it could be a completely non-AI related solution, but it's really just looking at the core processes and what the initiatives are, and then determining what the right tools would be in order to apply to the situation.
• 13:05
  SPEAKER_02: Yep.
• 13:06
  SPEAKER_02: And Seth, if we take take a deep dive into the classic problem that you mentioned, and a deep analysis will mostly reflect that there was a breakdown of data silos existed.
• 13:17
  SPEAKER_02: The system that took your order and two for one, actually that data never perclauted into actually the execution system.
• 13:26
  SPEAKER_02: So, and one of the things that you said, how do you bridge the gap between what boards and CXOs are deciding to get into operational?
• 13:32
  SPEAKER_02: Maybe data plays an important role as well, capturing those and translating those.
• 13:37
  SPEAKER_02: So I would say it is a data problem, and then AI could actually come in.
• 13:41
  SPEAKER_02: It's not only data for AI, but AI for data as well.
• 13:44
  SPEAKER_02: So, how you embed these two into every aspect of your business operations from your supply chain operations to your customer relationship management to your order execution is where it needs to be weaved in both data and AI and not treat it as a silo.
• 14:01
  SPEAKER_02: I think that would be one of the ways for me as I see and talking to exigs that we experience.
• 14:07
  SPEAKER_01: Sid, I had stopped to summarize your point.
• 14:09
  SPEAKER_01: Did you have some other points you wanted to make in that same context when you were using that example of standing up the new manufacturing plant?
• 14:17
  SPEAKER_03: The only other example I wanted to add was actually uh from our book, because in the book we used a case study from Audi.
• 14:24
  SPEAKER_03: So, Audi, as you know, is they're a big automotive enterprise based in Germany.
• 14:28
  SPEAKER_03: And their CDO made a very interesting point when they were doing a bit of analysis of where they are with their digital transformation.
• 14:34
  SPEAKER_03: They said, hey, uh, when we were analyzing the level of intensity of how good some of the challenges are, or how what is more challenging and what is less challenging, they found out, for example, that business pressure is not a big challenge versus maybe some other aspects like Suja mentioned, data silo was a big big challenge.
• 14:53
  SPEAKER_03: But when they double-clicked on this, why is business execution not a big challenge?
• 14:58
  SPEAKER_03: The assessment was that that's because not enough demand is coming.
• 15:02
  SPEAKER_03: So the business team were building something on their own and the IT team were going through a digital transformation on their own.
• 15:08
  SPEAKER_03: So sometimes it's not just what numbers are being presented to you as a leader or a C-suite, but also to understand why that number is like this.
• 15:17
  SPEAKER_03: For example, in this case, if the uh CDO just decided, hey, that number is low from a challenge perspective, let me just ignore it, then they would not have found out that there's not enough business demand coming in the first place.
• 15:28
  SPEAKER_03: So they could they didn't rectify the problem uh as a result.
• 15:32
  SPEAKER_03: So in the case of Audi, they luckily found this up.
• 15:35
  SPEAKER_01: So help me understand that more.
• 15:36
  SPEAKER_01: When you say business challenge, business execution was not a challenge, but then you found there was no demand, meaning market demand for the manufactured for the cars, or are you talking about business demand for an IT solution or solution?
• 15:52
  SPEAKER_01: Help me understand what that meant.
• 15:54
  SPEAKER_03: Yeah, it's a good point.
• 15:55
  SPEAKER_03: I should have clarified.
• 15:56
  SPEAKER_03: It means business demand for data and AI in the organization, right?
• 15:59
  SPEAKER_03: In this case, it's Audi uh logistics and manufacturing, which is a particular domain within Audi.
• 16:06
  SPEAKER_01: They didn't see that they had a problem that needed to be addressed with this approach.
• 16:10
  SPEAKER_03: Exactly.
• 16:11
  SPEAKER_01: And but they did have problems that need to be addressed.
• 16:13
  SPEAKER_03: So where was the where was the disconnect in terms of the disconnect between business and IT or sometimes within the different business units that we see, right?
• 16:22
  SPEAKER_03: And and this disconnect can happen due to people not connecting, this disconnect can happen due to data, uh, you know, as to Sujet's point, whether different systems and applications are not connected.
• 16:34
  SPEAKER_03: And it may it may work, it may have worked in the past because in traditionally a lot of these industries work at a very business domain level.
• 16:42
  SPEAKER_03: But to your original question, Seth, if companies need to really reimagine themselves with data and make that really next impact, this kind of disconnect needs to be first fixed.
• 16:51
  SPEAKER_01: So again, was it a challenge around the lack of understanding of how or where AI could help?
• 16:58
  SPEAKER_01: Because it doesn't, because it because obviously the business is always going to have some need around data and functionality and capability, right?
• 17:09
  SPEAKER_01: There's always going to be, you know, it's a it's a never-ending cycle, right?
• 17:12
  SPEAKER_01: They don't have perfect data, they don't have perfect forecasting, they don't have perfect processes, they have you know disconnected systems.
• 17:20
  SPEAKER_01: So what was the lack of demand for the business based on?
• 17:25
  SPEAKER_01: Was it that that they did not understand?
• 17:28
  SPEAKER_01: I'm just trying to deconstruct this to say what is the fix or what is the remedy.
• 17:33
  SPEAKER_03: Great question.
• 17:34
  SPEAKER_03: Uh let's take a specific example so it becomes clearer to everyone.
• 17:38
  SPEAKER_03: So let's take the example uh and and that will help then summarize this.
• 17:43
  SPEAKER_03: One of the examples was if you look at the shop floor of Audi, they have uh a lot of welds being done on the shop floor, right?
• 17:49
  SPEAKER_03: So you needed to weld the cars together and and you know the different parts of the cars need to be welded together.
• 17:55
  SPEAKER_03: If something goes wrong in the weld, it's maybe found out at a later stage in the shop floor, maybe like a quality uh control cycle.
• 18:02
  SPEAKER_03: But then you need to bring it all the way back in the shop floor for weld, right?
• 18:06
  SPEAKER_03: That requires a lot of cost as a result to be leaked.
• 18:09
  SPEAKER_03: Clear business problem, right?
• 18:10
  SPEAKER_03: You need to find out there's a weld problem or not.
• 18:12
  SPEAKER_03: Initially, they said, okay, let's look at AI as a solution, and they started looking at taking pictures of which are the welds and then started training a model.
• 18:21
  SPEAKER_03: Hey, this is a good weld or a bad weld, categorizing and classifying these welds.
• 18:26
  SPEAKER_03: And hopefully, when a new weld comes in, you take a picture immediately and it will have a certain level of accuracy, right?
• 18:32
  SPEAKER_03: The data leader came in and saw that this is a use case, but then B, they also found out that picture is probably one input.
• 18:40
  SPEAKER_03: Temperature might be another data point to consider because, in addition to a visual image, you might require uh a temperature or some kind of uh input to additionally better improve the outcome of the model and thereby increase the uh effectiveness of this, right?
• 18:56
  SPEAKER_03: This required basically both business understanding, but also a data and AI leader coming in and saying these would be additional parameters that would help improve, right?
• 19:05
  SPEAKER_03: This is what I feel is missing a bit in especially large enterprises where people still operate in silos and they all feel I they know their business problems, but when they come together, they might find a larger solution to the problem which may increase the outcome X.
• 19:22
  SPEAKER_01: So, and that's requires you know business technology and engineering know-how, because the engineers are the ones that are setting up the welding stations and you know, either automating that or having some manual process.
• 19:36
  SPEAKER_01: So they are the ones that need to say, by the way, another parameter is the temperature, you know, of the well to determine if it's going to be or what are the factors involved.
• 19:44
  SPEAKER_01: But again, that seems to me that it's there's there's an understanding of the nature of the problem, but is it simply saying that there's no understanding of the nature of that type of solution?
• 19:58
  SPEAKER_01: So it's a matter of they don't understand they don't know what they don't know what how they could solve that particular problem.
• 20:04
  SPEAKER_01: So that's why you're saying there's less business demand for that.
• 20:08
  SPEAKER_01: Right.
• 20:09
  SPEAKER_01: Exactly that.
• 20:10
  SPEAKER_01: So the answer to that then is education around what the possibilities are, and then ideation when you're looking at problems and bringing in a cross uh functional, cross-disciplinary team to do the analysis to say, okay, can we start solving this problem or improving this process, right?
• 20:29
  SPEAKER_01: If we have these recurring, so there's probably, you know, always hundreds and hundreds of problems in a manufacturing plant.
• 20:36
  SPEAKER_01: And the question is, when do you bring a team like that together to try to solve those?
• 20:41
  SPEAKER_01: And sometimes you don't know what the cost implication is, or or it's a matter of prioritization, right?
• 20:48
  SPEAKER_01: In other words, do we assemble this team to address this problem, or do we assemble the team to address that problem?
• 20:54
  SPEAKER_01: So in that case, it's a matter of collecting all of those problems and then applying some prioritization process to that to say what are the different dimensions.
• 21:05
  SPEAKER_01: There's cost, there's complexity, there's timeline, there's scope, there's data sources, there's all sorts of different constraints.
• 21:12
  SPEAKER_01: And so it seems to me that part of this is understanding what those processes are for doing that prioritization and understanding what foundational elements need to be in place to do that.
• 21:24
  SPEAKER_01: So when you look at teams that are focused on tools and models, the technical teams, a lot of that work, as we're discussing right now happens around the model, right?
• 21:35
  SPEAKER_01: The workflow integration, the compliance, you know, the data access, the process understanding, right?
• 21:44
  SPEAKER_01: Are the foundational elements that have to be in place before an organization can deploy with AI with confidence, right?
• 21:51
  SPEAKER_01: There's many different dimensions across knowledge, data, governance, and decision making, technical infrastructure, operational readiness.
• 22:00
  SPEAKER_01: So, what are the things that you see need to be in place as a foundation before organizations can effectively pursue these initiatives?
• 22:10
  SPEAKER_02: I think if I may take it, many of the things that you spoke about, all the AI technology related to the model, it is all the supply side of things, right?
• 22:19
  SPEAKER_02: How you're able to supply those capabilities.
• 22:21
  SPEAKER_02: The classic, and then you talked about the business problems, right?
• 22:25
  SPEAKER_02: So the classic gap we see in organization is this gap between demand and supply.
• 22:30
  SPEAKER_02: And like economics 101, you cannot be successful till this demand and the supply meet each other.
• 22:36
  SPEAKER_02: So that's the key part where we formulated the maturity model framework in our book.
• 22:43
  SPEAKER_02: How do you merge and bring the supply and demand together?
• 22:47
  SPEAKER_02: And that's how you chart your maturity journey.
• 22:50
  SPEAKER_02: So you must one side document all the business problems and the challenges.
• 22:54
  SPEAKER_02: These could be formulated as your use cases, the demand.
• 22:57
  SPEAKER_02: And the demand.
• 22:58
  SPEAKER_02: That's the demand.
• 22:59
  SPEAKER_02: Yeah.
• 22:59
  SPEAKER_02: That's the demand and AI.
• 23:01
  SPEAKER_02: And then you are working on the other side all the supply capabilities like AI technology, model training, evaluation of the model, all building up the data platform.
• 23:11
  SPEAKER_02: These are all the supply things, right?
• 23:14
  SPEAKER_02: In a classic organization, the supply and demand are very much far separated with each other.
• 23:19
  SPEAKER_02: And that's one of the reasons, like in that Audit case, like Sid mentioned, the supply teams are not having that visibility of the demand, and hence they have a feeling or a perception that the business challenge is low.
• 23:32
  SPEAKER_02: Whereas the business teams actually then who have those use cases and demands do not have visibility of what's happening on the supply side, and then they try to figure solutions on their own, or they feel this IT team or the AI team is slow to respond to their.
• 23:49
  SPEAKER_02: So this is the classic gap, and that needs to be merged through a maturity framework that we have formulated in our book, wherein the supply and demand must come together through the capability building, and you need to formulate that way.
• 24:03
  SPEAKER_02: And that's why we call data as the fourth pillar.
• 24:06
  SPEAKER_01: So it's interesting.
• 24:08
  SPEAKER_01: You know, when you say supply and demand, I have to go through additional mental gymnastics in order to convert that into something that I understand about the business because I don't ever think about supply and demand in terms of internal processes on the business side and the requirements and the ability to apply those.
• 24:30
  SPEAKER_01: So it's a different model.
• 24:31
  SPEAKER_01: And I think I would imagine that some executives may have the same challenge.
• 24:36
  SPEAKER_01: Maybe you could shift and talk a little bit about the elements to your maturity model.
• 24:40
  SPEAKER_01: What are those elements?
• 24:41
  SPEAKER_01: And then what are the stages of maturity?
• 24:43
  SPEAKER_01: Yes, sir.
• 24:44
  SPEAKER_01: Do you want to take that?
• 24:46
  SPEAKER_03: Sure.
• 24:47
  SPEAKER_03: So uh basically let's start with the stages of the maturity, and then I think the elements will come together.
• 24:52
  SPEAKER_03: So we've formulated three key stages of maturity.
• 24:55
  SPEAKER_03: So the first stage is more called the uh foundational stage.
• 25:00
  SPEAKER_03: This is where uh, you know, similar to the Verizon example or the Audi example, like a couple of years ago, which I mentioned today.
• 25:08
  SPEAKER_03: These are enterprises where the business and the IT teams operate in silo and they might be delivering data and AI value on pockets, but not really delivering it at a holistic level.
• 25:21
  SPEAKER_03: And the reason for that is perhaps from a supply side to Sujet's point, maybe they have some capabilities, but not all.
• 25:27
  SPEAKER_03: For example, a lot of enterprises today have you know good reporting solutions or they may have good data platforms, whether they build it or buy it, but they may not have maybe things like an ontology layer, right?
• 25:39
  SPEAKER_03: Uh a layer which allows them to understand different dimensions, what exactly the data means from different parts of the business, right?
• 25:47
  SPEAKER_03: So these are kind of capabilities which are missing from a supply standpoint.
• 25:50
  SPEAKER_02: One is like unstructured data that we have covered.
• 25:53
  SPEAKER_02: That's another capability.
• 25:55
  SPEAKER_03: That's another one which would be very important for AI.
• 25:57
  SPEAKER_03: Yeah.
• 25:58
  SPEAKER_03: And then from a demand standpoint, I mean, demand is not coming in from a harmonized fashion.
• 26:02
  SPEAKER_03: Perhaps some business units, to Sujet's point, they might be trying to figure solutions out on their own, right?
• 26:08
  SPEAKER_03: We have seen so many examples of enterprises having something we call as shadow IT, where they're trying to solve this on their own rather than coming towards the central IT functions, for example.
• 26:20
  SPEAKER_03: So these are kind of examples of the foundational stage of the maturity where So give me the, I'm sorry, so give me the three stages of maturity.
• 26:27
  SPEAKER_01: What are those three stages?
• 26:29
  SPEAKER_02: The fundamental stage, the scale stage, and the automated stage.
• 26:32
  SPEAKER_01: Okay.
• 26:33
  SPEAKER_01: And then what are some of the dimensions of those stages?
• 26:37
  SPEAKER_01: So what would be some of the elements?
• 26:39
  SPEAKER_01: So you mentioned the data silos versus data uh versus integrated.
• 26:44
  SPEAKER_01: What are the different elements of that maturity model?
• 26:47
  SPEAKER_01: So, you know, for instance, we might have governance is one dimension, knowledge readiness is another dimension, information architecture is another dimension, whatever it might be.
• 26:57
  SPEAKER_01: So, you know, we have different models for knowledge, we have models, maturity models for content, for product data, for customer experience, for orchestration, and then each one has multiple dimensions.
• 27:10
  SPEAKER_01: So, what are some of the other dimensions of your maturity model?
• 27:13
  SPEAKER_03: So, from a demand perspective, we have three dimensions, which is which we classify as the data intensity.
• 27:20
  SPEAKER_03: So these three dimensions are quality, compliance, and speed.
• 27:24
  SPEAKER_03: So that is something we've abbreviated to QCS.
• 27:27
  SPEAKER_03: Quality is what is the amount of quality required for your use case?
• 27:33
  SPEAKER_03: Compliance is how compliant it needs to be, whether it's internal or external compliance.
• 27:38
  SPEAKER_03: And speed is how fast does the data need to be or the AI needs to work, right?
• 27:43
  SPEAKER_03: For example, uh fraud detection may require real time versus maybe a marketing campaign may require data, which may be even stale by a few days, right?
• 27:52
  SPEAKER_03: So these are the three dimensions we are looking at from a demand standpoint to sujez uh demand point.
• 27:57
  SPEAKER_01: So the stages are fundamental silo automated, and then the dimension would be one dimension would be when you say demand, you're saying there's quality compliance compliance and speed.
• 28:09
  SPEAKER_01: So do you have a fundamental maturity stage around quality, a scaled maturity stage around quality, and an automated maturity stage around quality?
• 28:19
  SPEAKER_01: I'm just trying to understand how these fit together.
• 28:21
  SPEAKER_02: Yeah, Seth, if you may visualize, yeah, visualize the two by two, the y-axis is demand and x-axis is supply.
• 28:29
  SPEAKER_02: So you as an organization now then need to find your positioning with respect to the x-axis and the y-axis.
• 28:37
  SPEAKER_02: So when you come to the x-axis, is where all your data and AI capabilities come in, like you talked about governance, you're all like uh you also wrote your book, and Sid talked about the ontology layer, data ingestion, all your AI model capabilities.
• 28:52
  SPEAKER_02: So that's an assessment you need to make as to where do I stand on the supply layer, right?
• 28:58
  SPEAKER_02: And there could be we have proposed a calculation model in the book, but that's for later, that's going into too much detail.
• 29:04
  SPEAKER_02: And in the y-axis, then you figure out where do I stand on the y-axis in terms of the demand.
• 29:10
  SPEAKER_02: We have formulated two ways to figure that out.
• 29:13
  SPEAKER_02: One is based on this QCS framework, like if you take Q's quality, compliance, and speed, if you give the highest to like say 10 out of 10 for each of these three dimensions, and then you go through your each of your use cases, you figure out what is the QCS requirement in aggregate, or we have also formulated three metrics called TAV, EAV, and RV that is the total addressable value, expected addressable value, and realized value.
• 29:42
  SPEAKER_02: So on these parameters in the book, we have gone into the details how you find your positioning in the y-axis.
• 29:49
  SPEAKER_02: So these calculations of x-axis and y-axis will give you a point in this 2x2 matrix where you then identify which stage of maturity we are.
• 30:00
  SPEAKER_02: Normally, ballpark, you may guess, lower bottom left quadrant and y quadrant, and also moving upward is where the fundamental stage are.
• 30:09
  SPEAKER_02: And as you move towards the rightmost corner is where you mature from the fundamental stage to the scale stage to the automated stage.
• 30:18
  SPEAKER_02: Automated stage is where you have most use cases with high on QCS demand, and then you have maximized your capabilities with all AI tooling that you are able to deploy all the AI agents and everything to run and truly autonomous enterprise.
• 30:35
  SPEAKER_02: And scale stage is where you have enabled self-service of data and AI for all those business users who are able to leverage those capabilities.
• 30:46
  SPEAKER_02: So in a ballpark framework.
• 30:48
  SPEAKER_02: But this is actually a very detailed topic.
• 30:50
  SPEAKER_02: Hope I can simplify it.
• 30:51
  SPEAKER_01: So I'm just yeah, I'm just trying to understand like how we measure one of the dimensions according to your maturity model.
• 30:59
  SPEAKER_01: So fundamental scaled automated.
• 31:01
  SPEAKER_01: And if we look at quality, how do we measure quality according to those stages?
• 31:06
  SPEAKER_03: So this is a little bit where uh it so now let's let's take a practical example.
• 31:10
  SPEAKER_03: Let's say I'm an enterprise and I'm embarking this maturity model.
• 31:14
  SPEAKER_03: So from a demand perspective, I need to first look at all the use cases and programs of work which are coming from a display.
• 31:23
  SPEAKER_03: So different departments may have different use cases coming with demands.
• 31:27
  SPEAKER_03: Then we sit with these similar to how you create.
• 31:30
  SPEAKER_03: You just sit and determine what is the, let's take the quality example, what is the quality requirement for each of these use cases, and that will determine the score relative to each other's.
• 31:40
  SPEAKER_03: Another important uh you know principle here is we're not comparing one organization with another.
• 31:46
  SPEAKER_03: It's all about how you relatively scale yourself.
• 31:50
  SPEAKER_03: So, for example, if my 10 is the most quality intensive and what one is the least, you relatively place themselves.
• 31:57
  SPEAKER_01: So again, just trying to break down like one particular area.
• 32:02
  SPEAKER_01: So you mentioned you have a use case and then you have a data quality requirement for the use case, correct?
• 32:10
  SPEAKER_02: Requirement compliance requirement.
• 32:12
  SPEAKER_01: So, what's an example of a data quality requirement for a use case?
• 32:18
  SPEAKER_01: So, how would how do you measure that?
• 32:20
  SPEAKER_01: Are you measuring it on an aggregate score across uh a repository?
• 32:25
  SPEAKER_01: Or are you I'm just trying to get a sense of how because quality is always a challenge, right?
• 32:29
  SPEAKER_01: You can say data completeness, you can say, you know, data fill, the data uh model, right?
• 32:35
  SPEAKER_01: This the structure.
• 32:36
  SPEAKER_01: You it doesn't have the right structure.
• 32:39
  SPEAKER_01: Maybe the data is good and clean and up to date, but it doesn't have the right structure.
• 32:43
  SPEAKER_01: Or maybe it is has the right structure, but it doesn't have the right um um metadata tags, right?
• 32:51
  SPEAKER_01: So I might have a content model, it's a good content model, satisfies a use case, but the content is not tagged, so it's not curated.
• 32:59
  SPEAKER_01: Or I can have you know version controls and I could have uh that are versions that are not controlled, and I can have multiple iterations of a piece of content, and the model can't tell which one is the right one.
• 33:12
  SPEAKER_01: So then there's fitness to purpose.
• 33:15
  SPEAKER_01: Data quality can be measured according to lots of different dimensions.
• 33:19
  SPEAKER_01: So can you give us an example of a specific use case and a specific data quality score so that we can kind of understand how that fits in with your uh sure.
• 33:30
  SPEAKER_03: Let's do that.
• 33:31
  SPEAKER_03: So I mean, uh again, the this is where we are double clicking a bit a little bit.
• 33:36
  SPEAKER_03: So uh the book might be more helpful to go into the details, but just to paint the picture, let's take the six normal dimensions of data quality, right?
• 33:43
  SPEAKER_03: So, as you as you alluded to, Seth, you know, you have typically six uh dimensions of data quality, accuracy, completeness, uh, you know, uh, etc.
• 33:51
  SPEAKER_01: So you can't so just so we're on the same page.
• 33:55
  SPEAKER_01: Accuracy, completeness.
• 33:57
  SPEAKER_03: Uh let me let me think through it.
• 33:59
  SPEAKER_03: I might be missing something.
• 34:00
  SPEAKER_03: Uh timeliness.
• 34:01
  SPEAKER_03: Timeliness is uh definitely one.
• 34:03
  SPEAKER_03: So these are the standard uh frameworks uh that are present already today.
• 34:07
  SPEAKER_03: We are not reinventing the video, so it's accuracy, completeness, consistency, timeliness, validity, and uniqueness.
• 34:13
  SPEAKER_03: So these are the the six agreed dimensions of this.
• 34:17
  SPEAKER_03: Again, we are not being very stringent.
• 34:19
  SPEAKER_03: Every enterprise, again, with AI, there are more requirements also coming from data quality.
• 34:24
  SPEAKER_03: So every enterprise may determine what are these sub-dimensions.
• 34:28
  SPEAKER_02: And to add to that set, it is the it is the demand of the use case.
• 34:32
  SPEAKER_02: So if if a fraud detection engine needs what kind of completeness, what kind of a timeliness, and each of the dimensions.
• 34:39
  SPEAKER_02: It's not what you're supplying, it is what you require.
• 34:42
  SPEAKER_02: So for like, for example, a marketing campaign, the email ID must be accurate.
• 34:47
  SPEAKER_02: So that's a basic requirement of it.
• 34:50
  SPEAKER_02: So so it comes from the requirement of the use case of what is actually demanded by the use case, not what you are supplying.
• 34:58
  SPEAKER_03: Yeah, and maybe for that to extend to Sujet's point, the timeliness of that marketing campaign, they might be okay with uh with a delay of a week as well.
• 35:06
  SPEAKER_03: So that may determine as a lower rank, maybe related to another use case where, as I said before, there might be a fraud detection uh use case where there's a there's a requirement of this uh data to be real time in terms of timeliness.
• 35:20
  SPEAKER_01: Sure.
• 35:20
  SPEAKER_01: So then applying the maturity model, there's fundamentals scaled and automated.
• 35:25
  SPEAKER_01: How does how do we look at quality measures according to those maturity dimensions?
• 35:33
  SPEAKER_01: Or is it or is it not that much of a core correlation?
• 35:36
  SPEAKER_01: I'm just trying to understand how to fit some of these elements together with an example.
• 35:41
  SPEAKER_03: I think the way you can look at it is once now you've measured this from a demand standpoint, like we took the quality example, we will find out where you are in the axis, and that will determine your position in the fundamental scaled and automated stage.
• 35:55
  SPEAKER_03: Essentially, the higher you are able to meet the demand with the supply is when you will be much more towards the top right of this.
• 36:03
  SPEAKER_03: Imagine, like a Gartner quadrant, for instance.
• 36:06
  SPEAKER_03: So that will be like your positioning, right?
• 36:08
  SPEAKER_03: So the lower you are means you're not either delivering in terms of demand or supply.
• 36:13
  SPEAKER_03: So, for example, let's take a couple of enterprises.
• 36:16
  SPEAKER_03: An enterprise might have all the capabilities in the world.
• 36:19
  SPEAKER_03: Imagine every technology capability they need.
• 36:21
  SPEAKER_03: Let's say they have invested a billion dollars just for the sake of it too, and they've got every capability you can think of.
• 36:27
  SPEAKER_03: So maybe from a supply standpoint, and also they might have a huge team and they might have gone completely agentic and they've invested like this billion dollars to do this, right?
• 36:36
  SPEAKER_03: Maybe then from a supply standpoint, they are completely on the on the top of this.
• 36:41
  SPEAKER_03: So in the supply is kind of like the x-axis I said before.
• 36:44
  SPEAKER_03: But maybe there's no business using this at all.
• 36:47
  SPEAKER_03: Then from a demand standpoint, they'll be low, which eventually brings them overall that demand is low, but supply is high, so you're still at the fundamental stage.
• 36:55
  SPEAKER_03: So both of these really needs to come at the high point for it to be at the automated stage and the scale stage.
• 37:02
  SPEAKER_01: Okay, that's helpful.
• 37:03
  SPEAKER_01: Thank you for that.
• 37:04
  SPEAKER_01: Let's kind of move to a um, I'm I think there's still some gaps in my knowledge, but um, I'd have to go back to the book.
• 37:10
  SPEAKER_01: I don't know if I have the yeah, hence it can be.
• 37:18
  SPEAKER_01: So if if you ever want to send me a uh physical version, that would be great.
• 37:21
  SPEAKER_01: But um so what about the human dimension, skills, roles, change management?
• 37:28
  SPEAKER_01: You know, when leaders are feeling overwhelmed about preparing their teams, what do you offer?
• 37:35
  SPEAKER_01: What guidance would you have?
• 37:36
  SPEAKER_03: So I think the first thing here is like we said before, every enterprise broadly uh has some kind of an operating model today.
• 37:43
  SPEAKER_03: So they either operate in a centralized manner where these kind of functions for data and AI, but also for IT, might be centralized.
• 37:51
  SPEAKER_03: There are several enterprises where it's completely decentralized.
• 37:54
  SPEAKER_03: For example, if enterprises are growing by MA, for example, and they might operate completely decentralized, there might be legal entities or there might be federated operating models, right?
• 38:04
  SPEAKER_03: Where there's parts which are centralized and parts which are decentralized.
• 38:07
  SPEAKER_03: So first identifying what your enterprise structure is and extending that to the data and AI function is important.
• 38:15
  SPEAKER_03: We see several situations where it's the opposite, where you know, maybe for example, uh an IT function or a data and AI function is centralized, but entirely business is decentralized.
• 38:24
  SPEAKER_03: So it becomes very difficult to be able to map into that.
• 38:28
  SPEAKER_03: So that would be the first kind of step.
• 38:30
  SPEAKER_03: I don't know, Sujai, if you want to add something to it.
• 38:33
  SPEAKER_02: No, I think it goes, it uh one of the important parts is that you have to also follow your organization structure and model, how do you operate, right?
• 38:41
  SPEAKER_02: So you cannot create a data and AI structure that is totally disjointed and not connected to your ways of working.
• 38:48
  SPEAKER_02: That's a true recipe for failure.
• 38:51
  SPEAKER_02: So you need to first understand uh even the culture of the organization, what is the incentive model.
• 38:56
  SPEAKER_02: So there's a lot of study that needs to be done to finally get to the solutioning, right?
• 39:01
  SPEAKER_02: So that discovery of the problem statement and how you execute is truly important, I would say.
• 39:07
  SPEAKER_03: And then to add on to that, just one more point, Seth, is you know, then we need to also look from a data and an AI leadership standpoint on maturity of each of these business functions.
• 39:17
  SPEAKER_03: So let's say we are operating in a federated manner.
• 39:20
  SPEAKER_03: There might be certain domains which require hand holding from the IT standpoint, where you might require a lot more resources, right?
• 39:29
  SPEAKER_03: A lot more people to be staffed to help that, versus there might be certain businesses who prefer being more autonomous, but more use this as a capability and a service.
• 39:38
  SPEAKER_03: So identifying what kind of requirements are there for each domain and how they also classically operate might be a very important aspect for a data and AI leader to take us with.
• 39:50
  SPEAKER_01: Yeah, but their their readiness and their culture of say wanting to be on the leading edge and innovate versus wanting to be fast follower or learn or reduce their risk.
• 40:02
  SPEAKER_01: So that makes a lot of sense.
• 40:03
  SPEAKER_01: Um, a recurring theme in your work is that AI increases the importance of clean structure, well-governed data.
• 40:09
  SPEAKER_01: What core data and knowledge practices are essential for sustainable AI adoption?
• 40:16
  SPEAKER_03: So I'll I'll I can again mention something which we talked about before.
• 40:20
  SPEAKER_03: A lot of data teams today are well, I think, equipped by and by and large from a technology standpoint.
• 40:25
  SPEAKER_03: So they have enough skilled people, let's say data engineers or data scientists, or otherwise you they they know how to go about staffing and recruiting or upskilling their team to do this.
• 40:36
  SPEAKER_03: But there isn't enough people, in my opinion, in the data organization who A, first understand the business and B, also are really kind of focused on adoption.
• 40:46
  SPEAKER_03: So we also require a lot of kind of this business development or sales kind of people, I feel in a data organization who can really sell in that word, if I can use or help adopt the data landscape within the business unit.
• 40:58
  SPEAKER_03: So I think that mix of this technology, which is obviously required, mixed with the business and these kind of sales and business development focused people, I think are is an important mix.
• 41:09
  SPEAKER_01: Yeah, yeah, no, that makes sense.
• 41:11
  SPEAKER_01: So I know uh we're uh getting a little bit close to time here.
• 41:15
  SPEAKER_01: So I do want to kind of ask the question of where you see the most meaningful opportunities for AI in the organization over the next couple of years.
• 41:25
  SPEAKER_01: Which why what should leaders be paying attention to now?
• 41:29
  SPEAKER_01: And what steps can they take to prepare?
• 41:32
  SPEAKER_02: Yeah, I think one of the key uh challenges with AI is not about AI being ready for people, but are people ready for AI?
• 41:40
  SPEAKER_02: And that means talking about it's not the easy wins, are those productivity gains that you and I or Sid have in our daily work using Gen AI to summarize a mail, understand documents.
• 41:53
  SPEAKER_02: That is one aspect of productivity.
• 41:55
  SPEAKER_02: But next is that to truly get to this automated stage that we talk about in an autonomous enterprise with AI agents executing, is that the knowledge of those individuals needs to be captured, the business context, right?
• 42:09
  SPEAKER_02: Which are not properly documented today.
• 42:12
  SPEAKER_02: And it's not about only a process of getting documented, it's also about clarifying that incentive model to an individual that he or she has been in the organization for 10 or 15 years, his or her differentiation is the knowledge they capture.
• 42:26
  SPEAKER_02: So, why would they go and document everything for an AI agent to get that context and that person becomes not relevant in that organization?
• 42:35
  SPEAKER_02: So I believe that is where the key challenge of business value realization will come in beyond the individual productivity goal.
• 42:45
  SPEAKER_02: I think that's a key roadblock.
• 42:46
  SPEAKER_02: So it's not about technology, data problems will also be resolved, but ultimately it will land upon the people of the enterprise, how the leaders are clarifying that incentive model to each individual.
• 42:58
  SPEAKER_01: Yeah, being able to take that tacit knowledge, capture it, process it, be able to embed that in your technology.
• 43:07
  SPEAKER_01: I agree that that is uh the next uh evolution.
• 43:11
  SPEAKER_01: Any other uh thoughts, Sid, that you wanted to add?
• 43:16
  SPEAKER_03: Well, I I think Suj made a very relevant point.
• 43:18
  SPEAKER_03: I think for if I think more also from the C-suite standpoint, it's also reimagining your business.
• 43:24
  SPEAKER_03: So not just thinking about how I can just embed AI in my existing business process, but really challenging that business process itself and reimagining perhaps with AI that entire business process may look completely different, right?
• 43:38
  SPEAKER_03: And how and how you can do that for each and every part of your enterprise is where I think a C-suite and their understanding of how AI not just looks today, but having that proactive understanding.
• 43:49
  SPEAKER_03: Because as we all know, every you know, every week when we look at it, there's new models coming out, there's new innovations coming out, right?
• 43:56
  SPEAKER_03: So to be able as a leader to look at this two, three steps down the line and not just you know implementing AI what it is there today.
• 44:04
  SPEAKER_03: Because as we all know, I mean Chat GPT came uh three years ago.
• 44:07
  SPEAKER_03: How many of us are using the the base model that ChatGPT released with, right?
• 44:11
  SPEAKER_03: So that's that's important.
• 44:13
  SPEAKER_01: Yeah, great.
• 44:14
  SPEAKER_01: Well, Suj and and Sid, thank you very much for joining our conversation today and sharing your insights.
• 44:21
  SPEAKER_01: I really appreciate it.
• 44:22
  SPEAKER_02: Pleasure is all our Sid.
• 44:24
  SPEAKER_02: Thank you.
• 44:24
  SPEAKER_02: Lovely discussion with you.
• 44:25
  SPEAKER_02: Yeah, truly enjoyed it.
• 44:28
  SPEAKER_01: Yeah, and for our listeners, we'll include links to data as the fourth pillar and related resources in the show notes.
• 44:34
  SPEAKER_01: Uh, thank you all for listening, and we will see you next time on the Early AI Podcast.
• 44:41
  SPEAKER_00: Thank you for joining us on another deep dive into AI innovation.
• 44:45
  SPEAKER_00: Tune in next time when we introduce another industry expert and discuss how to maximize AI in your world.
• 44:51
  SPEAKER_00: The Early AI Podcast is sponsored by Early Information Science and Vector.
• 44:56
  SPEAKER_00: To learn more, visit early.com.
• 44:59
  SPEAKER_00: That's E A R L E Y dot com.
• 45:02
  SPEAKER_00: Thank you for tuning in.