Video: Navigating the Human Elements to Modernize your Data for AI Transformation | Duration: 4232s | Summary: Navigating the Human Elements to Modernize your Data for AI Transformation | Chapters: Introducing James Serra (29.23s), Introducing Chris Samulski (160.575s), Human Element Importance (337.62s), People and Process (435.845s), Data for AI (596.91s), Drivers of Modernization (774.57s), Data Evolution Challenges (990.08997s), Decoupled Architecture Challenges (1066.55s), Modern Data Architectures (1318.0801s), Data Architecture Evolution (1464.105s), Data Product Strategies (1935.64s), Change Enablement Strategies (2716.07s), User-Driven Data Transformation (2840.04s), Evolving Data Roles (2977.49s), Choosing Data Partners (3092.65s), Accelerating Client Success (3292.5051s), Customer Success Stories (3704.67s), Modernizing Data Architecture (3812.97s), Modernizing Legacy Systems (3898.75s), Importance of Data Modeling (4031.525s)
Transcript for "Navigating the Human Elements to Modernize your Data for AI Transformation":
Technology, but also aligning people and processes so that you can really unlock value of AI. Hello again. And I'm Chaitra Agastia, director for strategy and innovation at Argano, and I'll be your for this session. Quick reminder, if you have multiple issues or content questions, please use the q and a box. Closed captioning is available. Just click on the CC button, and this session will be recorded, and you'll receive a link after this event. So I wanna jump right in and introduce speakers for today. Alright. Are we ready to do this? I see. Ready to do this? Yes. Awesome. Awesome. Alright. It's my honor to introduce our guest who is a true legend in the data and AI world. Actually, I'm a big fan. He has a remarkable career that demands nearly four decades across the private, public, and federal sectors. He's currently a civilian industry advisor at Microsoft, and he helps federal agencies unlock mission critical insights to the power of cloud and AI. From architecting billion record data warehouses, NASDAQ and direct edge, He's spearheading some very good enterprise scale data fabric solutions with, you know, in his past clients as well as his current clients, curating really cutting edge AI solutions for the customers of the Microsoft NTC. He's done it across 25 top notch firms, and his journey from early days of coding, I think it was Turbo Pascal, to shaping the next frontier of government innovation is nothing short of inspiring. I want you to please welcome me, in welcoming, welcome help me in welcoming, James Serra, a data visionary and lifelong data evangelist and the author of best selling book on data, Deciphering Data Architecture. Welcome, James. Well, thanks. What a what an introduction. When you said four decades, I wanted to go, no. It can't be four decades. But but it is. And and even before turnover Pascal, I started in COBOL, but we won't go into my mainframe days. But thanks for having me. I'm really happy to be here. Also, Pleasure to be pleasure to have you here. Our next guest is also a powerhouse in the world of data and AI, Delphi. He has a career with over two decades in the space with starting with actually being an air force, you know, person. Right? And managing mission critical aircraft maintenance and leading cutting edge data engineering practices in enterprise tech space. He's currently serving as the vice president of delivery for data and AI at Argano, and he has driven many transformative analytics strategies and cloud data solutions for some of the most data intensive industries. I think he's probably one of the few I have met with deep hands on expertise across the Microsoft data stack, an excellent track record of building and mentoring high performing teams, and a unique ability to align technical innovation with business outcomes. He brings both vision and execution to the world of intelligent enterprise solutions. I think all delivered with this unique blend of charm and very humble attitude. I am thrilled to welcome him to the webinar, so please join me in welcoming Chris Samulski. Thank you so much, Chaitra. It's really exciting to be here today and talk about a topic that I'm so immensely passionate about. I think all too often we focus on solely the technology and we don't talk about the strategy anymore. We enablement that needs to happen to make these amazing capabilities and really get the age of AI working inside your organization. I had the pleasure of reading James's book earlier this year. We connected and I think that's really what got us on this road and I'm so excited for us both to have this little bit of a fireside chat and talk about both our experience and also the practical things for how you can help your organizations really progress their AI and overall data strategy. I think back to, early days within my career when I started, there was a a foundational book that set me on my journey, which was Ralph Kimball's Data Warehousing Toolkit. And, I think very much, all data professionals that came up during the early two thousands had a copy of that sitting somewhere on their desk. I know James did. It's something that we bonded over. And what I shared with James is I felt like there's been a gap, with the advent of big data and these new architectures and these new systems, we've been missing a philosophy book. And so, you know, James, I'm so excited that your book is out there. I really feel, and I've shared this with before, that, this is the the next evolution, of Ralph's work. And I really hope that, data engineers and data professionals will have your book book holding up a monitor stand, at some point in the future. Yep. I agree. And I I have had the chance to read that book myself, and I can say it really does live up to that phrase. I think it's not just about technology choices. Right? One of the things that James made me really go back and think about was how does it how do you think through the human organizational and strategic implications of data architecture. It's not start about technology but also the human elements of it. And that's exactly why, you know, James, your book has received so much praise from both industry leaders and educators is that and I think there's one review if you can, you know, head back there. It's, you know, a senior engineer at Google quoted, you have the ability to transform these complex technical concepts into clear, easy to grasp explanations, and it's truly remarkable to see that in through your book. Another code I would say is, you know, definitely we're looking at, a primer for college students and business professionals who who really wanna get to the bottom of it. Start from the bottom, see how it evolved, and then they can grasp those concepts. So that brings us to, you know, today's fun part as well along with this journey. We're gonna raffle off these books. At the end of it, you will be notified of the winners, and you will receive your books, a copy of these books. Five people five lucky people are gonna get a copy of James' book so you can have it on the shelf just like me and and get started with cracking that code of the human element and data architecture. So that brings us to our our goal. So, James, I'm I'm gonna, turn it over to you. What made you think about this human element in the journey of companies as they're going through these data architecture transformation? Oh, well, having been in the industry forty years, I still can't believe that. But I've I've been involved in a lot of projects, data warehousing type of projects. And in almost every case, when they failed, it was not because of the technology. It was because the people in the process on there. Now there was every once in a while, they would choose a technology that was really not suited for what they were trying to do. But more often, it was not understanding the work involved in building out a data route solution, not defining having the good the right people, not having defined roles, not having somebody, as another example, moving the project along and having appropriate checkpoints, not getting the users involved in there. All these little things I've seen over the years that said, well, wait a minute. This this really has nothing to do with technology, but rather the people in the process. And and I think it's kinda amplified. You agree with me, Chris, especially with everyone talking about AI? Absolutely. I think when you think about the core thing that any data modernization program strives for, it's adoption and trust in that data. And that's at the core of getting back to this people and process conversation. What we do, the data that our businesses create, that they collect, that we synthesize into intelligence, that we hope to have the goal to have AI action on and gain efficiencies and give us these new transformative capabilities, it can only work if people trust it. And if only if they're aligned on how they're going to govern it, and ultimately how they're going to move that journey forward. So I really think that at the core of this, the imperative to make AI work, yes, you need to be in the cloud. Yes, you need these modern systems, these modular approaches and just, you know, all of the architecture we're also gonna talk about today. That's Yep. The plummet. But the people and the process are what's gonna make it be a success for your company. I agree. And that's exactly the goal of today's session, folks, is we're gonna explore how modernizing your data platform and your people practices is actually the key, the true power to unlock the AI within your organization. Such a big business imperative today. Right? So, again, I wanna pull up a poll here that's coming up shortly. Let's see if we can get that poll out. Awesome. Alright. How confident are you that your current data infrastructure supports AI driven insights? I love this question. This is really good. Well, I see quite a few words coming in. And as as we okay. Okay. We're we're starting to see some more words pile in. Alright. Somewhat confident. I think I'm everyone's leading to somewhat confident. I see a little bit of not so confident as well. Very interesting. I think we're gonna see at the end of it how we can get you to be confident in your data journey, right, along with some tips and tricks and keys to success. Thank you so much. And keep bringing those votes in, and we're gonna head back, to our presentation for today. Alright. K. And sorry. Give me one second. Alright. So here we are. Awesome. Yeah. I think I think one of the things we are starting to see is, you know, everybody is asking why are we starting with data foundations? Why are we starting with, you know, this this this kind of new imperative? If AI is a new imperative, why why do we need to do this? James, what are your thoughts? You've you've seen a lot of clients go through pain points, in unlocking AI value. What are you seeing in the industry? Yeah. AI has just completely transformed the value of data, And people are now using data whether it's unstructured, semi structured, or structured data to get more insights into their business than ever before. It's it's opened up a whole bunch of other doors. The data is all the the answers you get from AI are only gonna be good as the data, and Chris kind of touched on this before. It it we used to talk about junk in, junk out. And if your your reports only gonna be good as the data that's collected and cleaned and mastered and aggregated. And now it's the same thing applied to AI. So your AI that you run against large language models is not gonna be accurate if the data isn't good, that it's built the model the models are built on or the way the questions are created, and that's where you get into prompt engineering and whatnot. But the the value that I've seen now out of structured data and the unstructured data with large language models is it's just amazing. And I think we've just scratched the surface. It's amazing in thinking that two and a half years ago, we didn't know what Chat2BT was. And and look how far we've gone. And within Microsoft using Copilot for everything, whether it's transcribing your calls or using it to help you create an email, it's it's incredible. Yeah. Yeah. Absolutely. Chris, you you see some challenges. Right? If if if we are starting to ask the question of why, I know that and we see from the poll just we just did, they're somewhat confident. They're, like, you know, mostly, yes, we are somewhat confident, but we still need more information. What do you think is the biggest driver for data modernization? I think the largest driver that a lot of organizations are seeing is complexity. Right? As we continue to see that shift to the cloud, many organizations still are operating on organically evolved, fragile, complex environments. And, you know, this is I think why we see such synergy around tools like Microsoft Fabric, Azure Synapse, Databricks, even outside the Microsoft space with things like Snowflake. These platforms, are gaining traction because the things that we used to do as data professionals, the nuts and bolts of doing modeling, of doing data movement have become easier with these tools with the scale of the cloud. I think that many organizations have started that journey, but where they're probably struggling is governance, adoption, and really, like, getting to that center of excellence model. And I keep going back to that. Everybody for the last twenty years has had this goal of self-service analytics, self-service platforms. We wanna do more with less. This has been the promise of computing since the PC came on the market. Right? And I think we're actually poised at a time now where AI is gonna make that promise real, but it's only gonna be accessible to organizations that really go into this intentionally. Right? So we have to break free of those shackles of an organically evolved data estate. It needs to become a part of company culture. And again, looping back to it's not just the technology. The technology is now easier than it's ever been before, But really where our our challenge is in aligning people, gaining consensus. It's in, how and where is the right place to pull that data from. And that challenge only gets more difficult the larger the organization. Yeah. Agreed. I absolutely agree. In fact, sometimes when people want to customers want to launch their AI journey, one of the foundational elements of AI readiness is actually looking at your data foundation, and we see those cracks. Right? Sometimes a lot of times those cracks really matter and halt your journey to progress before you've you have to fix it. You have to remediate it. Even bring your people along with it. And as we have seen, James, in your book, you know, it's so important that human element that everybody is aligned as they go through from the left hand column of that journey to the right side of that journey. Right? From from being that more reactive to being more predictive and proactive is where you're really unlocking the AI value. Thank you for sharing that. I think, as we were looking at what's really pushing us into this transformation from traditional to modern architectures, I can't help but think about this big data and the v's that James actually, if you if you read one of those chapters, you you outline it really well, the big v's of of data. James, do you wanna talk a little bit about about that? Sure. The data I I sometimes miss the old days when big data was a megabyte, maybe a couple of my two megabytes are there, and it would come once every week. And now if you look at all of these, it has greatly changed in the size of the data, the type of the data, the speed of the data. In large part, that came about when we had the Internet come around. And now all of a sudden, there was all this information people can pull down easily, and then there was data that gets accumulated whether it's just based on a large database or social media or data from IoT devices. Explaining data. Yeah. Yeah. All that was great for helping you make better business decisions. Can I look at real time data and on the fly make adjustments on my supply chain? Or is Twitter suddenly saying something that Kim Kardashian just talked about our product and I get notified and I have to now move shit products all over the the country in there. Those are actually use cases I've seen with customers on there. And then the challenge becomes how can I build an environment, a warehouse that's gonna be able to handle all this data depending on the size, the speed, and the type? Yep. And and and and a lot of times, you know, we face challenges. I've seen clients that don't really understand the decoupled architecture, but then we are starting to see that decoupled architecture. Chris, you you you seem to have some experience bumping into, you know, areas where the decoupled architecture really is taking off. And because of that, we're able to scale data. Yeah. I think this has been, like, the big piece, and and James is dead on that, like, it's been the rise of the internet in the information age now giving way to the AI age that has made all of this happen. We have more data being created every second now than has existed for the mass majority of human history. And we now have systems that have to work in a fundamental communication method, the internet, that is actually unreliable. We talk about this all the time in delivery in our projects about the well architected framework and how to properly work within the cloud, building for resilience, building for scale, building for performance and disaster recovery, and all of these what if scenarios that can happen when you don't control the fundamental underlying architecture. And at its core, the Internet is still kind of a request and response method, which means also real time is technically not perfectly synchronous. And because of that, it's caused this rise of having to create decoupled resilient architectures. And so, any of these new systems that you look at, that's sometimes one of the first cultural changes that we have to get our users over the thought process of. If an organization's moving from an on prem system, or an organically evolved system, they may have a perception that their reports are in lockstep with that transactional application, whether that be your CRM or manufacturing system or their ERP. They're used to running reports that reflect a transaction that just happened. And it's not that that's impossible in these cloud native systems. However, it's a balanced trade off. Everything drives cost. And again, that's why we really prescribe to this well architected framework. Ultimately, that is a decoupled system. There is an integration that you have to drive to get into these unified platforms like Fabric and Synapse and Databricks and Snowflake. Managing that and answering the core questions of what's the impact if this system's down, what's the impact to my business if that integration bringing in, sales orders can't meet. It all ends up coming though back to a human element. It's an ability for us as data professionals to be able to align on an agreement for how we're gonna serve our communities and ultimately drive both our value to the business when it comes to how do I deliver that time to insight, but also how do I give a confident and predictable experience. So I find organizations at times that are early in that journey have a hard time understanding this concept because users wanna push back and go, I need my data right now. I need to be able to do this, like, within the second. And then we show them the cost, and they go, it's not worth it. And there's this misconception that the journey is not worth it. So really coming into this and having more strategic view, what is the value of your data? What's the velocity of that data? And really critically thinking about what is my decision making speed? How fast can I affect change in my organization? Once I have captured this insight, that is the best way to assess. How do we design that system overall? And these new architectures that have come about through the age of a, through the age of the internet, now let us do things we've never been able to do before. We can store everything and figure it out later because we have cost effective scalable storage. We can follow multiple paths for integration following something like the Lambda architecture where we can do batch processing and streaming processing. But all of those render out in different experiences. Yeah. I I I think that that actually, you know, makes us think about, you know, how the data architecture itself has evolved with these big drivers, right, to this evolution. We saw the decoupled architecture. I think the best part was schema on right to schema on read, which really drove some of those platforms delivering these data architectures. I want us to kinda head to our our our next, you know, topic here, which is all about the different flavors of data architecture. And and and I I think I think you may have or you have not, then welcome. These are the four modern data architecture options that are very popular, but I'm sure everyone has heard of them or probably even started implementing some of them. I wanna start, James. Which one's your favorite? I'm gonna go with data lake house and Data lake house. And that's in large part because, essentially, that's the behind the scenes architecture that is used with Microsoft Fabric. Yeah. I agree. Agreed. With that, I wanna launch our poll. You know, I just wanna know from everyone, you know, what kind of, architecture you use today that's really successful. And it's okay if you have a layered architecture. We'll talk a little bit about that, about some layering of this architecture. Awesome. Let's, go to our polls tab and share. Alright. Spreadsheets. We see some spreadsheets. Okay. Datamart, reporting database. Yes? Spreadsheets are the number one analytical tool in the world, and they're never gonna go away. And that's okay, Morgan. Actually, there's a contest on who can do better spreadsheets. I'm sure I think it was fun watching that. Okay. We we have a lot of words coming in. I I guess, looks like data lakehouse is the most popular. I see a lot of folks doing data lakehouse. Thank you so much. And keep those full, you know, keep pulling, keep submitting your answers. We're gonna check back on them and see how we can help you. Awesome. I think this warrants us to kinda deep dive into some of those. Chris, let's talk about that medallion architecture. You know, I I love that we have a nuanced approach to this. Why do you think this layered approach is the most effective for our for the data modern data architecture? Yeah. I think, you know, we come back again to that decoupled design. Right? Like, this overall data life cycle that you see these modern tools follow, going from ingestion to storage to transformation to modeling to serving, It's going to be the common pattern across all these data architectures. It's about the nuance of how or what technology or experience we leverage in each one of those steps. And I think there's a consequence to dealing with a system that does schema on read. While, yes, we can store everything and figure it out later, our goal is to not allow our data lake, that fundamental storage technology, to go ungoverned and get messy. And we find, you know, I think I share with James that, the lake house is kind of my preferred fat pattern, and it's because it's got the most growth potential to go in other directions and be able to, evolve into elements of data mesh or data fabric or evolve into a modern data warehouse when appropriate. But this model of the Medallion architecture, this idea of progressing from raw data into a different storage medium that gives us the building blocks of data to ultimately the analytical products that we're going to serve to people is kind of a core foundational design pattern that you see across all these architectures. And I want to actually hone in on the layer that's not talked about often. I think everybody intrinsically understands bronze. Right? Store everything, figure it out later in its native format. I think everybody understands gold. Right? This is the thing that's gonna build my report or help my AI. Silver's where the magic happens. Silver allows us to view our work as data professionals as Legos. I now have what all of those kind of, best practices that come from the object oriented programming world, this idea of classes and and building blocks I can reuse. If I start viewing my data that way and then make that accessible, I now can make n number of products and not have to go do through that ETL cycle extract, transform, and load every time. This pattern is what let us flip that on its head and say, we're gonna extract, load, and then transforms. Right? Yep. Yep. Agreed. So you're going from ETL to ELT. And question, for James. James, do you feel like organizations are struggling with one of these layers a lot more than the other layer? Yeah. I would say that the silver is. I because a lot of customers don't understand the time and complexity of cleaning data, and that's where that magic happens in the silver layer. Too often, they they think all the the data I pulled in from these systems is fine. It doesn't even be clean. And then they get into it, and they're like, well, why do we have birth dates in the future? Why are people 300 years old? And then they don't think about, oh, wait a minute. We're pulling all those customer data. Now we have duplicate customers. I don't realize that was gonna happen. Now we have the master data and come up with golden records in there. Or I I don't know how to join all this data. We need some kind of common data model. Oh, that's a lot more work. So the silver layer is really time consuming, and too often customers don't put enough time in the project plan for that data cleaning. And then you get into the the governance aspect of that, and let's avoid reinventing the wheel. Let's make sure we're pulling in data that's not already there. That's where you need to to think about update a catalog. And then lastly, governance as related to security. Can Can we make sure the sensitive data is not seen by other people in there? And that to do that, you have to make sure that sensitive data is being tracked and being secured, which is yet another challenge that people have and that's mainly combined to that silver layer. So not we're not just thinking about transformations, but we're also thinking about applying governance and best practices and the right security. So think of this as as the place where you really are creating this sculpture from being this plain brown color that you've just figured out to actually having colors that it can render and look really good when you land it, but you're choosing the right colors. You're choosing the right templates. It's it's the right way to go about it. I think that brings us to, you know, some of the modern data warehouse architectures we have seen. And and, Chris, this this is something that I wanted to explore a little bit deeper, especially in the machine learning space. You know, as as we take our data through this architecture, how has it helped clients be more ready for their real world AI scenarios? Yeah. I think, you know, the the benefit and the strength of a data warehouse is it follows a more, megalithic design towards its model. Right? It's meant to work, cross domains. It's meant to be able to bridge from finance into sales and into product, and it's a cohesive model. It takes a lot of work to get to, that gold layer, which is, you know, ultimately a relational database feel under the hood, but served by the fundamental, you know, data lake technology to get us there. I think this, however, gives us a lot of ability with that flexibility of the data lake in that, transformation step to do enrichment. Right? So I think about, like, the traditional things that we would have done in a traditional data warehouse that we always struggled to execute on. I need to be able to do a demand forecast. I need to be able to do time series prediction. Now that I've got so much more of a flexible storage technology, I now have multiple ways to be able to orchestrate my transformations to load that model. It's a lot easier to be able to do that machine learning infused into that data warehouse after the fact. So I'm not struggling with being able to do that because now I have T SQL or tables as my only storage concept to be able to deal with. Right? Maybe I've got an amazing time series, algorithm that I'm leveraging from the latest open source community, and it's written in r or Python. I don't have to figure out how to get that into a stored procedure and orchestrate it and all that. It just goes into my notebook as a cell native in that and boom, it keeps going. So this pattern, I think, really helps enable that, but there's also a consequence to this pattern, which is that ultimately I'm still landing things in SQL, and, those tables have to be structured and they're more rigid. I I was thinking as you were talking about the medallion architecture and all these in general, to bring this back again to people. Right? We're all working with smaller teams. We're all working with, you know, less resources. And the goal, right, is AI to extend us and be able to help us do more with less, which is being demanded by, you know, all of our businesses today. This architecture is great, but I think its challenge remains that it's still very IT heavy focused. I think at the right time, these models make sense when you need to be highly governed, highly secure, highly predictable for your schema, and ultimately how you serve that. But in my mind, and I would love to hear from James on this thought, I see this as still being on a spectrum still higher on the IT governance than really truly enabling a self-service model. What's your thought or your your, your experience in that space with modern data warehouse? Yeah. I totally agree on that. This is I'll call it the old school approach on that, which we had to do back in the day because the tools weren't there to make it easy for end users to access the data. I always think back of when the data lake first came out, people thought, I'll do everything in a data lake. We don't need relational databases. Just throw everything in there. And then they went to the end users and said, everything you want is sitting in files and folders and go use Hive Yeah. And have at it. And it just it failed miserably in there. And and then we figured out, well, the data lakes will have its value, but in specific use cases. Well, over time, those use cases have expanded. And even I was surprised when I said you're always gonna need a relational data warehouse, but then these tools came out. And I said, well, maybe we can get away without having this relational storage. And now I'm in in fully embracing that in very rare cases, you don't need relational storage. You can do everything in the data lake house within a data lake and brought about by features of Delta coming about and making a data lake more data or health like in there. But you still wanna model the data because you wanna present to the end user a model a relational model and not just files and folders. And and so the the main point is because of the technology advancing so much, a lot of the work that used to be done by IT does not have to be done anymore or can be done very easily by the end users. And you look at Microsoft, which is amazing in that it's a platform to build solutions, not just for engineers, but now for your anybody and then your power user and that. And so that's because of the ease of use of technology on there. In the end, it's all about self-service BIs as Chris brought out. I agree. And and can you do enough in IT to make self-service BI a real thing as and allow the tool allow it to be easy for the end users to do a lot of that work and not have IT as a bottleneck. I agree. I I think as as the data volume grows, as there's more expectation from our customers, from the end users to see different types of reports at the speed at which they want, it it becomes just it becomes a bottleneck. We wanted to be more self-service, rendering in, on time, live, and trying to extract that report as demand grows on demand. And and I and I wanna kinda highlight and and emphasize one thing throughout this this conversation, the idea of data products. I mean, if you had asked me data what data products are five years ago or ten years ago, I would be like, what what do you mean by data products? But it's starting to take more, I would say, emphasis as we talk about AI, as we talk about, different roles, doing different things in the data architecture. We talk about what are the data products that they're responsible for. Chris, you seem to encounter the notion of data concepts every day as we talk to our customers. Right? We we try to deliver value through the data products, but these platforms help us build and that eventually drives value. What are your thoughts on that? Yeah. I think when you think about a data product, you know, I come back to this concept we talk about, which is it needs to be a small enough, data model, typically a star schema or a very wide flat table, two different modeling approaches, that speak to a business problem, right? I mean, at the core of this, we still come back to what am I trying to measure, what am I trying to measure it by, and what's its value? And I think, you know, as you think about those data products, how you choose to ultimately cut those up comes back to who your customer is. And it's why you've as you go through building these systems, you've gotta have this this culture of pulling the business in and getting those requirements from them, to really truly deliver these things. We say it. It's something that I think every single person on this call has always said or heard that IT led IT projects fail. You have to have the business input to understand their perspective and more importantly, build their semantics, their language into that. And I think as you circle back to AI, that's the other important piece. Right? When you think about these data products, they're the thing that controls how we describe our data. When we look at this chart up here, the metadata, how we describe that data, it's, synonyms for product or item or finished good depending on what our industry is. We gotta marry all those perspectives together, and that's inside of that data product. Yeah. Agreed. And I think when, when you're thinking about different data products and you have this wide variety of options, James, how do you help clients figure out which is the right choice for them for the right data product, I guess. Yeah. It's when we look at the architecture, I tell customers, learn about all the architectures and then choose the bits and pieces that are gonna work for your situation because every customer is different. They all have different data sizes and speed and types on that. And and I I wish it was a world where we can just open up a playbook and ask a bunch of questions and go, this is the architecture that you need. But the reality is it it's gonna be different. Every I've done so many ADSs and it was never the same solution. And so when you think of when you understand the architectures and decide which approach you're gonna take and then build out that architecture, somewhere in there, you're gonna have data product or data of the product on that. And then you have to define what do you mean by a data product on that. And there's there's maybe the the light approach is I'm gonna say a data product is just a semantic model in Power BI. And then there's the the the farther approach which the data product product is all encompassing and and maybe there's an API call that you you go and request the data and you request the metadata and it's self cleaning and all this other stuff that is much more complex and and quite rare that anybody's gone down that path in there. And and so it's gonna matter depending on your organization. And it's and it's ideally, you may have what you think should be a data product and then realistically, what are you gonna be able to implement? Because that's the technology challenge in there. Something sounds great in theory, but if you say, well, I wanna create this data product and people go, we don't have the technology or the people to build it like that, then maybe you scale it back and say, okay. Maybe it'll be a semantic model. Maybe we'll follow some kind of contract that everybody's gonna follow in their particular data solution to have data as a product. Yep. Agreed. And and then I think, your your book, James Serra, you know, talks about these mistakes, costly mistakes even that's costed organizations millions of dollars because they didn't go through an ADS session to identify what their business objectives are. What are the data products that really drive value? Chris, you've had experience, driving some ADS sessions for our clients. How have you seen that really help our clients in making the right choices to those ADS? And what what is ADS? Yeah. So ADS, architectural design session or architectural design workshop. These are honestly my favorite workshops that we deliver in general because it I always joke that to be a successful data architect or data consultant, Our job is kind of one part DBA, one part MBA, right? We've got to understand business, one part therapist. Right? Like you see so much business trauma come out in these sessions. Everybody thinks it's gonna be talking about data pipelines and T SQL and stored procs and notebooks and arguing about that. But really, the focus is how do we serve the client's needs best? And you find as you go through these workshops that are intentionally cross departmental. You have IT in the the room, you have the customers of the data, you have people that are involved in the life cycle of that data there. People start talking about what has caused their silos, what has caused their access their accessibility challenges. Be honest. Like, I've been in a couple where there's actual finger pointing. Well, I can't get this because you're not doing that. And you end up having to, as the consultant, as the the mediator in that room, help navigate those very often human challenges. Agreed. I I can't I can't I have been through some of those sessions and they're so valuable to not just, you know, make the right choice, but also enlighten yourself through that journey and learn about these things, you know, and and and and that makes you be a better decision maker. So if if you have done this and you're ready to choose, your, you know, data architecture, I I I kinda want to hear from you. What are the top key things that they should keep in mind when they're when they think they're ready to choose? James Serra, I wanna start with you. Yeah. And this this brings up some good points here, and and it was funny that Chris Samulski brought up it's like therapy. Because before I started doing ADSs, an experienced architect saying, beware it's gonna be like business therapy. And I'm like, what are you talking about? And then after the first one, I was like, oh, I I know exactly what you're talking about in there. And and and the ADS is vital to prevent or hopefully prevent any mistakes that you made in your technology decision or even in the architectural flow of the data, and you address those concerns early before you start building something. So I would recommend do an ADS and then before you go down a path of building out a a solution in there. Because an ADS is also important in doing discovery. And and when I'm in front of a customer, I'm asking a ton of questions there, and it could be things about what's your skill set, what's your timeline, what's your budget, and the architecture is gonna change based on those responses on there. So so the the discovery piece is really important there, and a lot of times these people have never been in the room and discussed all this. And so this is where a lot of the arguments come out or maybe the misunderstandings come out, but they're all taken care of way early in in the project on that. And then and then finally, it's understanding these things that like the no one size fits all is that you you need to keep in mind that you need to build a solution that's gonna handle data no matter what the size, the speed, and the type. So don't just look at your needs today, but what you're gonna be needs in in tomorrow, in the next month, or the year. So you don't go a year from now and go, oh, no. We didn't think of this. We didn't realize we need real time data, and now what we build is not gonna work. I agree. I I think one of the things you mentioned in your in your book, James, was that it's changing how data is managed and processed. And and how it's valued means the teams need to really think, you know, differently and and interact and and see how they interact with the data. You're almost challenging those norms. So you wanna really invite those people who are gonna use it to be part of these decision making sessions because eventually, we are these are the consumers of those data products. So it's it's not just merely technical making those decision, but where's the human element in it that's eventually gonna use and consume it and provide feedback even to get better? And and and that kind of I just wanted to kinda emphasize on on why I I kinda wanna double down on the human element of this data are, you know, transition that clients are going through. It's not just tech. Chris, in your in your opinion, you know, how do you how do you think clients can leverage some of the best practices, and what are you what are your top key tips for success? I think one of the the main things is, it's gotta start with education, and it's gotta start with, a cohesive vision of what the end state is gonna look like. Right? And what I've kind of just talked about is change enablement. Org structures, teams, collaboration, all of this is gonna look different than what a lot of organizations who have not adopted these architectures yet are gonna feel. Then when we talk about AI, we always talk about the human in the middle. Right? So we have to keep that human element involved. So starting with your roles and responsibilities, how does my data admin database administrator, what does their job look like in this new architecture? For my business analyst, what does their day to day look like? If we aren't talking about that in our ADS and as we're forming this project, you're gonna have this very human reaction to pull away from this. And I think early on in my career when we were building data warehouses, you know, in relational systems, that was the first hurdle we had to get over. Right? I remember literally having, someone, in, when I was working with the Air Force, looked at me and said, you will take my spreadsheet from my cold dying hands. And how how to make that person feel comfortable because that's how they saw their value in the organization. They saw their value as I infuse myself into this product and without me, how is it possible? If you don't go through the process of helping that person reinvent what their view of their job is and how they can also participate in that, your project's also gonna fail. Completely agree. I I mean, it's it's one of those things when you go through a data transformation project and you realize that one of the key drivers for success in this project was actually the people and the culture that you transform that's gonna sustain your success. Right? It's not just about the transformation that happened, but sustaining the transformation. And and, James, you kinda laugh when when Chris, said, you know, I'm gonna take the spreadsheet from my cold hands. You know, you'll have to take that. And and and what do you think has been like, if I were to pick one tip, the number one tip that I wanna keep in mind, what what what in your opinion would be that one tip? As as Chris mentioned before, get the users involved early in in the process there. If I go to a project and I ask IT what have the end users said about what you're building, and I go, well, I know what they want. I don't need to talk to them. That project is gonna fail a % of the time on there because they that you have to get understand the users. The first thing about users is they hate change. So if you're gonna go in there and change the way they do things, they have to see the value, and that's showing them the art of the possible. And that's saying we're gonna build a solution that's gonna make your job easier. It's gonna save you time. It's gonna make you look better than your manager because you're doing things with the data that they never thought of on that. You're gonna get better answers on there. You're gonna save money. Those are the things that excites the end user. We want to get involved and want them to six want you to succeed even though it may be changed. But you're always gonna fight that battle that, hey. I use this Excel spreadsheet, and I build solutions in great dashboards and Power BI. And the first thing the end user goes, where's the export Excel button? Because they that's what they know and they love. And you either show them the things they can do in Power BI, they they can't do in Excel, or you give them an option to export the Excel. But you have to keep them happy happy. And that's a big change I've seen in the last few years with projects is a lot of them now are being driven by the end users. They see the value in the data, and they want to drive it as opposed to IT. Sometimes IT, they they get it, but it just won't work for them. So they're kind of reluctant. But if you get the business users involved, it's gonna be better. Absolutely. I I think that's a great call out. Thanks, James, for that. I think the the roles itself have evolved, in this transition to modern data architecture. Chris, now if I were to pick maybe three or four roles that that have been heavily impacted in this transition, which ones would those be? I think, it's gonna be ones that don't have super clear, lineage for where that person is, inside of an organization, and I'm gonna point to the data steward, the data analyst, and the business analyst. I think a lot of organizations struggle to understand who is a data steward, and in my mind, this is really the simplest way to put it is this is the data product owner. This is the person who understands all of the, where did this come from, where and how is it managed, who has access to it. This is the trust builder. Sometimes that's a technical role, sometimes that's a business role, sometimes it's a hybrid. And I think that, like, when you think about trying to chart a course towards change enablement inside of an organization, if you're not thinking about who your data steward is, that's one tip I'll give you now, right, is that person is gonna be your champion because they run around and build consensus, and they build trust in that system. I I think in the past, we've often naturally thrown that to IT and thinking that, you know, it's it's gotta be governed. There's governance somewhere in this title, so that's IT's role is to govern. But another part of this is consensus building. It's it's trust building. Right? And often that's done best with the people who are gonna use that day in and day out. Awesome. Thank you so much for sharing that. James, you know, it almost feels like a democratization of certain functions, beyond IT as we see this evolution to modern data architecture. What are your thoughts on that? Yeah. And that kinda gets in the the idea behind data mesh is you decentralized this and you have other groups doing things to take the load of IT, which can work. It was interesting in the survey that one person said they're building a data mesh. So, the takeaway from a lot of what we've learned with data mesh is maybe nobody's building a pure data mesh, but they're saying, well, can we offer some of these responsibilities? So maybe somebody's domains, they clean the data before it's pulled in to a central location to create the analytical data in there. Maybe those domain teams who know the data best should be cleaning the data on that. Can you get that buy in from those others? Can we maybe have a federated solution where IT is not having to do everything on there? So it's that balance of that and but it gets challenging if you think of, like, data governance on there. Well Yeah. If everybody's if all the data is being copied to a central location, that's one spot that's gotta be governed for the data, and IT will do it. If if instead you have all these other domains keeping the data and creating their own analytical data, now you may have dozens of spots that have different people that are governing the data. And so how can you make sure all of them are doing it and all of them are doing it in the right way? Well, somebody's gotta govern that. That's a that's a maybe a central group. And maybe you work together to get that right balance between the two, but it makes things more challenging. I agree. And and a lot of times, companies partner with partner companies, to to sort of help, you know, evolve these roles or execute some of these roles to help them guide through in the initial phase as they kinda ramp up their maturity as well. So, James, the next question is, in your experience, how does how does somebody choose the right partner? What sets what criteria should they be using when they're choosing the right partner to do some of the data modernization work? Have they done it before? What's their experience on that? Do they know nuances of building out these solutions? And it doesn't have to be that they're gonna they build something exactly like you're looking to do in there, but overall, what is their experience level in multiple products in there? Because a lot of it is not just getting the deep technical expertise, but do you have do they have people that are good at running the whole thing in there? Do they understand the pitfall that'll happen? Do they realize people can't work forty hours a week in there, and do they put in enough buffers for other things on there? Are they realistic in their timelines in there? It's really hard to find smart, experienced people when it comes to building data warehousing solutions in there. So that's where even at Microsoft, we struggle with finding partners who who have that ex expertise in there. And then finally, it's are they willing and able to work with the customer to transition their knowledge over so that partner can walk away and the customer can keep things good only and not have to rely on the partner for everything. Yeah. I I completely agree. I think a lot of times, the partner also needs to bring the customer along in their journey in terms of understanding these foundational elements as well, the new modern data elements and how they work. James, in your experience, you've seen, you know, some instances where it was more about the change enablement, than the technical implementation that that was the key to success. Yeah. As as I said, people, they changed. So as a a partner who has got experience, they can say, here's the way we've seen it work when you implement change in your organization. Here's how you get end users to buy into what we're doing on that. How can we get excited? How can we show them the art of the possible? Maybe it's early on, we we do some training with them. Or maybe early on, we have an envisioning session. We show them all cool things they can do that they had no idea they can they can they can do those things on there. So it's really important. We keep saying this, but it's so important to get end users involved early in the project. And, Chris, how are you thinking about, you know, helping our customers? What would be that secret sauce that would be kinda help them accelerate that success? Yeah. I think, ultimately, it comes back to, in my mind, some of the core values of Argano. Right? Like, we we talk all the time, internally and externally about transparency and working together. I think transparency is a big part of it and trust in general, right? Like, when we deliver, you know, James, you hit the nail on the head that, like, ultimately, if we can't transition this this solution over to a client to be trusted, to be adopted, to be used, we failed as a partner. And and that, you know, we've not going to lie to anyone and say that we haven't had our lumps there. But what we've learned through those years is within the industries that we're we're working in, that we have deep expertise. We have the reps, right, where a lot of a client clients this may be the only data platform transformation project they've ever been on. Yep. This is probably that consulting team's twenty fifth or fiftieth that they've done. Right? And so being able to always approach things from an education space, understand your client, where they are, and having open discussions. Right? Yeah. Trying to build trust is one of the most important things of this, and, again, we keep coming back to it. Bring the users. Right? And make sure that we're getting people to see a vision, as to where we're going. I think the biggest takeaway, and we'll probably talk about this next, is start small. Get some wins under your belt. Get people excited, then expand. Right? It manages risk, and we have a methodology at Argano that helps manage that risk while showing value early. And Yeah. I think that's been one of our key success factors. I agree. And and, James, in your book, you kind of quote, an example of how a client went on prem, couldn't scale and grow fast because of their r and d needs, and they needed to really meet those demanding, requirements around customer insights. And and I think one of the, you know, when I was reading through that case study, I was like, wow. They just took that, pushed that into cloud, scaled it, and they were immediately able to see the benefits of that modern data architecture. And and and, Chris, going back to, you know, how we help our clients by staying starting small. Right? And you mentioned about a methodology that we would follow sort of to lift off. Could you kinda walk us through something that the secret sauce, right, where we kinda help clients just accelerate their journey to the modern day architecture? I think, you know, what we've got up here, on the screen right now is what we see as a typical journey to get started. Right? We start with ensuring that there's executive sponsorship, that there's buy in, that there's a road map that's developed. That first step of a quick two hour briefing to talk about platforms and choices and values. It's like a first date. Right? Like, we talk about values. We talk about experiences. We talk about capabilities and how technology is going to enable that. We identify in that what we think the roadmap is going to look like. From there, we get into the technical stuff. Let's go look under the hood. Let's go look at the data estate as it exists today. Let's evaluate its challenges. Let's start bringing users in to get their perspective so that we can identify a use case that's going to be something that's going to be impactful back to the business to get people excited. And then whether it's Microsoft Fabric or it's just Power BI or it's any of the awesome data solutions that we have inside Microsoft, starting with a pilot is a great place to start. Something that's a quick transactable, you know, then get excitement back to clients very quickly. Four to six weeks. Right? Get in there, connect some data sources, show the process of going from, bronze, silver, gold into a data product to set that vision. We'll set the right course for us to go into any modernization project. Thank you, Chris, for sharing that. This is this is wonderful. I think it it almost I can see the human element through it as you're not just establishing trust, understanding the statement, and empathizing with the client, with the customer, but also taking them along in the journey of helping them make more informed decisions. I call it the decision intelligence. Right? Because it's a new journey for a lot of clients. They're they're new to modern data architecture. They wanna learn some of these concepts before they make those decisions. And this kind of a methodic approach kinda gives them both the benefit of accelerating that implementation using well architected framework. At the same time, almost like increasing their value in terms of understanding this modern data architecture elements, as well for the organization. With that, thank you so much, James, Chris, for being with us today. I had I had a lot of fun, asking a lot of questions and nobody asked me. But, you know, I wanna let the audience ask questions. So please feel free to, I know we are about time, but, please feel free to stay back if you have time and ask questions. K. Let's see. We have a few questions. K. Okay. Well, as as the questions pour in, I may have a few, that I have, written up as well. James, what's your, most notable story or journey with a customer without I'm I'm sure, you know, you can anonymize all that. But which one sound which one is very profound close to your heart? The the ones that fell are most memorable on on there. The the most of them, like, succeeded in there, but the the ones that failed miserably, a couple of cases where they somebody a customer did say, we're gonna do everything with a in a data lake, and we're not gonna use relational database anymore, and we're gonna get rid of all of them. And as much as I tried to say that that's not gonna work this and this was ten years ago. They they fell midrude in there. The ones that I've seen involved in that work really well are and and I I like the assessment and all those things you're doing. That's awesome. I wish more partners were doing that. And and being realistic too because I'm in presales and everything works great in theory. But when it comes to building a solution, this is where you need a partner who's got the experience to say, well, what you're doing, Microsoft Fabric will work great except for this one spot where maybe we gotta do a workaround or we've had performance issues with that. So they know the ins and out of the product and and the areas that may not work well. And so my best experience is being an edge a really skillful partner like Argano and having the ability to predict the challenges that you may have down the road as you take on that project. Yeah. I agree. And and the book. Right? The book kinda kinda launches you off as well, which, by the way, we're gonna raffle off that book. It's amazing. It's a must have as as you've seen. Chris introduced a little bit of excerpts of those books, and super excited to be raffling that off today. Chris, you know, next I think a question for you is if if you were to picture, a a very tough problem statement and and kinda deliver it using, a typical modern data architecture pattern, what would that look like? A tough pattern. I think, again, I kinda come back to the the toughest projects we've done have to involve with organizations that have not kept up with change. And so I think some of the the most challenging projects out there are folks that are still coming off of mainframes, who were used to real time access, who, you know, are moving all of their data into spreadsheets, ripping that culture apart, and being able to do it in a progressive yet gentle way, I think, really takes a lot of talent and tact, to both get there from a talent technology standpoint. Right? There are so many things that we've gotta think about from how do we impact those systems, how do we gateway that from maybe an on prem into the cloud? How do we structure that data? Does anyone remember how the table structure of this thing still works? Right? Like, organizations that haven't had those reps of doing a lot of change enablement, that initial lift off feels really hard. But again, I think those are the most rewarding to be able to deliver that transformational, piece of insight. We've done that for some clients, and I can think about this, you know, again, I'll anonymize this, but some of the most proud projects we've done is help things, you know, CFOs go from where, yeah, it's taken me to extract this data out of my mainframe, and it took me two and a half weeks to close the books. And I don't even have time to do cash flow analysis to Yes. I have it on my phone. Right? And and I'm on the beach, and I can sit there and ask Microsoft Copilot and Microsoft Fabric, you know, what's, what's my cash flow analysis look like right now? You know, we often communicate in cycles. Right? What's my cycle time for cash? We now have models that can do that and it didn't take a giant ERP project to do that. It took connecting those legacy systems that maybe we're not ready to transform and bring them into this modern world, and it's a great way to start showing the value of that modernization effort. That that that's a powerful story. Thanks for sharing, Chris. I I think that that pattern, the very difficult pattern of somebody using these, you know, legacy traditional tables to generate a report that's as critical as driving the financials of the company, to being, see to using it on a phone, on a beach is is really transformative. And and making decisions on cash flow, which by the way is so important in this day and age is every CFO needs to know what the cash on hand looks like, what the cash flow analysis looks like. We do have a question in the panel, in from Adam. In terms of modern data architecture rules, is there any space for a data data modeler rule? Anyone wants to take that? No. I'll take that. I'm begging you. Please have a data modeler. If if you want your project to fail measurably, don't have a good data modeler in there. Did I was a DBA for many years and the worst cases of performance I've seen where somebody didn't know how to model data, especially when you get into star schema on that. So there's there's that modeler who's gonna make sure that the solution you're building in there has the right relationships between the table, the data's in the right tables. And it's it's a hard skill in there. There there's a lot of ways to model data on that. So make if you want your project to be successful, you have to have a solid data modeler, not just for building out the the table, say, in the data lake, but then the person who's doing the reporting. And say you're using Power BI, have somebody who's really good at building semantic models out in there because the report writers are used to be the one that did the modeling and the report writing, and that's kind of gone away because they realize company realize you need to have monitors who are really good at that. And and they will create the model that the report writers will then use, but if they create a bad model, the performance is gonna be horrible, data's gonna be wrong on that. I've seen that in too many use cases. So in the end, data model role, it not only should you have a number of data modelers, but make sure that you get somebody who's really good at it. I think, you know, to piggyback off of that, James, that is the critical spot to make AI work as well. So, we think about Copilot, the Copilot experience inside of Microsoft Fabric. You're not going to get value out of that if you have all of your key fields still available. If someone hasn't taken that extra step in that Power BI semantic model to hide their keys, to make the labels of their fields human readable that align to the semantics or the words and the nomenclature that the business uses. If you have duplicate relationships, right? I've seen models where product is in two different tables. You ask Copilot, what was sales by this product month over month? It's gonna give you an answer. It just might not give you the right one. So you're gonna really need that critical data modeler. Okay. The modeler. And make sure to bring those semantics that are available to that semantic engine. Couldn't agree more. Again, thanks so much. These are wonderful insights, and keep those questions coming. We will leave the recording. After this, webinar, we you'll have information about the recording, and some key links to those resources, and you can post back questions even post the webinar. We're here to answer those. Thank you so much again, James, Chaitra, for joining us today, and I hope you have a wonderful day. Thank you, Chaitra. And thank you, James, again for, doing this inspiring session with us. And, Chaitra, thank you for being a a wonderful emcee today. I hope everybody has a great rest of their week. Take care. Take care. Bye.