Name: Building the Business Case for Self-Service CoEs: Value, ROI, & Scalable Growth
Uploaded: 2025-07-23T16:30:46.289Z
Duration: 50 min 12 s
Description: Building the Business Case for Self-Service CoEs: Value, ROI, & Scalable Growth

Transcript for "Building the Business Case for Self-Service CoEs: Value, ROI, & Scalable Growth": And we're live. Good morning, everyone. Great to see you all joining us today, and welcome to the final session in our building the next gen COE webinar series. A couple of housekeeping notes before we begin. Please feel free to use the chat to share your thoughts or reactions. We always love hearing from you. If you have any questions, please drop them in the q and a panel. We'll tackle some as we go, but we'll also leave some time at the end as well. And, yes, we are recording the session. We'll send out the recording and slides afterwards, after this webinar meeting. So no need to take notes unless you absolutely want to. So over the last five sessions, we've explored how COEs support self-service, onboarding, governance, and AI adoption across the enterprise. But today, we're bringing it all together by tackling one key question, and that is how do you make the business case for the COE itself? As self-service programs mature and AI agents become part of the modern enterprise toolkit, COEs are increasingly being asked to prove their value, meaning to show how the work they do drives measurable outcomes, aligns with strategy, and delivers long term ROI. And that's exactly what today's session is all about. Today, we're excited to be joined by John Tudor, director of business architecture at Dataiku, and also Claire Gubian, VP of business transformation at Dataiku. So without further ado, let's jump in. I will pass it over to John. Awesome. Thanks, Renata. Good morning, good evening, good afternoon, everyone. Thank you for joining us for our sixth and final session. So today we're focusing, as Renata said, on really how do we think about value as it relates to self-service programs and COEs. And what we have here today is we'll discuss about both terminology around how you think you can measure around these programs, both in terms of starting them, but also keeping them going over time. So the first thing we wanna start out with is really this quote. And it's if you can't measure something, you can't understand it. And if you can't understand it, you can't control it. And if you can't control it, you can't improve it. So really what you'll see today is a lot of focus on metrics, on KPIs, and how we measure. Because if we don't measure something, we really can't do something to improve it or make it better. So how do we use data to really improve ourselves, improve what we're doing in terms of a program perspective? So the first thing we'll talk about here is this idea that, you know, what are some of the major value KPIs that organizations have around these programs and how they start to think about them as a scale. On the left side here, you'll see a standard approach that will recommend to many organizations for thinking about how to do this. So first, how do you define the priorities of the organization and tie those to those KPIs? What are the strategic imperatives of the organization? So what is in the mission statement? What are the goals of the company? What are the specific goals of the business unit? And what are the goals of the data analytics teams? From there, you wanna define the key stakeholders you wanna resonate with. Generally, these stakeholders can be across IT from a CIO leadership perspective within the business, all the way up to the CEO as well as VPs leading various divisions, and then your finance leadership across the organization who generally impacts your ROI analysis. You also wanna think about and brainstorm and really map KPIs into this. So what KPIs can be brought up for the group and implemented to help drive value and show these value to these key stakeholders? Further, you wanna make sure you look at the data collection, ideally doing it in an automated way, sometimes through surveys. But a way to pull that data together regularly is so you can review it on a regular basis and guide direction. And then finally, you wanna make sure you're important to communicate. What is the regular cadence to talk about these metrics and drive them throughout the organization? Now we generally see three kind of common value KPIs within companies. You'll see them here on the right side. First and foremost is return on investment. So probably not a surprise to anyone, but the goal of any type of investment from a business context is to get a better return on top of that. I'll spend a lot of time here today talking about ways to calculate that ROI and how we think about building that as an organization. Another key metric is this idea of active users and the number of users who are active within the software. For me personally, this is the metric I've always looked at because ultimately people vote with their clicks. So the goal is to ensure you have as many active users as possible within your data products across the organization as well as using our self-service tools such as data. The higher that active user account, the better you're doing as a team to help support those users and show back to the organization. Finally, there's this idea of business process adoption. So as we all know, when we're releasing data analytic products into the business, it doesn't necessarily mean those products are adopted or utilized. And if they're not, that means you're not driving the value that was intended for the product. So you're really killing about tracking your values and organization. It's also ensuring that you see business process adoption happening. And this can happen through looking at the unique number of users of the product per month or other metrics associated with that. So to carry us forward here, I'm gonna pass the ball to Claire, and we'll start to dig in a little bit more into how we calculate some of these items. Thanks, John. So yes. Indeed. Let's go into the first one, the return on investment, and how to calculate that one. Return on investment, so you definitely want to look at benefits versus costs. Here, we have detailed how to think about the benefits side of the house or value gains. And we get this question a lot from our customers, which is how how can I, showcase the value of a self-service COE? We're gonna deliver many use cases. We're going to empower the organization. How can I translate that effectively in a dollar amount? This led us to creating the value framework, the Dataiku value framework, which, shows the four value pillars breaks it down in four value pillars. Three of them are our platform value gains or how a self-service COE helps deliver operational excellence. And the fourth pillar is our use case value gains, how a self-service COE helps build valuable business applications. If we take a look at the operational excellence pillars, the first one is speed and agility. We all know it. Speed is the name of the game these days, and allowing your data people to do to go faster when they're building data products is super compelling. Helping them save time on the manual tasks, on the repeat tasks is, very valuable as it helps gain insights much faster, and faster insights, makes better decision faster decisions. This is also about enabling new users to deliver insights, meaning that you can get more insights. And what the the last piece about speed, is the fact that if you can deliver a product faster, you might do more of them, and you might be okay to fail, faster and succeed faster. So you're basically unleashing more innovation for your organization. The second pillar is stack efficiency. And here we look typically at what's the cost to deliver one project and what's the total cost of ownership of, having a self-service COE. And here we include the people aspect of it, the tech aspect of it. And here, when you're thinking about it, what are the different tools you wanna leverage, and, how are you going to power it, and, what type of training and what type of enablement you're going to be requiring, your teams to to have. Here, typically, a a platform like Dataiku that helps you cover the the end to end, data product life cycle from data ingestion to model building to model monitoring, from agent, creation to agent orchestration. You are may making sure that you are optimizing your cost, optimizing your IT spends, and also, making the most out of your cloud investments. The third pillar is control. So we saw how we can unleash more analytics at a faster pace, but we also wanna do it in a way that is controlled and governed with the right guardrails in terms of costs and in terms of, of access, who has, who can build projects, who can access the data. In the self-service world, right, this is something you really want to do well. Why? Because you wanna avoid potential fines, potential, security breaches that can cost a lot of money to your organization. So that's the third pillar. And then, the the fourth pillar, which is the use case value gains piece, so business applications. Self-service COEs, I've seen many of them, and I've been always very impressed about the amount of dollar that they generate for their companies. Some huge projects generating millions, some smaller projects generating in the hundreds of of thousands or the tens of thousands. But, typically, you are, unleashing, projects that will help increase revenue such as next best action, next best offer projects that will help save costs, such as predictive maintenance projects, reduce risks, such as ESG projects, etcetera. So if we go to the next slide, let's take a, a closer look at the time savings piece. Right? So we saw the speed and agility, pillar is the number one value of a self-service COE powered by Dataiku. And here, on this, again, thinking about how to frame this in a way that can be easily explained to people who are less in the weeds of what a self-service COE does or what, a date how a data product life cycle works, we've broken down a data product life cycle in five key steps. Data ingestion and data access, data wrangling and analysis, model training evaluation and validation, then how you deploy this model, and then eventually how you will, monitor, the evolution of the model. And, and for each one of these, steps, we look at what drives time savings. In the case of data ingestion, it's, the fact of having everything in one place. People don't need to be looking for the data and also there might just be data that they didn't know about. Making, datasets that have been very useful for a team available to the rest in a very easy way is also a great way to inspire more people to use these datasets and spend less time looking for them or make ensuring that they're clean and overall building trust in these datasets. The second one, the data wrangling, analysis part, this is often the the part where people lose the most time prepping the data, cleaning the data. And here, leveraging ways, best practices, notably thanks to Dataiku, to prep faster with, visual recipes where you don't necessarily have to know how to code to be able to do that, and also being able to do it in a way where every, people who are, stakeholders of the project, meaning the data experts, but the subject matter experts too can take a look at the project and really iterate and go faster as they can all collaborate on the single platform. On the third piece, the model training evaluation validation part, here again, taking, going faster by leveraging models that have already been built, being able to compare the performance of each of these different models and being able to analyze drift very easily helps save tremendous amount of time. On the deployment piece, having one, a way to not only design your model, but then to put it in production in a seamless way where you don't necessarily have to rewrite it or leverage third parties to be able to do it. And, and being able to then to deploy it for whatever makes the most sense for your organization, real time batch, again, helps save tremendous amount of time. And then last but not least, model monitoring, having, all motors models that are governed in one in one unique way with, similar guardrails enabling to look at potential drift and, also coming with the documentation that can be necessary for the regulators, in certain industries. All in all, these time savings equate from 70 to 90% versus traditional tools. So this is a very, compelling, value driver, and we will see how we can use this value driver to then translate it in a compelling business case, in a few minutes. But before we go there, the second piece where I want to highlight, is and if we can move to the next slide, is on the use case, side. So here, we get a lot a lot of the questions we get from self leaders of self-service COEs is which use cases should we be, prioritizing, how to qualify them, how to prioritize them. And, we always advocate to really have to put in place a strong methodology here and, eventually, even a review committee, to to go over them. But besides the beyond the process, if we take a look at this use case Canvas template that we put together for for our customers and if we break it down, the first big piece is starting with a business problem. The temptation and especially, these days is to do AI for the sake of AI. It sounds great, but we all know that AI is not, the the the the finality. Right? II is the mean. And so starting by understanding what is the business problem from the business stakeholders, why is it a problem, and whose problem is it is absolutely paramount. And then once you've understood the problem, what is the potential solution? And, is there a specific use case? Is this like a typical AI use case that we're talking about? Does this require a, a specific build? And then what type of solution, can be leveraged? Is it actually not really an AI use case, but more like a standard analytics use case? Do we need Gen AI? Do we need to build an agent for this? Right? Of course, the answer will vary greatly depending on the business problem. Then defining who are the stakeholders, who will build the use case, who is the executive sponsor, and who are the other potential stakeholders. Again, we we need to build trust here, and it's not just about building the model. It would also be about who will be consuming it and how it will be embedded in the business processes. And having these key stakeholders from the design phase will be a tremendous help in the success of the use case. Data. We need data to do, to do analytics and AI. So what what data is required? What are the data sources? What type of data, will will be used? Modeling. What time of algorithm is used? What, Dataiku features will be, leveraged? And then last but not least, consumption. Thinking about the last mile from the beginning. Who are the end consumers? Who will be the people actually consuming the insights or the actions of this agent? How many where are they located? How will they consume it? Will it be through an app? Will it through be through a dashboard? Will it be through something else? It's important to have that in mind when you're building the use case. And then looking at the value piece. Right? So this is the bottom row, which is what is the potential business value that you are estimating from this use case? And here, it's about defining a baseline. Right? So looking at where you're at today on your key, KPIs. So if you're thinking, for instance, about a churn reduction, you wanna understand what's your churn rate today, what's the impact it has on your revenue. And so you will be able to measure the, incremental value that you're delivering thanks to your churn prevention use case. And then you also want to understand what's your cost. How what what what are the resources you're gonna need in order to build this use case? And, usually, you end up with a nice little metrics where you have value business value generated versus cost and feasibility. And this will definitely help you see whether this use case is worthwhile doing or not if, benefits outweigh, the costs. In the cost piece, there is a in the the in the feasibility piece, there is also another very important aspect, which is the risk piece. What are the potential risks tied to this use case? And, in building this use case, how could they impact the use case, and how can you mitigate these risks? Sometimes that's a reason to not pursue a use case. And then it's not it's it's a good thing to have put a business case together to convince your stakeholders, but you wanna be able to track the value. You've set the baseline, but you do want to have a mechanism to track the value of your use cases so you can report back and potentially also kill your use case if it it it doesn't make sense once you've actually started developing it. So, next moving on to the next slide. Now on this one, we, have summarized how you can actually compile this in an ROI calculation. So we've talked about speed and agility, stack efficiency, control, and, and use cases. How do you actually put this in a business case in a way that is, simple enough to articulate? So first of all, you wanna build your business case over a few years. Right? And we typically we recommend the three years view. Why? Because you, will, gain in efficiency, and, you will be able to reuse data, reuse use cases over time, and your, your value, creation will be, in some way exponential. So on the speed and agility piece, we like to talk about adding headcount without hiring, which is another way of talking about productivity gains. And this is about looking at the time savings, applied to the number of users, their salary, and you can export a productivity gain from that. Stack efficiency is typically how you're saving on, licensing costs, maintenance costs, and infrastructure costs, tied to the fact of using maybe a reduced number of tools and maximizing your cloud investment. On the control side, it's about running a more efficient MLOps piece and the MLOps process and also avoiding potential fines and and security breach related cost. And then the last but not least, the more data products. So people could argue, okay. Great. I have a big better productivity of my data analyst and data scientists, and I I have all these new people who are now building, data products. But, so what? Well, if you're freeing up this time, you have time to do more projects. And if you're doing them faster, you will recognize the value faster. And here, either you actually have the value of these use cases, in which case you can input it there, Or what we often do, if that's not the case, we take an average value for a project, in this case, $250,000, and then we apply the time savings piece, which increases by the same amount, the number of projects. We take this and we put it in perspective with the costs. So typically, your tech cost, your people cost, your maintenance, and your training cost to extract the ROI. I'll pass it on to John now to walk us through the other, components of the value drivers. Awesome. Thank you, Claire. Appreciate it. Hey. I'll just get a quick pause. We do have a questionnaire from Sanraj, so I'll go ahead and answer that. The folks watching here, feel free to ask questions at any time. I'm willing to take them throughout the presentation. But to take this one, so Sanraj says, how will the digital transformation program managers ensure AI governance and risk management are embedded? So excellent question. That's not something we're gonna be focused on today. Today, we're focusing on value. That said, if you go back to the fourth webinar in the series, we actually focus very specifically on this idea. Right? How do you apply actual oversight and governance and really best practice standardization as part of your analytics processes. To cover that at a high level, we usually recommend a few different things in terms of overall observability as well as having strong architectural governance. So really making the easiest path also the right path for users as they work through your analytic and, machine learning workflow. But then, additionally, we also take a look at really, I'd say release governance and how do we manage releases of products to ensure they're aligning quickly with risk management. From a data perspective, that comes through our, govern product, which allows you to apply business process flow on top of release of analytics, as well as through our project assessment tool that allows you to really apply test driven development practices to your data products before they go to production. So in our mind, building out those process as effectively would help you drive more of that risk management and, governance management and oversight. But we highly recommend going back and watching the recording of their fourth webinar in the series, which directly focuses on that. But, excellent question. So with that, I'm gonna go ahead and jump back in here onto our slide content. And, really, the kind of next thought here is how do you start to demonstrate success? So taking what Claire just talked about, the idea of collecting strong use cases from across the business, showing those efficiency gains, showing the impact you're having. Right? Well, a step that once you started to collect that, once you start to show that to our eyes, how do you actually demonstrate that success to the organization? How do you go out and showcase the value of what you're bringing? And, really, this can lead to many outcomes as an organization. First and foremost, this is how you show the value of the investment. It's not just the calculation behind closed doors, it's just finances. This is something you share across the business. The idea of showing here so we can cause a change in how we operate in our culture as a business to be more data and AI driven. It also demonstrates here are the possible. We found that in many cases, when you have folks who have a similar background to someone who say is in engineering, talking with another engineer, and they share what they're able to do, that resonates a lot more stronger. Right? So they're able to show a much stronger or the possible for the individual. That's part of the value shouldn't have success. It can also drive additional user adoption. Right? As you see your peers have success and have good things happen with them, you also have a desire to go do the same thing. So, again, it can drive a bit more growth there. And, eventually, this is how you support additional investment and growth. So as you show their ROI, as you show the success of these use cases, you you position yourself well to say, hey. What is the next tier? How do I keep growing this out across the organization and company? Now in terms of highlighting wins, there's a few key elements. First and foremost is make sure you're focusing on those target personas. So if you work with a particular division of the company, look to show success within the terms of that division. Right? You wanna focus on what will resonate with that division. As an example, if you're working with supply chain, maybe you don't be highlighting sales and marketing use cases as a simple example. I think a big one, especially for those of us who come from more of an IT background like myself, is to make sure you frame as a business case. Right? Ultimately, you need to be speaking in the language of this so that what you say resonates in the terms of the business. Generally speaking, that's not a conversation around your data architecture, about what data lake you have or how you're handling your processing and what the ingestion flows look like. It's more of a conversation of, well, hey. This impacted our cycle time for delivery of these parts of the shop, or this impacted our ability to improve our tolerance levels and our engineering design. Right? So you need to make sure that you frame from a business perspective in a way that'll resonate with that group. And I think very closely tied to that and probably the most important element of the slide is it really should be the users speaking to their experience. As someone who's, you know, let's see who's for many, many years now, but a part of coaching our customers on this as well, the important part is that you want folks who actually did the work to speak to that work, and you want them to do it in a way that they're speaking in a heart to heart way with others who may be interested in doing that work. Having a centralized team say, hey. Here's all the great things we've done. Here's all the great work by your peers in the business. But not having those peers present that really doesn't have the same effect as having those peers directly sit down and present those directly to those other users. That will truly resonate and start to grab the other groundswell of interest around what you're doing. Again, we wanna make sure we have a line highlighted within the audience, so ensure it aligns with what they want to get out of the session. And for those of you who've been throughout this entire series, I'm going back to our first session. We talked about the idea of a branded mental model or the success model. This is a great time to plug this back in. You wanna make sure that you're aligning what you talk about from a use case perspective, what is done on those use cases back to that mental model. So if the users went through and did some data engineering and produced a data science output, you may wanna highlight, hey. This is where the users did the stitch and the science components of this project and how it ties back to the original kind of mental model for the program. That'll help drive consistency and communication. On the right side here, we'll talk about ways to really highlight wins. These are different ways to reach out to the organization depending on your org structure or team definition. This may change. But first is really roadshows with the business and IT. A A really effective way to do this is simply to go out and have these roadshows and have the individual users within those groups speak about what they're doing to others within those groups or to their IT counterparts. Similar to this, as you can get through organization or team meetings, All hands, staff meetings, these are all great places to come in, highlight that work, and maybe have a more discreet discussion if it's a smaller group and a team meeting or maybe a broader organization discussion in all hands. If your organization has stand ups, runs in a scrum methodology, you could also have collaborative stand ups where you share these ideas and discuss, hey, what's working, what's not. For those who've seen our previous sessions on more building community or session five, you talked about communication and newsletter communications. That's a big part of this as well. You should be looking to highlight these success stories in your newsletter communications, both to highlight the program, but also to give recognition to the great users who built the work. And then finally, you know, if you have ambassador meetings, again, going back to previous sessions, but this idea of ambassadors who are close, you know, friends of the program, folks who work from the business or maybe power users, you wanna make sure that you're also having these use cases to those potential users because then they can go back to the business and share further. So with that in mind, we'll go ahead and pass it sorry about there. I'll pass back to Claire here. What we're gonna talk about a bit more is a focus on how do you start to look at the qualitative facts, are there other elements related to how do we actually get this, you know, user adoption across the business. So, Claire, I'll pass back over to you. Thank you, John. And, actually, there's a question from Koichi, that ties, nicely to this one too. One non what non quantified benefits might you want to include in the business case in addition to quantifiable business benefits that drive calculable calculable ROI? So I thank you for asking this question. I I really love it. Indeed, we we do spend most of it our time trying to translate aiming to translate all the powerful substance that comes from this all serve civilian dollars, but there's also a lot of nonquantifiable business, benefits that we notice. And I think one of them is employee satisfaction. And, why do we highlight that? Because we know there's potentially a lot of churn that, can happen in the teams. And and with churn, there's knowledge a lot of knowledge that can go away. There's a lot of competition in this market. And so having people who are fulfilled with the tools that they're using and fulfilled with the way the org is working is actually a big deal. Right? So that's a great example, that is not as easy to quantify in ROI, but it definitely has a tremendous, indirect effect on on the organization. Now on this slide, we in, you know, in the ways of highlighting success and showing, the importance of, of the user, buy in, one very compelling way of doing so is, showcasing quotes from users, stakeholders that highlight why, they what, what the the value they're getting from the the self-service COE, what they're able to do. And beyond the number, it's really sharing the voice of the users and the voice of the stakeholders. And so we always recommend, including quotes and sharing a variety of seniority levels and, of teams. Right? You wanna be showing typically when you're driving an AI transformation that this is being embraced by many people in the organization and not necessarily, the people also that you are expecting, to you're you're expecting the most. Right? And so here in this example, as you can see, we have a head of supply chain that is sharing how, this has helped to accelerate the time to market of their AI projects and then lock in $500,000,000 in value over three years. But then we also have the CDO here talking about speed to delivery, talking about by far the most robust and scalable platform allowing teams to accelerate the build out of data pipelines from preparation to ML. Right? So these, sharing these these user feedback and this team leader and key stakeholder feedback is is is really a great way, to to talk about these maybe less quantifiable benefits. But, again, that show that the transformation is working. We often talk about, you know, the problem of of trust or people not wanting to change the way they do things. And, you know, you often have a visionary in a company that is seeing things 10 steps ahead, and you want to show the buy in. You wanna show how people are embracing this to inspire others. Right? And one great way to do that is quotes, but also having these people actually sharing testimonials. If we move on to the next slide, the the the another great way that I've seen, and this is this is one of our customer, Mercado Libre, who does this really, really well. So a big pillar of the AI transformation and a usually a big responsibility of a self-service COE is to help with data and AI literacy. Because, again, we all know that there's a huge value to be got from empowering the business, with the power of of AI agents and and data, in general. And, and so in the case of MercadoLibre, what they did really well and, is define key KPIs to measure how well they're doing on their data literacy, journey. And so they've set a a target of a percentage of people that they want to get to a certain point. And, and then they've also added, KPIs around frequency of use, last time used to make sure that there is continuity in the time in in in time. And this is, again, talking about the usage part that, John, was was, particularly, talking about. Right? It's one thing to drive business outcomes, ROI, but then really the usage part. Is this anecdotal still in the company, or is this really, is there really a wave of change and adoption happening? The other piece that I really like about the MercadoLibre example is that they have defined very specific training tracks where, people, can, embark themselves depending on their starting point and how far they wanna go. And, they can obtain the right certification, and they can move up in their data literacy journey. And each, track and each person is is is, I mean, is is tracked in order to measure the progress, that's that's being done. I'm passing it back to to John, to talk about our business process automation piece adoption piece. Sorry. Awesome. Thanks, Claire. We've got a few other questions here in the chat. So I'll go ahead and, hit those next. So one question we've got is, what should the c we focus on as for six to twelve months to lay the groundwork for measurable business impact? So actually, to answer that question, again, big visuals person, what I'll kinda come back to here is actually the slide that we shared here at the beginning. Right? This idea of these value KPIs. In my experience, it really comes down to these kinda key moments in the left, see if you can figure out what exactly you want to align terms with terms of value. As we're talking about here today, there's many different ways to calculate that. Right? ROI, active users, business process adoption, but also things like the qualitative benefits, the overall digital transformation of the company, just like Claire was talking about. There's many ways to look at this, and I think the best way to be successful is to have a very clear definition of what those priorities are, what the stakeholders are that you wanna make sure you're aligned to, and then what KPIs actually matter for that. Right? You have to get that alignment upfront to make sure you're speaking the language of those who are key stakeholders to make influence and make sure you have that impact. Now the other part of this, and I'll kind of point to the data collection element in the screen, or that I'm down the left, is that really automating that process and that capture and really getting a clear definition of value, I think, is very key. Something that we'll see a lot, and it's even personal on the slide. Right? This idea of getting KPIs at scale is that in a self-service environment, it's very difficult to capture this at a broad level across many, many people. So you need to think about ways you can build systems to capture that information. A common thing that we'll see customers do as an example is use, Dataiku to capture the value of projects as they go from more of that non project production state. They'll look for the capture that value before you operationalize or release the product. That's the one way to do it. There's also ways to do it in terms canvassing the business or tracking business process adoption, which we'll talk about here in a minute. But I think it's very important to get clarity on what those metrics are as well as align how you, calculate them. Claire, I'll also pass to you. Do you have any, additional thoughts on this question, before I jump to the next one? Thank you, John. I think the one that you mentioned is, yeah, the, like, the the one around alignment. And it goes back to what we were saying about the use case selection process, which is really you wanna be solving a business problem. And, and and and the other piece I would also highly recommend is identifying your quick wins, basically. Where are you gonna be able to showcase value, fast and well? I and then, sorry, this ties back to another question that's in the chat, which is what are, like, the pitfalls to to avoid? We what I often see is COEs embark in a journey, and they take too much time to show success. And we all know, stick execs, stakeholders are impatient, and there's a lot of pressure to deliver results on AI right now. And, you know, even our internal COEs, they get the question all the time. Okay. What's the value? What what what what success have you showcased? Right? And so you need to be able to buy time with your executives because it will take time in certain case. And so how do you do that? Well, what number one with what John said, which is basically aligning on the priorities and the problems you wanna solve, thanks to to AI. The you wanna first solve you wanna soul first. Sorry. And the second one is finding projects that show that demonstrate the value, the value of the upskilling part, the value of the transformation part. Right? And sharing those success stories. Success calls for success. So never be shy to share success. And when you share success, share it in a way that everyone can understand. That's my final word. Good quote for the session. Thank you, folks. That's a good one. So just to kinda read the other question that Claire answered. So what mistakes should COEs avoid when trying to build a value case too early without stakeholder alignment? I think Claire answered that very well. The last thing that I would say is, you should always have stakeholder alignment. I think you need to really dive in deep and look at what does a business really need. Again, going back to our first webinar of this series, we talked about deriving vision and strategy directly from the organization, from the feedback of those who are actually be using the technology and doing this work. I think the same thing applies here. You have to make sure that you're building your values, so you're focusing on what you're doing from the ground up from what's gonna really resonate with the business or what's gonna actually be help the business be successful. Right? So I think they all kind of tie together here, but the idea is that you make sure you really build it out from a people perspective and you hit that success early and use it to drive more success. So awesome questions. So So pulling us back into our, regular scheduled programming here. So tying to one of the metrics we just hit up on the previous slide. Right? This idea of business process adoption. It really ties down this idea of, you know, when we think about when the business is adopting a new data product, whether that's self-service or developed by data professionals, in order to realize the value of that, you need to make sure there's actually use of it. Right? And Dataiku is very strong in supporting this because of its web based interface and observability across the ecosystem. Right? So one way we see this happen with a lot of customers is use the proxy of, you know, number of unique users who, you know, consume a data product on a monthly basis and use as a proxy for when something is becoming more valuable or entrenched within the business process. I have seen other organizations use different metrics such as the quality or, you know, other machine learning models. Once it gets to a certain level of quality, they're comfortable moving into more productionized state. You may see different types of metrics that tie into this. But I'd say the most common is usually this kinda user adoption metric. Right? So when you think about this, the first thing to think about is, now how do you measure it? For those who've been around for this series, some of these will look familiar, but you wanna measure things like how many queries are happening against the tables in the ecosystem for the products. Now how many result queries or API calls go against your machine learning models? How many have used for visualizations? Right? Or how many problems against your LN environments or AI agents? All those are different ways of quantifying or looking at the utilization of something. Now importantly, you need to make sure users are not are really, using web based tools to do this, not desktop tools. If you don't think observability into what they're doing, you're not gonna see those activities happening, and that's much more akin to shadow IT. So you need to make sure you can actually see the activity. Also, you wanna make sure you have maybe more of a centralized data storage or a mesh environment. If you have distributed datasets across the company and you can't really manage or see those datasets, it's gonna be very hard to measure this. So you can make really that central pane of glass, which again was a focus of some of our earlier sessions, primarily webinar two. So one example you'll see here on the right side is this idea of an innovation funnel. Right? If I have many, many data products that are being built in organization by various, self-service teams, I wanna start measuring or categorizing the level of those data products. Right? What is their maturity? So as an example, a POC could be something that's not automated or within the design of Dataiku with, no five unique users or less per month. So we're just trying a new proof of concept. We're trying something new. Right? A pilot could be something that we've automated. We're now trying to embed this into the business process that's been scheduled in the Dataiku's automation node. And then we've got six to four eight nine unique users. Right? We're really trying to see, is this gonna get traction? Is this gonna impact people the way we want it to? And at least for me, the key element is really this idea of a product. Right? This is when you've got something automated and it's really embedded in the business process. My usual recommendation, again, this is organizational size dependent, is this idea we have 50 or more unique users per month. Once you get to this level, there's a pretty good chance that this will become core of the business process and probably replace previous business processes. So that's interesting from a conversation standpoint, but it also shows you this idea of, you know, value through adoption. We're announcing we are directly impacting the business process and realizing the value that we want to realize. And then finally, I think it's important to have this layer of a critical product. Right? Which is really the idea that an executive is stepping and saying, hey. This is key to the business. In particular, I find the supplies, and domains where maybe you don't have as many unique users, but the data product is really important. So as an example, think of finance. There's usually this, aspect of closing books on a quarterly basis, and maybe only five individuals look at that per quarter. However, they may be impacting hundreds of millions of dollars for that one data product these five individuals look at. So you need to be careful when you think about it and have situations you can say, hey. This is still critical. Even if the number of users isn't necessarily indicator of that. Now kind of going to the left bottom side here, there's also this idea of how do you manage this from a portfolio perspective. Right? As your data products get more impactful in the business, you wanna make sure you have the right level support and management around them. This is really where the idea of innovation funnel comes in. So the idea is that the more data product is used, the more effective it is. And you wanna make sure you have the data products that get to the best level of, reuse since you can drive more usability across the organization. Now importantly, on the bottom left, the best practice we usually recommend is say, okay. If your data product has over 50 unique users per month, let's go talk with the IT or senior leadership group over that particular division and say, let's disseminate what we're gonna do with this project next. Now do we keep it in self-service, potentially accept the risk of someone in self-service running that project themselves, which could represent a risk. Let's say that person gets sick. They're not in the office and run a project, etcetera. Do you move it in IT because you realize it's become a critical of the company, and now we're gonna take this new innovation from self-service and roll into a full blown IT product for the broader company. Do you combine with existing? Do you have multiple initiatives happening and it makes sense to take the best of breed of both and put them together? Or you just remove it entirely because you realize that even though the product is getting adoption, maybe it's not exactly where you want the organization to go. But net all, this is a really great way to look at your actual adoption from a business perspective and sure you see that realization. And importantly, it's also a great way to tie the work that happens in the self-service directly into your IT teams and leadership and drive more innovation, across the organization. So with that in mind, I think we're here at time. We've got a few other questions, at least, one other question over here on the left. For those here on the call, we'd encourage you to add additional questions. If you wanna see the other five, sessions of this webinar series, please scan the QR code here. You'll be able to access some from an on demand perspective. Both that in mind, let's answer our last question over here on the left. So after working with customers, do you have a by vertical listing of quick wins, DX and or x and data team should go after? I can answer that a little bit, but I'd be curious on Claire's thoughts on this as well. So, you know, it's kind of interesting. Right? I think it comes down to the, well, really the state of the company. Right? What is the company's current status when it comes to data? What I'll say is that for a lot of organizations, and especially those that may be larger or those that have, you know, been around a longer time, maybe have more employees, more complexity. What I tend to find is that the initial wins usually come from more the analytics side or data engineering. And that's because in the organization, you have many disparate datasets across the company, and there's actually a a data engineering problem. I need to harmonize these datasets if you have a single kind of view. So as an example, I have hundreds of shops around the world, but other shops are reporting different information. I have no way of comparing those shops or making decisions from a leadership perspective. So I need to go through and engineer my datasets so that I can see that similar pane of glass. And I think that tends to be the quickest lens because what you find in self-service is that those employees being empowered by Dataiku and by these technologies, they have a pretty good idea about what's not working, and what they wanna do is automate that and improve that process. So I find a lot of that, I will call it lower hanging fruit, so to speak, usually comes from this idea of how do we do some initial data engineering and solve some major problems that can be solved for a little bit of work in automation. Then again, over time, you may see more growth as it comes to machine learning, simple models that solve predictability and generally speaking, that's where I see a lot of the initial value, early on. Again, if an organization is more mature, if they're getting more into AI, then it might apply to more simple AI use cases. Maybe you start with a, you know, an l 11, a basic chatbot to improve a process that may be your low hanging fruit before you explore further into a agentic AI and fully automating processes through AI. Right? So there's different ways to look at it. But in my experience across many organizations, I do find those lower hanging data engineering use cases tend to be pretty prevalent. It can show value pretty quickly. Now Claire, I'll pass the ball to you. Anything you'd like to add to that or to the maybe an answer for it? Thanks, John. No. I think that's a great example, that you just, shared. And, and I definitely agree with it. I would say and, again, this is, you know, thinking about what I've seen be very successful with with COEs that that use Dataiku is they typically went after this identifying business processes that had a lot of manual work. And, I'm I'm thinking it'll be about a customer that improve their contract review flow. And so I would go after these business processes. Right? Internal business processes because you wanna try with internal processes first before going external. And look for these processes where there's a lot of manual steps, a lot of content review, and where you can, build some automation to it or, as, John was saying, some basic chatbots that can already help the chiefs greatly, and that's how you build your internal sponsors. Right? And then people will come with other internal business processes that they want to get improved. So I I would go after that. And then what I've also seen be very successful in COEs is, having COEs build the POCs, as John was saying, on on some use cases that might be more complex for the rest of the organization to do, which is why it's really good to have them done centrally. So they're proven, they're rock solid, and then they can be, spread out to the whole organization, factories, brands, depending on how, your your company is organized. Awesome. Awesome. Well, thank you, Claire. And I was gonna say, I think that's all the questions we have. So we'll be wrapping up here shortly. But I wanna say if there's any other questions, please add them. I want to thank Claire for being part of the session today. And then also since this is the last part of our webinar series, we wanna thank all of you that have been attending the series. It's been a fantastic, experience for me. I really enjoyed the questions and interaction. So hopefully, you found the series to, be helpful for you and something that you found to maybe ignite some thoughts in your mind, maybe drive some conversations in your own organizations. But if you're ever interested, we're happy to follow-up and talk more from a Dataiku perspective, and I'd love to continue these conversations. But thank you all for attending the series. We've thoroughly enjoyed it. With that, I think we'll be wrapping it up here. Anything else we'd like to add, Renata, from a Dataiku perspective before we wrap up? No, John. I think you wrapped it up beautifully. Just, of course, a huge thank you to you too and Claire for a thought look thoughtful look at how teams can articulate value, align with leadership priorities, and build the business case for the COE. I think you mentioned, for the audience, in case you missed any of the earlier sessions, you can catch up on the full series by scanning this QR code that'll take you straight to the webinars where you can fill all your information and get access to all of our replays. And if that's all the questions that we have today, I think that's it from us. Thank you everybody for joining in, and we hope to see you guys again soon. Other than that, enjoy the rest of your day, everybody. Thank you, everyone. Take it easy. Thank you. Thank you.