Designing Data Strategy for a Data Driven Organization | Data Analytics | Community Webinar

Key Takeaways

really what attracts me often. But this is going to be a little bit of a different talk in that I'm going to talk about the things that offer you a leg up um to improve your odds of success and I'm going to categorize these in these three different domains with these foundations. Right? So organizational design, data design I'm going to speak very briefly about and then analytics um design as well. Okay. So where what's my story my let me provide you a little backstory of what motivated me um kind of with the the current writing that I'm doing and and some of the research that I've I've done. So 10 years ago, Dell bought a software company I was working with and um I was actually brought on board about the same time um of it of the acquisition. And uh so Dell, if you're not familiar, they're they're just a marketing Michael Dell and and the whole company, they're just a marketing machine really. I'd worked for Texas Instruments. We made we we competed with Dell. I really think that we probably had a better engineered product than Dell had with uh the laptop business, but uh they kicked our can. And they kicked our can because they were really good at marketing. They could spin spin um and and market. And so I mean Dell could even make oatmeal sound sexy, uh quite honestly. Um so that was the first place that I kind of heard this soup. They were into big data at the time. that was the spin-off term that was so popular and everything was going big big big and and that sort of thing and I I was thinking well you know small data is still important and by big data do you mean technology like Hadoop and Spark or do you mean big data in a different paradigm anyway there was just a lot of hype going on and and Dell was a master of of the the original big data hype and then I I've been affiliated with Northwestern University and SM MU and uh CUNI and so they all had programs and I was involved in in their analytics programs and then about 2015 roughly uh there's a little bit of variance there but for for all three schools they went from a master's in analytics to a master's in data science. Now there were some curriculum changes uh in that but a lot of it was just the name right because analytics was somewhat p um and data science was kind of the new term. So we see this shifting of terms and I'm going to talk more about that but I'm just kind of setting the story for for what got me involved and and interested me in this in this topic. Um started doing some research. I mean, when you hear AI, you think that's something that's new or, you know, relatively new, but actually AI started in the 1950s um um Dartmouth uh in in the 1950s. So, um and then personally, I I have a a master's in data mining. Um, and data mining was all the rage for a while and now and then that kind of spun off because data mining didn't, you know, it it got a little older and then people were thinking, well, that's kind of like data dredging and so machine learning kind of replaced that term. So, you really don't hear much about data mining. You hear more about machine learning and AI and data science these days. Um, then there was another one. Uh, you don't need statistics. All you need is this new thing called machine learning, right? Well, you know, most people don't realize you can't do any machine learning without statistics because the whole paradigm of statistics is taking a subset called a sample and using that sample to make inferences and create models for a future or existing population. So, you can't even do machine learning without statistics. So, this whole thing about statistics is dead. uh that kind of being a statistician that kind of ticked me off but anyway I uh I wanted to kind of dispel uh some of the rumor there as well. So, and you know there's a whole lot more here, right? So, BI, business intelligence, there's decision support, there's there's visualization, optimization, etc. Um, and there are differences in these terms, but most people don't realize the the subtle nuances um of the terms and everything. And so, we we're always looking for the killer app, the killer algorithm. Um, and there is none. And in fact, one of my favorite theorems is the no pre-launch theorem. It actually is in computer science, but it applies to machine learning and AI as well. Um, in that there is no uh super killer app. There's no button that you push and it solves every problem. It's all about knowledge. And that's the reason places like data science dojo come in because you know they they train and and uh teach people uh what's what essentially. All right. So this is kind of the So I we created a threebook series and it's it's all it's all analytics is is the main title and the reason for that is um I'll show you in a minute why uh we kind of came up with analytics is the overarching term but it applies to AI and data science machine learning etc. The first book is really about um demystifying this whole field of all these different terms and all these the soup that's involved there. And so yeah, so I just started really thinking about what I just went through with the last slide a lot of the different problems and and this this first book took about six years. Uh it was about six years in the making. Um, and you know, again, I've been around for a long time, so I've been thinking about the shifts and and that sort of thing and and did a lot of research. And so, and in the end, like I said, that the book title was was difficult. Um, but again, out of all the terms, I think analytics is probably the overarching best best term. And so, all three books of this series are it's analytics. And then the they have different nuances. The first book is about really the hype and demystifying that hype and then uh analytics literacy, right? Basic literacy for for all this stuff. So, we're about to get into um some of that hype, some of those terms. But let's let's just pause for just a second and and analyze why there is so much hype, right? So um I I don't know when the first time you heard we need to be data driven uh in your career. Um so maybe that's been a a while ago. Um for me it was 30 years ago. So we're still saying the same thing. Um we need to be more data driven. Along with that we need cleaner data. That's another thing that I've heard for 30 years. Um so what's happened during all that right? Um, I I remember a a professor at Baylor say, you know, whenever you read something, you should say, why is this person lying to me? Um, and I thought that was crazy, pretty cynical. Uh, but not really anymore. So, you know, and I'm not saying that any of these are necessarily good or bad. You know, they're they're reasons for it, but as objective, I mean, I I like to think of any data scientist as is someone that can be objective, right? So the whole idea behind data is, you know, it's it's non-emotional emotional it's, you know, it's it is what it is. And um hopefully it's it's factual. Um sometimes it's not, but uh it represents the best truth that we have. Let me just put it that way. So um again, none of these is good or bad, but it is. But we need to understand where some of the stuff is coming from. So again, so marketing hype, why uh why do you see so much marketing hype? Because you know, marketers have to have something new and sexy, right? So you don't sell data science, you now sell ultra data science or you sell instead of big data, you sell ultra big data or ultra AI or whatever. In other words, you have to change the terminology to make it sound like you're applying something new. You can separate yourself from from everyone else. So that's kind of the where the marketing hype comes from. The second thing is self-promotion, right? And you know, I'm doing a bit of self-promotion right here, right? So I'm with you and I hope you leave today with something meaningful, but you know, I'm I'm I know I'm biased uh because I have my view. I have my experience and that's what I'm sharing with you. So, I know that I'm not 100% objective. Um, I am a Beijian um as well. And so, one of my things that I love about Beijian uh there there's a Beijian saying that says the only objective person is the person um that knows that they are subjective. Let me say that again. The only objective person is the person that knows they are subjective. In other words, we know we have biases. Um, and that makes us more objective than we would be otherwise. And then media hype, obviously, you know, same thing, right? You can't sell old news, you have to sell uh new news and, you know, spectacular news. Um, so, and the same thing happens for for politicians, right? Um, even the politicians that say that they're public health experts but are really politicians, they're they're very biased. Um, so anyway, okay. So, looking at some of the soup that we have out there, and the reason that we titled this it's all analytics was because we were doing research. it it doesn't take you very long to find you know well over hundred analytics types out there. Um and so in the blue circles you'll see different types of analytics. Some of these are functional um like visual analytics or predictive analytics and then you have all the domain specific analytics like financial analytics or healthcare analytics. Um so but they're they're they're all different different types and then you have you know analytics that that use uh data science techniques, machine learning techniques, um AI, visual BI, u statistics, etc. So those those terms are as well. We put together well over 50 terms that we added to the book and we could have gone on and on and on. Um, but it seemed best to me that analytics was more encompassing and that's the reason for for the for the book title. So how what should we do? Um, well the first thing we should do is we should speak the same language. That's the reason data literacy is so important. And I love these two quotes and I'll read the first and you can go through the second one. But you know, imagine an organization where the marketing department speaks French, the product designers speak German, the analytics team speaks Spanish, and no one speaks a second language. That's essentially how a datadriven uh business function works when there is no data literacy. And that's Casey Panetta um from from Gartner. And the second quote is similar. It's by uh Valerie Logan um which I think it's the data lodge that Val Valerie's doing now and she's um trying to educate on on literacy as well. Uh so this this is really important right I mean how do you how do you function as a business if you don't have at least a basic understanding and that's my challenge I I think or I would challenge you to have some basic data literacy where people are speaking that's where again places like um data science dojo come in handy because they help you speak a common language um across different different use cases across different industries etc. So we need that and you know th this stat 2020 h you know 50% of organizations lack data literacy skills um I don't think that's probably moved very much it's probably still uh 50% today I don't think we've gone very far in the last two years looking at the other metrics that I'm very familiar with we are making slow changes um overall okay so Um, why is it important? Because it it's really too broad. Analytics, again, we we just talked about that quite a bit. depending on industry, the types of analytics that you're involved in, whether it be functional analytics where you're a practitioner in say predictive analytics um or prescriptive analytics or visual analytics versus, you know, a domain or indust industry where you're doing sports analytics or financial u etc. So, it's really pretty broad, but we need a basic um language that we we can all all speak. And number two, analytics is pervasive, right? It it it really hits every level um within or it needs to hit every level of the organization and it's becoming more and more so. So um and it's hitting not only professionals but even even consumers. Even consumers are becoming pretty pretty savvy on on analytics as well. Um, number three, as I stated before, everyone has a bias, right? Um, and so I' I've pretty much already said this already. I, you know, I I don't mean to be lying and lying might be too strong. I'm biased, right? I have my opinion. So, we just need to be be smart about that and be an intelligent consumer of information, right? whatever you read, whatever you see, you need to understand the biases that are coming at you. Um, and so data literacy really helps you with that because um, you understand and you can filter that information as it's coming across. And you know, the bottom line is you don't have to be an expert in everything, right? So you can really start off um, just trying to get basic literacy. everyone in the organization with some sort of basic literacy and don't panic, right? So, it it just takes a while. Um, and just just be patient and and try to do what you can we you can do. So, why is why is this hard? This this is a quote and I'll give you a minute um to look through this. This is a quote by Dr. John Cromwell. He is uh he wrote this actually as a Ford into one of our books. He's a brilliant surgeon. Um but he's also a great D uh data scientist and you it's really rare to find um physicians that are are great data scientists and we need more of them. But I love the quote uh many who need to know don't even know what they don't know. I thought that was very poignant and um something to point out. And so that's one of the things that makes it hard. And then then silos is another thing that makes it hard. We'll talk a little bit about organizational structure in a minute. But most organizations are way too siloed. They are you have the all the different functional units. You have marketing, finance, sales, logistics, um order entry, whatever, you know, depending on the organization. But the the cross functional units don't speak uh often. And so that makes things difficult as well. So a few things that have been brought out by Howard Dresnner. Um and so and he did this comes a lot from a Ford that he did in in one of our other books. And so he's a he's kind of the developer of this idea of information democracy. He's been doing this and he actually coined this term I think 30 years ago. And so information de democracy a word a world where everyone has timely relevant and actionable insights to carry out the task associated with the role and align them with the overarching strategy of the organization. what he's talking about is making sure that everyone has the right information at the point of action within their work, right? Uh and that's one of our goals, right? As data scientists, that's what we want to do is make sure that we send the right information in the right context in the right format to the right people wherever they might be within the organization. Um, and you can see there that only 43% of the people think that they're actually getting the right information for decision-making. Um, and only a third of organizations claim high or extremely high data literacy, right? So, and then we come full circle, right? When we look at the last bullet there, clouding things is constant barrage of technologies, techniques and buzzwords that bombard us each each and every day. So, um, really good points made by Howard Dresnner. The next point is leadership. And, you know, like I said, I' I've got a whole lot to say about everything here, but let me just say that leaders really only have two principal jobs. It it's really that's it. They develop a strategy, right? And the strategy is really kind of what should we do? And of course that's simplifying but you know it really is you know they set the market what what market you go out after um within your organization what the what the organization is about what what its charter is um what it should be doing and then how do you do it how do you get it done is actually um or I shouldn't even say necessarily how you get it done but make making sure you get it done is probably better when I talk about execution. So you you can't do that in isolation. So effective leaders make sure that they are very knowledgeable at and including every level of the organization. So you don't set strategy in isolation because if you set strategy in isolation then when it comes to executing that strategy you're going to fail. So you've got to include um the whole organization in setting that that strategy up and making sure that you set the right strategy not just a strategy and that goes hand inhand with the the quote here by Edison you know having a vision for what you want is not enough vision without execution is a hallucination so you know again you have to not do necessarily what you perfectly what you want to do but what can be organizational design. Um so the just kind of recapping here uh let's just pause. So one and two go hand in hand. Uh number one you need everybody to speak the same language. You know some people are going to speak at a more fundamental language or basic level uh language than others. Um but everybody needs to have some common core um terminology across the organization and that will be dependent upon each and every organization and then you need to design the organization for information democracy right you need to be able to figure out how you're going to enable everyone within their jobs with with anal the right analytics to get that job done better um to do a better job with the right information and that could be automated decisioning, that could be dashboards, that could be deep data science. Um but but again the whole organization speaking data and acting upon data is the is is the goal and that leads the organization up for success. And the the last point is, you know, we've been layering now, right? So kind of we started really kind of from more of the bottom and now we're moving to the top with leadership. Um but again leaders must engage the entire organization. I I've said it but I'll say it again. They should not create the strategy in isolation. Um if they do that they're going to fail. Um a a brilliant strategy is just worthless without execution. So uh I love this quote. If culture eats strategy for breakfast, then structure eats culture for dinner. And that's by Sophie McCall. So, um I'm I'm enamored very he's got some great great stuff out there. Um and the bottom line is there is a definite difference between culture is is it takes a long time to set up. uh you know structure is pretty lowhanging fruit and so you can start with structure and then you can move towards culture but let me just kind of go through an example of the difference and why it's so important. So you know if you're in a small organization um you can really garner an atmosphere of team and shared values. I mean if you know everybody wins together or they dies die together. So let's suppose you're as an example you're part of a biotech startup with 10 people and people work together because it's it's it's either everyone succeeds and becomes really wealthy or everyone fails and you're out of the job and so you're you're pretty much heavily invested you know in that um you know your your contribution is is vital to the whole so your your your stake is and the outcome of the entire enterprise and the enterprise must succeed um or you'll be out on the street. So you know contrast that to what happens if you're in a organization of 10,000 people right so first typically in in that size organization you got a lot of different layers uh you got staff department managers directors senior VPs vice presidents senior vice presidents executive vice presidents got these fancy titles that are created to make everyone feel more important but u it's really easy to get lost in a large organization and that's the reason all these layers and structures are are created and at the end of the day rank matters and it matters actually a great deal. So there are ways of um structuring the organization um that that can improve your odds of getting everyone to contribute maximally, right? And so Safi goes through um and he defines these things called artist and soldiers and and the different roles that they play. And so especially on as you grow in an organization, if you're working with a larger organization, you need to separate these these two role types out. And u I'll provide his homework. Um if you're more interested in that topic, you should uh pick up his book called Loonshots. It's a it's a great great books uh Safi Ball. Um and it even goes in with the second tidbit there that I wanted to share. Um you know celebrate your your failures as well as successes. Um and the quote is for successful AI project celebrate your graveyard by you know Sam D. Um, that was a quote and it it goes it it actually ties back into that artist versus soldiers because artists you allow to fail quite quite a lot. They take a lot of risks. Um, it's where you're creating and you have a lot more flexibility and a single failure doesn't mean everything versus your soldiers. If you're on a project timeline and you have to deliver, you need soldiers to manage that type of thing. So these these first two quotes kind of go um uh hand inand what else I was going to say there maybe it'll come back to me. Um so the the third point and you can look this up but I I thought this was interesting you know Google's quest they they they did a lot of research and they they did a lot of experimentation on developing uh the the perfect team. And so the these are the two bullet points that that really separated um great teams from not so great teams. The great teams had people um that felt free and willing to to talk up and they also felt like they were uh listened to that that leadership believe that they had something to say and it mattered. And then the last point is back to that silos thing that I was talking about earlier. Um, you really in any organization, I mean it's kind of hard to in a biotech of 10 people, but if you get any size, you should really consider some sort of matrix structure within the organization as far as structure. Um, a center of excellence, a COE is is a good idea. Um, I've been on COE's that were uh very effective. And then the other is a cop which is a community of practice and um and those can be quite quite beneficial as well. So we've kind of gone through here. Um so I'm just kind of repeating here and again you don't see a lot of technology so far. I'm going to talk a little bit about technology in the next next few minutes, but um the interesting thing is technology is a lot easier quite honestly than these organizational things because technology selections are cut and dry. Um and they're just and because they're cut and dry, they're more objective, they're just easier. It's the squishier, softer stuff that that's harder and and that often gets overlooked and that's the reason that we need more of it. So, I'm going to break down and and talk about data for just a a slide or two. You know, you probably heard this data is the new oil, but really, you know, data itself is just really a cost. It's a liability. It's it's it's um you know, it takes money to source and clean up data. It takes money to keep it. It takes, you know, money to secure it. Um anybody heard of a data breach? Uh that's a liability, right? when you got data sitting around. Um so literally companies are spending millions of dollars um on this and then if you look at the the second bullet point there right after spending billion actually billions of dollars um 90% of that data is never even analyzed. So it sits there um maybe rolls off at some point but a lot of money is spent for that. Um, and so how do you turn a corner? Again, back to data literacy. You can monetize that data, which is a subject we won't go into. Um, certainly create and deploy analytics is something that probably is um on everyone's mind on this on this call. So, and then again, you know, technology is important, but knowledge trumps technology every time. And so again, that's one of the important points of places like u data science dojo um is helping you um with that with that knowledge. So why is data so important? So um it's really the cornerstone of improvement. You can't do any any sort of improvement without data, right? Um and I could go into this for a long time, but I won't. Um because we're kind of running out of times. I just have a few more slides, but you know, you can't do any sort of improvement unless you know where you are and where you want to go, right? And whether you're you're in the day if you're at day zero and then day one, if day one is better than day zero, then you know you're improving. If not, you're staying the same or or getting worse. So data is the the cornerstone of improvement. And it's because data measures process, right? Any sort of data point you have pretty much measures the process. And life is full of processes. I would almost argue that it's difficult to label something that's not a process. Everything from getting up out of bed in the morning to brushing your teeth to working out to, you know, going to work, checking your email, everything. Every single one of those is a process. And you can capture data on almost every single one of those things. And so, um, the problem with with with data is that, you know, a lot of it's not collected. Um we need to collect the the the most meaningful data for the problems that we want to solve. Um you know and we there is opportunity to be more data driven. Um certainly and we have to collect the right data because the everybody's seen the data explosion where we're we have these um extabites [clears throat] now that are being collected um uh of data but are we collecting the right data? Um and I would argue that often we we don't. So data considerations the way I kind of break this out in in the writing um is that you know from a very high level consideration you have these pi you have pipes right so you have pipes that carry the data and those could be um you know part of the movement of data so you could have you know APIs application program interfaces you could have events you could have micros services you could have streaming um etc and so that's all about the data movement and then you park the data at some point, right? And and those are in the data stores and you know the data warehouses, enterprise data warehouse, um you know the big big data, the data lakes, the the cloud data, the reservoirs, etc. So that's kind of where you you store. Then you kind of have a hybrid if you've run across data virtualization. Um and so again, I'm not going very deep. We have a whole section in in one of the books about this um this particular topic and and you know best practice on on how to do that. And then you know there's um curated data um where you can buy data um you know and then you should always as a data scientist you should always be thinking about next generation data uh well and that's really twofold. Number one, you're solving problems today. What data problems do or what data do you need in hand to solve today's problems? But you should also be thinking about what data am I going to need tomorrow to solve tomorrow's problems and what technologies are coming down the pike to help me um grab that data. I mean, you know, again, some of the the these pipes that have become available make it so much easier now than it used to be um to to grab to grab data. Um so that that's a piece. And then in the end, you really need good data governance and data management programs, right? So you need to be able to make sure that that data is of the right quality. Um and you need, you know, to make sure it's ethically done. um what what you're storing um all the security pro protocols that you're not storing data that you shouldn't have. So there's the the liability from that standpoint and then just keeping the data uh up to date uh with good data management and then the analytics design for success, you know, having all the right stack and tools and everything for what your business is doing. uh you know from the data to discovery and acquisition, the visualization, all the exploratory data analysis, you know, the uh all the stats, the reporting, um preparing the data, getting the data in the right format, amputation, whatever it might be. Um there and then feature engineering, finding those features relevant to the problem that you're trying to solve. Um and then obviously model build and and model selection then the evaluation and testing of that model um deployment and then once you put into production you need that that model into production then you need model monitoring and governance and of course this is very circular and it should be long-term planned right back to that Netflix thing you know you need to be sure that you can deploy that model you you need some sort of business case to go with all of this this full flow because if you don't, you could end up with a model that um engineering cannot deploy or it's too expensive to deploy um and does not return the um uh the ROI that your model would suggest that it that it would return. And of course you have the even the technical debt um you know that refers to um the long-term cost associated with the models that you create um and the software engineering um you know the model dependencies uh the data dependencies and then back to you know within the last few years really growing piece is the ethical and legal um impli implications, right? So, um everything in healthcare from like HIPPA to uh GDPR, uh etc. making sure that you're you're within the the the realm there. And so, this last slide is essentially um kind of I've talked a lot about the the different stuff that we've written about. You'll see it's all analytics. Um parts one and two on top. Those two books are out. Um the executive guide which I refer to a lot on the um creating the strategy there. There's a lot of information there. We have a healthcare book coming out later. But if you want to connect with me, there's a few ways that you can do that. and they're listed here on LinkedIn or uh my website or you can email me or if you want interested in one of these books, you can just type in my name. It's very simple, Scott Burke uh on Amazon or u I've got a a channel with some content you might be interested on YouTube. So anyway, happy to support in any way I can. Perfect. All right, thanks Scott. Um, I am going to share my screen really quick and talk about tomorrow's webinar before we answer some questions. And if you do have questions, please post them in Q&A. Um, and I'm going to give you time to type those out now. Um, so one minute while I share my screen and Scott, if you just let me know when you can see this so that I know everybody can see it. Got it. Perfect. So I'm seeing I'm seeing the data science. Yep. Yep. Perfect. So, next week we have uh Neil Lizer with us. He's going to be t taking us through an introduction to convolutional neural networks with TensorFlow. And so, we're going to be talking about how they work mathematically. He's going to give us an intro to TensorFlow. And then we're going to build our own model for image classification um using TensorFlow. So, make sure if that's something that's interested you you to you uh tomorrow, July 20th, tomorrow. Yep. at 10 am Pacific. Uh we'll be doing that. And Neil is a data scientist at um I'm not sure if it's IWAKa or IWCA. I hope it's IWAKA because that would be a really fun name to say over and over again. But they are a fintech startup and and he's a data scientist there and Neil also hosts the AI stories podcast where he invites um data professionals and tech leaders talk about their careers and how they use AI in their day-to-day lives to impact our world. So, um, make sure to join us tomorrow for that. And, uh, Scott, I am seeing one question. Um, so feel free to, um, I'll read it out loud to you, but if you want to, uh, pull up your last slide again so that people can know how to connect with you and all that while we do this. Um, that'd be great. So, I am seeing one uh, one question from LS. Um, and Scott, just let me know when you're ready and I can I don't I don't want you to have like your focus on two two things. Sure. Sure. Yeah. Yeah. Go ahead. Um, so where does the data design happen? Is it in marketing or in data science or in some other um department within the organization? Sure. Yeah. So best practice probably would be depending upon the organization. If you're doing if you have a project team um you know it's it or a program that's really if you're supporting a project you want the data design to be there right so if you're familiar with like crisp DM like uh uh it's it's a methodology for machine learning but part of it is is it's a circular design and it's it's getting the right data to support a project. Now, if it's if you're trying to get your data design right for the overall organization, um then that's a mixture of the business and it um with certainly with um Yeah. Yeah. that well overall the the the business, right? So, um you want to make sure to that point of Dresnner that you want all the the data to be available at the point of action. um you that's what you want. But does that help at all? I mean, I'm not sure if I'm answering the question in the right way or not. I think so. I mean, I think, you know, it kind of depends on the organization that you're working with really. Um yeah, but yeah, I think that's it's a big it's a big it's a big question. Data design. Um I I would say that there's fundamental data design. That's where you're you're in you're you're determining what your technology is going to be. I mentioned data virtualization um which is kind of a hybrid of the data warehouse um with all the APIs and all the events streaming in right so you don't have to have a lot a physical data warehouse it's actually where you create more of a logical data warehouse um that's data virtualization um or whether you want to have well and you can supplement an existing data warehouse to a data virtualization it doesn't have to be one or the other but you know a lot lot of depends upon budget There's a lot of factors. So, yeah, it's a little little difficult. Yeah, for sure. And that's the only question I'm seeing uh for you, Scott. Um I know somebody in the chat was asking where can they find the recording. Um that'll be uh it'll be on our YouTube channel. So, if you get subscribed to our YouTube channel, it'll be up there. Um and then if you did RSVP on our website, um we'll make sure to send uh the recording to you via email. And um it'll also be on our web page on online.datascience.com. So there are kind of three places that I guess you can find it and I'm sure our social media guy um will uh we'll we'll share it on across our social social media channels. Um and then I'm getting another question in chat. This is from Maru. uh what is your experience and how do you deal with the lack of focus of teams in data science and machine learning when they work on customer value? I think that's okay. Yeah, I was say you may want to pull up the chat and uh and and that one reading because I think there's two parts to this question. All right. Okay. Um now which one is this one would be one of the most recent ability of ability of okay ability of these AMI teams to work as a team. Um so really one of the I I mentioned very briefly the uh the the center of excellent and community of practice right so I I'm a strong believer in both of those. There's a difference between those two. So the the community of excellence is where you essentially have a core just like you'd have like it you'd have a core analytics AI data science team right so you'd have different people maybe you'd have some people that were great at visual BI you'd have some statisticians you might have some data scientists you may have you know other other adjunct people there so you you create a core and they're matrixed though they're matrixed with the business so you have these teams working with say marketing marketing gets time across that centralized team just like an IT department typically is. So it's kind of that model. Um so so that is a center of excellence. And then there's a community of practice which exists kind of outside the day job a little bit. I mean you should be given time um by the the organization should give you time to to work as part of the community of practice. um and that is a crossf functional and cross rolebased. So I was I was the leader for example I was a leader of a healthcare committee of of practice and we had people across different roles right so we had people from across the organization some of those people were business people some of those people were more technical people and they would come together and we would talk about best practices and so it was a great way for the technologist to lead to learn from the um the the the business and vice versa. So that was one of the the things um there. So um so the majority of AI and decision support teams rejecting agile philosophies. Uh yeah um that to me is going to be a leadership problem, right? a management problem because uh it sounds like I mean or or at least my experience there is that there's not enough investment there's not enough um weight coming in from the top uh for that. So yeah, I would go to my if that's if that's it, I would I would very diplomatically go through through my boss and my leadership chain to see and address the the problem there. Um because it's not something that typically unless it's you know personalities or something which everybody can work at. Uh it typically has to go up the chain before it uh that can be resolved. Um, all right. Oh, uh, I'm I'm reading as we're going through. So, I see that it's part of it again, it's it it's part of an education problem, too, right? Um, on what agile is. Yeah. Okay. And I'm not seeing any more questions, so I think that will do it for us today. So, thank you so much, Scott, for being here and taking time out of your schedule to um do this presentation for us. And thank you for everyone who joined whether you're in Zoom or on one of our live streams. Uh we appreciate you being here with us today. And if you are able to make sure to join us tomorrow for that introduction to convolutional neural networks with TensorFlow. All right. Thank you everyone. Hope you have a good rest of your