From Data to Dollars: Building Practical AI in Large Organizations

Welcome to the Lean AI Podcast, where we're flipping the AI conversation on its head by focusing on holistic strategies and tactics that drive AI adoption, rather than focusing solely on overcoming the technical challenges of AI. In every season two episode of the Lean AI Podcast, we talk with corporate AI leaders just like you, who've uncovered the secrets of driving successful adoption with far less wasted time and investment. Our guests challenge established views and offer disruptive perspectives, providing you with new, actionable insights. All right, Siddharth, welcome to the Lean AI Podcast. Thanks for having me, Ben. So let's just dive right into maybe a little bit of your experience in this field for the benefit of our listeners. Sure. I'm actually a computer science engineer by training and I started my career at Morgan Stanley, the investment bank. So I was building softwares, designing data pipelines, and working with all kinds of trading data. So it was a very hands on experience and introduction to the tech world, which gave me a good foundation early on. A few years in, I decided to go back to school to study business, mainly to better understand how tech connects to strategy and decision making. And since then I've been at Royal bank of Canada here in Toronto. And over the past decade, I've had a chance to work across different areas, starting in enterprise strategy, then moving into digital transformation, and most recently diving deep into data and AI. These days I am focused on building enterprise data products and leading AI initiatives, so basically figuring out how to turn ideas into real solutions and get them to the users. I've won few different hats along the way, which has helped me to see how tech, data and business all comes together. And I'm excited to share a bit of that with you today. Yeah, well, it's great to have experiences across all those different buckets. And so that's why I'm really excited to talk to you today. So let's jump right in. In our last discussion, you described what I think is a very useful Venn diagram for our listeners who are the corporate AI executives out there trying to build things that people actually want. So maybe you can describe that and kind of the gap that you mentioned to me. Oh, yeah. So I often like to describe this concept of AI success in organization using a simple Venn diagram because it's. It really helps paint a picture of what's going on inside, in. Inside a large organization. Yeah. So imagine two circles. One circle is made up of business leaders. You know, your P and L owners. They are the people who wake up in the morning and think, how do I sell more credit cards? You know, how do I reduce the cost of operations? You know, their focus is very outcome driven. So revenue growth, cost savings, you know, improving customer experience and, you know, those kind of things. Now in the other circle you've got the technologists and know, especially people working on AI and data. So these are the data scientists and the engineers and the architects who think in terms of structured data and unstructured data, you know, machine learning models, LLMs, and they are constantly trying to improve what data and AI has in their ecosystem. Right. Their focus is usually solution driven, so building better data sets, you know, improving data pipelines, increasing model accuracy and performance, you know, things of that nature. So the interesting part is these two groups don't always speak the same language, right? Business people are thinking $in impact while AI teams are thinking in data pipelines and model accuracy. Right. And that's where I believe that disconnect emerges. And this is the key, the real magic happens in the overlap, the small space where someone understand both sides, speaks both language and sort of connects the two world together. So that's like my idea of this Venn diagram explaining how things are in the organization. I love it. And then what's the gap that you mentioned earlier? Yeah, okay. So you know, in, in my experience there just aren't enough people who truly operate in the middle space, you know, who understand both the business side and the AI side. But for AI to be really successful, um, I think especially in corporate environment, you need people who can bridge that gap, right? Folks who can take the business run, translate it into something AI can solve and make sure that it actually delivers impact. So these are the people who understand both sides, like the data and AI pipeline, you know, especially how data is collected, cleaned up, you know, transformed and then used in different type of models, whether it's your basic classification model or something more advanced like LLM. So they are great problem solvers, but at the same time they understand how business is run, what are the priorities, how decisions are made in the business, and what actually are the outcomes that matter in business. And they can speak both language fluently, both tech and business. And I like to refer these individuals as AI product managers. For me, they are the glue between tech and business, making sure all that AI horsepower is actually solving the right problems. They are the ones who ensure AI initiatives stay lean, stay focused and are outcome driven. So whether it's like boosting revenue or you know, saving cost or just getting things better, they are the people who can make most impact in future in the AI space. I like that and I think one of the struggles that we were also talking about that's related to that is, you know, you've got some of these use cases that are really kind of out there and maybe maybe a little bit more of a moonshot and then you've got some, I don't know, some things that maybe are a little bit more kind of quicker wins. You'd mentioned in our last conversation, you know, kind of the mix between those two. I think that might be really interesting for our audience. Yeah, I think, you know, there is a reason to think and reason to believe that we need to focus on something which is, you know, really important, which we can deliver quickly. And this is how I like to think about it. You know, the pace of change in AI right now is just wild. You know, you've got companies like Google and Amazon and Apple cranking out these new AI features faster than ever. Stuff like Chat, GPT, Gemini and Apple Intelligence, you know, just constantly pushing them. And they are able to do it quickly because, well, they often are the ones who are building the core stuff to start with. But that gives them a big edge when it comes to speed and getting things to market and getting things to user. But I think the key is the ripple effect it has. You know, the ripple effect is that user expectations are now sky high across the board. You know, it doesn't matter what industry you are in. You know, people are already experiencing AI in their phones, in their browser every day. So now they expect that same level of intelligence and simplicity from every brand and, you know, every experience they interact with. And I think that puts a lot of pressure on the organizations, you know, no matter the industry, to keep up. Right. So you really can't afford to sit on AI ideas for years. And that's where I think the idea of understanding what is important comes into the forefront. What I've seen work well is essentially focusing on delivering small, meaningful AI features early and often. So, you know, to isolate what is important to you and then make sure that you kind of have a continuous delivery mechanism against it is important. It doesn't have to be groundbreaking on day one. Like for example, we recently built a call classification and prediction model for contact center agents to reduce handle time. We rolled it out to small pilot group and not only to test that it works well, but also to get feedback from the agents and, you know, buying from the business. So when I think about, you know, balancing between what you do, what you don't do, where you know, you might get the best benefit, I think this kind of approach of, you know, quick wins, help people buy in, show real value and keep the energy up and give you room to learn and it create. That's my perspective on, you know, how to manage this conundrum. That makes a lot of sense. And I think it's interesting the, you know, just talking to people in the EAI space and going to conferences, I'll say, well, they will say rather, yeah, we're going to, we're going to test the model and then, you know, if it, if it tests well, we're going to release this product. It's like, well, just test the model. Yes, but also test the experience, test the user experience, the customer experience. Do people actually know how to interact with this thing? Is it intuitive to them? Because I think you're right. The bar has been raised. There's proof points out there of how great it can be where it's seamless and you're not even really thinking about that. It's AI. And so I think that, you know, that's really solid advice for the, for the listener. So what other, what other pieces of advice could you give to, to our audience in terms of. If I talk about few things that I have seen, you know, in my time that has worked or not worked in this space. Among those, if I were to kind of talk about something simple but foundational that I've applied personally repeatedly in my own work, I've seen few things. One is, you know, you should lead with business or customer problem and not the technology. It means it's easy to get like as we are talking about Apple and we are talking about Amazon's and the AI features, it's easy to get swept up in the Hulk, right? So whether it's lln, whether it's generative AI or it's agentic AI, but if you start with the tech, you often end up building solutions that nobody really wants. You know, it's instead, I believe, start with a deep understanding of the business challenge you're trying to solve. You know, what's poke, what's inefficient, you know, where's the opportunity to create value? You know, that's where your AI effort should begin. The technology, I believe, is an enabler, it's not a goal. Right? So I think that's one of the key advice. I think the other key advice around thinking about this AI problems is prioritization. I think that is important in an organization. I believe you can't do everything, especially not in AI where resources are very limited and sticks are High, you know, so the key is to focus your time, funding and talent on a small number of AI projects that actually matter to your end users. You know, that focused attention is what gives you the best shot at building something impactful and you know, get it, getting it across the line as well into the hands of the users. So I think those are the two that comes in mind. There is one word which I talk to a lot in my experience with my teams, and that's around hiring the right people across the AI value chain. And that's the key part. Success in AI is not just about having great data scientists and engineers. You know, you need the whole team, you know, businessmes who really know the domain, AI product managers who understand whole ecosystem who can connect dots and solve problems, and data scientists and engineers who can scale and build, you know, so, and maybe most importantly on, you need people who can, are not just technologically strong, but also great communicators and collaborators. You know, AI is a team sport, so you need the right players and they need to work together. Well, that's my general advice to the teams I work with, with myself as well as, as part of, you know, what I have learned over time. And I have seen results coming out of these, you know, essentially few, few learnings. So initially in this episode you were talking about the Venn diagram of, you know, business leaders, P and L owners, and then also technologists. And the overlap is actually where there's a gap. Right. And then you specifically mentioned AI product managers there, I guess. And you just talked about the importance of finding the right team members, the right people. What makes a great AI product manager in your experience? Essentially they are accountable to ensure that AI initiatives translate into real business value. Right. So the honest lies with them to bridge the gap between technical capabilities and strategic priorities. You know, identifying high impact opportunities, shaping AI solutions and guiding them to successful execution here. So their success is measured not by the model accuracy alone, but by the tangible outcomes delivered to the business. So I think about AI product managers typically possess three to four core skill sets. First, they are strong problem solvers who can quickly identify the most important questions that need to be addressed, bubble them up and continuously keep everybody focused on those key questions and understand what outcomes are actually going to help answer those questions or what outcomes you're driving. Secondly, they are effective collaborators. They are able to work across functions and communicate fluently with diverse teams. And finally, they have this broad depth in breadth of understanding of both data and AI landscape as well as the business domain. Sort of enabling Them to make sound judgments and sort of inform decisions as they sort of build AI solutions, tie it back to the problem they were trying to solve, place those AI solutions in the right channels, help them get to the users, let the users adopt those solutions and try to get back the feedback so that you continuously kind of having this flywheel effect improving your solution over a period of time. And they are the glue who kind of manages all this and keeps it on track. Hi, this is Jonathan Burtfield, senior director at the Lean Startup company. If you're a corporate executive looking to drive broad adoption of the AI centric products you're developing with far less wasted time and investment, we invite you to join us for a free 45 minute one on one consultation where we'll help you understand key tactics for validating use cases early in the development journey, to identify the optimal sequence for rapidly driving to scale and to navigate the potholes that have tripped up other leaders in similar roles. Head over to leanstarter co contact AI to reserve your spot. You'll find this link in the show notes, but don't wait. Spaces fill up fast and we don't want you to miss out again. That's LeanStartup Co contact AI. Let's make successful incubation and scaling of AI centric products a reality for your team today. And I think many of our listeners are corporate AI executives. In other cases we just have corporate executives that are leaving some sort of AI transformation or they have a, an AI development group that's within their portfolio of businesses that they manage. I think this is a really important point for that second group of executives because you know, thinking about in a lot of cases, why isn't AI taking off for us? Oftentimes it's because you've got, you know, the, the business leaders, including yourself in the second group and then you've got the technologist. But there's not that, that overlap that you were, you were mentioning Siddharth. So that, that there's, there's the gap. Right. So you need to hire people and they're, they're hard to find. Right. And they get paid a lot these days. But they are out there who can really connect the business goals and you know, actual real development and those, that is the missing skill set in, in a lot of companies. Yeah, for sure. And I think it comes from the fact that people have spent time to learn both sides of the world over, over their career and really understand from ground up how business runs and decisions are made and from the ground up how basically Core data works and then stepping on to how then AI is built upon the data which works in the hands of users. So those come with time and come with experience and kind of spending time with, with the core people who do that type of work. And I want to correct myself. I think I said it's the missing skill set. I think the skill set is out there. Right. It's, it's very valuable. I think it's a missing role is what I should have said. Right. So that role isn't, isn't actually defined in a lot of companies and that's, that's a problem. It drives a lot of waste. Yes, it does. I would agree with you. And that's why I like to believe that this is the role that would be important in future as AI takes off and kind of becomes the focus of a lot of companies to increase their revenue, reduce their cost, improve engagement with clients. And this would be a role that would kind of help get us to success. It's interesting. The Venn diagram is a, is a helpful visual aid. Some of the, well, a lot of our clients at Lean Startup Co and even some of the guests on this podcast are kind of trying to fill that AI product manager role. And they've got either a history more on the technology side and they've kind of shifted into product management or they were product managers in the non AI space and they're trying to really get themselves up to speed on AI now. And so you've got I guess both, both types of AI product managers that are, that are out there trying to, trying to make an impact. So I wanted to pick up on the thread you mentioned prioritization and I'm, I'm all in on prioritization these days because there's so many companies that are just burning themselves out, working really hard and they're working on the wrong thing. So you know, this is this. We could take six hours to just talk about this one topic. But how do you prioritize what to do with AI at a company? I think it's, it's, it starts by looking at what are the two or three real business priorities that the business wants to move needle on in 12 to 18 months timeframe. And that's where it has to come top down from the leadership. Leaders like us within the organization have to make those tough decisions. And maybe it's just improving your call center efficiency or boosting developer productivity. You know, whatever it is, getting that clarity helps focus energy resources and funding where it matters. So I think it has to Be more of a conversation on what is important to our end users and how is the business landscape changing and what lies at the intersection of two that we need to definitely focus on and then understand where we could want to put our AI dollars where we can have most impact in those priorities. Especially diving deeper into the business processes underneath those priorities and seeing what really fits, not just for the sake of AI, but, but for the sake of making the right impact that you want to drive. So let's go one level deeper on that. Let's say that you, you know, you've, you've looked at the strategy, maybe even, you know, held some workshops where there's the stated strategy and also maybe the unstated strategy or constraints. And so you've kind of got that and you say, look, we've got these two to three themes that we're interested in developing some AI powered products or workflows. How do you then kind of go from there to what to actually assign to some teams to start testing or validating? Yeah, once I think you've gone through this tough process of going down from, you know, thousand things that you want to do as an organization to top three or four priorities, I think the key thing is to then build a team around those priorities. And, and I talked about this earlier, you need to have business domain experts who understand that business will help. And then you have to have people like we discussed in length the AI product managers who can, who understand that domain, they understand why that business decision was made to in the first place, how those outcomes can be driven in the marketplace and what the customer wants. But then you need to hire the right level of data engineers or data scientists and analysts. And I don't mean to say you need to hire from scratch, but you need to look around the organization and say who are the best people who we can draw on to kind of execute this particular project and bring those people together and give them the problem to solve and autonomy to make decisions while also have a lens on, you know, what is the end goal that they need to sort of achieve and have that governance structure on top. And if you give that time in the right framework, you would see magic happen in six to 18, six to 12 months, I would say. So, you know, key topics are then hired the team, give them the right reasons and the behind the prioritization of the business topics and then let them do their magic. I like it because I, I keep coming back to the Venn diagram. I guess you're, you're describing a, a, a cross functional team that's in that overlap right between the, the two circles of the Venn diagram. No, I want to point out, I think the whole Venn diagram is important. The business domain people comes from the one circle which has business teams. The technology people come from that circle which has technology teams. And the product manager comes from the overlap because they understand both. So essentially the cross functional teams are a combination of people coming from across that Venn diagram. But having the right skillset for that particular problem drawn from that entire Venn diagram is what makes the magic happen. And that cross functional team, it sounds like, you know, there's a, maybe a build, borrow, buy kind of decision to be made. Because if you've, you know, I don't know, let's say you've got a team of five, for example, all five of those don't need to be on the team that's dedicated to AI. By definition, depending on what the theme is, you might borrow a team member from, from, from the business who understands that domain or maybe even two. The other thing is you could buy a team member saying, look, we need expertise in this case, so we're going to hire a, you know, a 1099 or a consultant or a, you know, whatever it is, an external expert whose expertise we need right now. And so I think it's interesting, it's not just the team that you've built that reports up through, for example, you, it could also include, you know, other people that you can borrow from around the enterprise and also maybe even some external resources. The mix isn't, isn't set. It just depends on the skill set that you have, what skills you need and what business domain expertise you need. Again, because you're operating in that gap or the overlap between the two circles of the Venn diagram. Perfect. I think you laid out the secret recipe really well, Ben, there. And I guess, you know, we've seen, and I'm sure you've seen this as well, but we've seen all sorts of different permutations of that where sometimes it's all the core team, sometimes it's almost all borrowed and just one person from quote unquote, the AI team. And in other cases we've seen companies that just outsource the whole thing. That's it, that's pretty unique and probably not recommended in, in, in most cases, but it does have its, its appeal in a certain number of cases. So we've talked about a lot. Let's talk about the negative stuff now. So what, what are some pitfalls that you Would kind of call out as maybe watch outs, that maybe it seems like a good idea, but you've seen people get burned time and time again. So among many things that you would want to avoid, one pitfall if I have seen in the past, and I think it's important to call out because of its impact now, is jumping into AI because it's the buzzword of the moment, you know, and that's often this fear of missing out and nothing wrong about it, you know, but companies think, hey, everyone is doing AI, we better get on board. And the instant is to hire a few data scientists or spin up a project quickly. But I believe that knee jerk reaction usually doesn't pay off well. And we go back to the idea of, you know, know, what really matters is to understand how AI can support your actual business process. You know, where does it fit, what's the value it drives? And the reason is AI is not one single thing. You know, it's, it's a broad toolbox, you know, it's, it's like large language models might be perfect for tasks like summarizing documents or generating content, but if you're trying to predict customer churn or identify fraud, a machine learning model might be a great fit. So trick is in aligning the right AI capabilities to the right front. I think otherwise it's like, and we were talking about cars earlier, so it's like buying a high performance engine without knowing what kind of car you're building. You end up with a powerful piece of technology but no direction, no integration and so no results. So it's why it's important to start with the business need and then work your way backward to the right AI solution, not the other way around. And sometimes because of how AI is progressing and there is so much need to move faster, we do the other way around. We bring AI first and then think where, where it can fit, but it has to be the other way. I like that. Yeah, that's a great call out. What else, what other pitfalls would you recommend our, our listeners avoid? Um, I think there are fewer things which people sometimes struggle with and all the organizations have face those challenges around, you know, getting out of the POC state and scaling the solution itself. And I think there are a few learnings that I've had which I think I would, I would put in the same bucket of saying, you know, avoid those pitfalls. And in that sense, when I think about it, you, you got to go back to a core idea of you prioritize what is important first. Don't spread yourself thin. That's a pitfall I've seen happen where you focus your energy and funding on multiple AI and data initiatives and eventually, because you have spread your resources, your energy, your focus and the expertise of business domain, then eventually, after 18 to 24 months, none of those have made the impact you were looking for. So that's one thing. The other one is, you know, which is more technical in nature, is actually idea of this ecosystem, you know, in terms of when you're making your tech choices. And it's very important because it has long term implications. You know, it's tempting to just go out with the, you know, go in with the newest shiny tool or infrastructure choice that that's in the market, you know. But if your team is working on an isolated data or AI platform that doesn't play well with others in the organization, you'll end up rebuilding a lot of things from scratch. The goal should be to build on shared platforms, reuse existing data assets and models, you know, and just create that interportability between your teams. That way as your AI footprint grows, which is likely to grow in next five years, you're actually accelerating delivery because you are using and leveraging what already exists in your organization that they have spent time and effort and money on and have perfected it over time. So these, I like to think as things that you need to get right early on so that over a period of time you're not diverging, you're converging. So those are a few of the things that comes to mind in terms of pitfalls for me. No, those are great. Yeah. So another thing that I think is important from an adoption perspective is you, you know, taking care of our data. And it doesn't come intuitively when we think about adoption. But you know, think of this concept like, okay, you know, if, if you want to make a delicious dish, it all starts with ingredients, right? Fresh, high quality ingredients. And the same goes for AI, right? So you need clean, organized, trustworthy data. It's the foundation, right? You can build the most advanced models in the world, but if you don't, if your data is off, your AI would be off as well. You know, you and AI can get more powerful, but it's going to be always as good as the data it's fed. And so taking care of data is must for AI to produce results that actually people can use with confidence and that drives adoption. So I think the idea of adoption starting from your core pillar of data is kind of draws your attention to this whole flywheel effect that, you know, people like to use what is fulfills their need. The need is only going to be fulfilled when what you're presenting to them is they can use that confidence. And how can you generate that confidence? When you focus your attention on the core of it, which is the data itself. So you know, that's, that's one of the important things that I think about in my day to day work, you know, in terms of adoption, which people miss just because adoption is seen to be something that comes as go to market and you usually interact with end users in terms of, you know, increasing adoption. And that's, that's very important as well. The other thing about you know, helping the users, not just users as client per se, but the users within the organization to adopt AI is as important and that's where a lot of change management comes into picture. A little bit of training and guidance come into picture. But it all goes back to the idea that, you know, the, the better product you deliver to them, they're more likely to use it and they're more likely to be able to achieve their goals. And you know, it all starts with data. I love it. So we're getting really close to our time limit here, so we're going to have to go into our wrap up. So I always like to ask at the end of these, you know, what are the three pieces of advice that you would give your fellow corporate AI executives? We're really trying to build the right thing either for internal customers or external customers with a minimal amount of wasted time and investment. Yeah, I would bring us back to some of the topics we just discussed and I think just wrap it up with them and say, you know, you should first start with the right problem. You know, start with the business and customer problem and not just think about technology. One second is focus where it counts, which means prioritize ruthlessly. Think about few areas that is going to make most impact and then finally build the right team around those particular priorities. And when I say right team, hire the right people across the value chain of AI and that way you give AI a chance to be successful in your organization. Super powerful advice. Simple but hard to follow. So hey, I really appreciate your time today. It's been a pleasure speaking with you and hope to talk to you soon. Likewise, Ben, thank you for having me. Thank you so much. The Lean AI podcast is brought to you by the Lean Startup Company to find out more about our Lean AI approach to generating more AI wins with far less wasted time and investment, including our training workshops. Pilot programs and full implementation offerings. Visit us at leanstartup.co Search the Lean AI podcast in Apple Podcasts, Spotify or wherever you get your podcasts so you don't miss a future episode. On behalf of our entire team here at the Lean Startup Company, thank you for listening and sharing with your colleagues and friends. If you found this episode insightful.