/ Insights / View Recording: Why You Shouldn’t Invest in GenAI Insights View Recording: Why You Shouldn’t Invest in GenAI October 17, 2024Join us for an insightful webinar titled “Why You Shouldn’t Invest in GenAI” as part of the AI Symposium 2024. This session will delve into the complexities and challenges associated with investing in Generative AI technologies. Our expert speaker, Brandon Dey, Head of Engineering at Concurrency, will guide you through the pitfalls and misconceptions surrounding GenAI investments. He will discuss the unclear ROI, production workloads, use cases, and implementation complexities that often accompany such investments.Learn why traditional tech selection frameworks may not be sufficient and how industry differences can thwart general solutions. We will also address the talent gap and the importance of having a clear strategy for what to build, buy, and where to partner. By the end of the session, you will gain valuable insights that will help you stop thinking in terms of build vs. buy and instead focus on creating a path forward that is both practical and effective. Don’t miss this opportunity to gain a deeper understanding of the GenAI landscape and make informed decisions for your organization. Transcription Collapsed Transcription Expanded Brandon Dey Hey everybody. My name is Brandon. 0:0:12.52 –> 0:0:13.452 Brandon Dey Happy Friday, junior. 0:0:14.92 –> 0:0:19.332 Brandon Dey I’m excited to be here chatting about Gen. AI and whether or not you should invest in it. 0:0:21.182 –> 0:0:26.262 Brandon Dey I’m gonna jump right into the presentation and then folks can trickle in as they as they arrive. 0:0:27.982 –> 0:0:30.942 Brandon Dey The recording will be available. 0:0:30.942 –> 0:0:33.22 Brandon Dey I believe it’ll be made available to you guys afterwards. 0:0:34.582 –> 0:0:36.502 Brandon Dey So folks should be able to catch up on the whole thing. 0:0:38.942 –> 0:0:41.662 Brandon Dey All right. So appreciate you guys showing up. 0:0:42.222 –> 0:0:44.102 Brandon Dey My name is Brandon. 0:0:44.342 –> 0:0:46.462 Brandon Dey I am head of engineering at concurrency. 0:0:47.502 –> 0:0:57.742 Brandon Dey Really, I’m responsible for the the P&L of our group and then also on the, the more fun side of the the House, the technical design and implementation of the. 0:0:58.522 –> 0:1:16.962 Brandon Dey The software that we design and implement for for customers concurrency is a professional services org in the Microsoft ecosystem we build enterprise AI enabled applications to solve various problems. Typically for manufacturers, retailers, financial services, and healthcare. 0:1:16.962 –> 0:1:19.762 Brandon Dey But we’ve we’ve worked across other industries too. 0:1:20.642 –> 0:1:24.602 Brandon Dey My background and then we’ll get to more interesting things outside of myself. 0:1:25.382 –> 0:1:29.382 Brandon Dey Is applied math and software. I’ve been doing this for about 10 years. 0:1:30.272 –> 0:1:35.672 Brandon Dey Before concurrency, highlight AI at another vendor in the in the space called 3 cloud. 0:1:35.672 –> 0:1:40.832 Brandon Dey Before that was my LED machine learning at Publix, Florida based grocery store. 0:1:40.832 –> 0:1:49.912 Brandon Dey Before that was an R&D at Kroger and worked in financial services, doing algorithmic marketing and so on at Fisher Investments. 0:1:50.712 –> 0:1:56.392 Brandon Dey But we are here to talk about Gen. AI and why you shouldn’t invest in it. 0:1:57.422 –> 0:1:59.782 Brandon Dey And that’s a sort of provocative. 0:2:3.742 –> 0:2:17.382 Brandon Dey And I’m doing that because as you guys might have noticed, the data industry is filled with a lot of hype right now and often creating value with AI tech can feel like a scavenger hunt. 0:2:18.942 –> 0:2:24.622 Brandon Dey And that’s often perpetuated by vendors depicted as this little raccoon here. 0:2:26.302 –> 0:2:39.422 Brandon Dey Who? Who claim to be experts in all kinds of stuff just in the interest of expanding market share and so on. I think that can be really problematic from it can be problematic if you’re a serious engineering executive or lead. 0:2:40.532 –> 0:2:47.412 Brandon Dey In it, who just wants a very frank discussion on what you can do with Gen. AI and what you can’t do. 0:2:49.822 –> 0:2:54.262 Brandon Dey Everybody’s familiar with this hype cycle and. 0:2:56.22 –> 0:3:8.342 Brandon Dey We’re we’re definitely riding the wave with generative AI, and that’s not necessarily bad if you can cut through the hype, and that’s really what I wanted to do today. 0:3:10.222 –> 0:3:11.422 Brandon Dey Because if you don’t. 0:3:12.572 –> 0:3:15.172 Brandon Dey You’re left with really all these problems. 0:3:15.972 –> 0:3:25.372 Brandon Dey Which are you don’t know what the ROI is or could be with with Gen. AI you don’t know what the production workloads could be. 0:3:25.372 –> 0:3:27.172 Brandon Dey You don’t know where you should use it. 0:3:27.612 –> 0:3:31.652 Brandon Dey You don’t know how hard it would be to implement should you decide to do it. 0:3:33.222 –> 0:3:42.102 Brandon Dey And that can be really frustrating, especially since there aren’t known solutions to that with new technology. 0:3:42.562 –> 0:3:42.962 Brandon Dey Right. 0:3:44.782 –> 0:3:51.582 Brandon Dey How you typically go about making technology choices with with the various frameworks that exist usually don’t help. 0:3:53.462 –> 0:3:57.502 Brandon Dey Different industries have different ways of thinking about how they invest in technology. 0:3:59.182 –> 0:4:5.902 Brandon Dey And and importantly, there’s a talent gap, right? LL Ms. have been around for a while, but popularly for the last two years. 0:4:5.902 –> 0:4:11.262 Brandon Dey And you that means you have two years of of runway for people to have. 0:4:12.12 –> 0:4:14.132 Brandon Dey Developed expertise in it and. 0:4:14.342 –> 0:4:21.982 Brandon Dey You know, mastery of a complicated domain like generative AI takes a long time. 0:4:22.22 –> 0:4:27.782 Brandon Dey And there’s a dearth of talent, basically, that that can actually deliver this stuff. 0:4:27.782 –> 0:4:30.62 Brandon Dey So all of these issues really. 0:4:31.612 –> 0:4:41.452 Brandon Dey Converge on why? I think you shouldn’t invest in generative AI, but if I were to say that more specifically, it’s not. 0:4:42.492 –> 0:4:43.812 Brandon Dey It’s not that you shouldn’t. 0:4:43.812 –> 0:4:48.932 Brandon Dey It’s. It’s what you should build, what you should buy and where you should partner, right? 0:4:50.542 –> 0:4:51.982 Brandon Dey Out of the gate, I think people. 0:4:53.822 –> 0:5:1.262 Brandon Dey Feel in my experience, clients have articulated that they feel like they need to invest in this technology, lest they be left behind. 0:5:1.542 –> 0:5:4.822 Brandon Dey And there is a a spirit of truth to that. 0:5:5.22 –> 0:5:9.662 Brandon Dey But then there’s a good deal of noise that could really lead you lead you astray, and so. 0:5:11.212 –> 0:5:12.52 Brandon Dey I’ve tried to. 0:5:14.382 –> 0:5:16.942 Brandon Dey Wanna develop a framework to help? 0:5:16.942 –> 0:5:30.222 Brandon Dey Think about that’s really important decision on on where to build and whether that’s internally or with a partner and what you should just buy off the shelf if at all and so. 0:5:31.772 –> 0:5:33.772 Brandon Dey Today my goal really is threefold. 0:5:33.772 –> 0:5:35.212 Brandon Dey I want you guys to learn something useful. 0:5:36.932 –> 0:5:45.532 Brandon Dey I hope that you stop thinking about build versus buy and hopefully you have a path forward wherever you are. 0:5:47.762 –> 0:5:50.2 Brandon Dey And so we’ll get, we’ll get rolling. 0:5:50.682 –> 0:5:52.922 Brandon Dey Feel free to leave questions in the chat. 0:5:53.642 –> 0:5:55.602 Brandon Dey I believe that should be turned on. 0:5:58.322 –> 0:5:59.602 Brandon Dey But let’s let’s dive in. 0:5:59.602 –> 0:6:3.482 Brandon Dey So remember this this scavenger hunter. 0:6:5.52 –> 0:6:16.372 Brandon Dey This AI vendor basically trying to help organizations create value with AI, but everybody’s an expert. You can do everything all the time, always with AI. 0:6:17.252 –> 0:6:27.692 Brandon Dey That makes it really hard to navigate the landscape and that’s why we need a map to help us help us do this. And so I’m gonna unpack. 0:6:29.372 –> 0:6:29.732 Brandon Dey This map. 0:6:30.892 –> 0:6:43.852 Brandon Dey Because it’s, it’s how I reason through technology choices and it’s how I help our clients reason through technology choices, specifically around generative AI and other AI applications, right? 0:6:43.852 –> 0:6:54.652 Brandon Dey Whether it’s demand planning, your forecasting things or next best action systems where you’re trying to make intelligent moves in the face of constraints. 0:6:55.492 –> 0:7:1.612 Brandon Dey Or other sort of NLP applications. Regardless, this map helps you reason through technology choices. 0:7:2.492 –> 0:7:4.852 Brandon Dey Across the board and so. 0:7:6.412 –> 0:7:8.572 Brandon Dey This is really what what the map looks like. 0:7:8.572 –> 0:7:13.372 Brandon Dey This is there’s a lot going on on this picture, and I’m gonna unpack it in the following slides, but. 0:7:15.132 –> 0:7:16.332 Brandon Dey Importantly, you have two axes. 0:7:18.12 –> 0:7:19.852 Brandon Dey That you can basically. 0:7:21.912 –> 0:7:26.192 Brandon Dey Use to characterize technology. And so where you put it on the map. 0:7:27.732 –> 0:7:27.972 Brandon Dey Basically. 0:7:29.692 –> 0:7:32.452 Brandon Dey Lets you conclude what you should do should you build. 0:7:32.452 –> 0:7:32.892 Brandon Dey It should. 0:7:32.892 –> 0:7:33.292 Brandon Dey You buy it. 0:7:33.292 –> 0:7:38.92 Brandon Dey Should you partner with a A partner to build it? 0:7:38.252 –> 0:7:43.932 Brandon Dey And so on the bottom axis, the horizontal one you have the technology maturity. 0:7:43.932 –> 0:7:45.892 Brandon Dey Basically, how evolved is it? Is it? 0:7:46.492 –> 0:7:48.412 Brandon Dey Is it not even invented right? 0:7:48.412 –> 0:7:54.372 Brandon Dey That’s all the way down in the the vertex here all the way to the the far right, which would be. 0:7:54.992 –> 0:7:56.312 Brandon Dey This is a utility service, right? 0:7:56.312 –> 0:7:57.952 Brandon Dey It’s it’s been around forever. 0:7:59.492 –> 0:8:0.892 Brandon Dey There’s essentially no risk. 0:8:0.892 –> 0:8:1.652 Brandon Dey It’s very cheap. 0:8:1.652 –> 0:8:2.812 Brandon Dey It’s a commodity, right? 0:8:2.812 –> 0:8:11.812 Brandon Dey So this is technology maturity on the horizontal axis, on the vertical axis, you have the the visibility and the value chain that the technology is. 0:8:11.812 –> 0:8:16.332 Brandon Dey And so this is, you know, at the top you have very visible stuff, right. 0:8:16.332 –> 0:8:23.412 Brandon Dey Think of like a aui or an interface of some kind or something that touches users that they. 0:8:24.532 –> 0:8:25.172 Brandon Dey Can interact with, right? 0:8:25.252 –> 0:8:26.532 Brandon Dey So this would be any kind of. 0:8:28.262 –> 0:8:29.702 Brandon Dey Like basic application. 0:8:31.252 –> 0:8:35.292 Brandon Dey On the bottom here you have things that are invisible, right? 0:8:35.732 –> 0:8:39.132 Brandon Dey Basically very far removed from end users. 0:8:39.172 –> 0:8:40.332 Brandon Dey It’s not kind of. 0:8:40.332 –> 0:8:44.292 Brandon Dey It’s not clear what its utility is and. 0:8:45.892 –> 0:8:51.572 Brandon Dey In the space of the actual graph here or map rather you have different interpretations on. 0:8:53.292 –> 0:8:54.892 Brandon Dey How to think of that technology right? 0:8:54.892 –> 0:9:4.92 Brandon Dey And so if you were to draw a line horizontally across you, that would represent the sort of minimally accepted. 0:9:5.772 –> 0:9:5.812 Brandon Dey Return. 0:9:6.782 –> 0:9:8.822 Brandon Dey On a technology project that you need. 0:9:10.372 –> 0:9:11.612 Brandon Dey To justify investing in it, right? 0:9:11.612 –> 0:9:12.932 Brandon Dey That’s sort of like the minimum. 0:9:13.612 –> 0:9:17.332 Brandon Dey Now obviously you don’t wanna do things below that. 0:9:19.572 –> 0:9:23.772 Brandon Dey And that changes a lot, especially how, given how mature it is. 0:9:27.232 –> 0:9:28.512 Brandon Dey I’m gonna skip over. I’m gonna. 0:9:28.512 –> 0:9:30.72 Brandon Dey I’m gonna move away from this slide. 0:9:30.72 –> 0:9:31.712 Brandon Dey I I sense that it might be a little too abstract. 0:9:31.712 –> 0:9:33.792 Brandon Dey I’m gonna talk about an actual example here so. 0:9:35.372 –> 0:9:36.412 Brandon Dey Think of. 0:9:38.452 –> 0:9:46.372 Brandon Dey A manufacturer who is who builds widgets and they need to manage their supply chain that you know takes raw materials. 0:9:48.172 –> 0:9:48.492 Brandon Dey In they do. 0:9:48.492 –> 0:9:53.852 Brandon Dey Some manufacturing of them are on them and then they, you know, sell it, sell the output, right? 0:9:53.852 –> 0:9:59.332 Brandon Dey The widgets now they need to manage that supply chain, their inventory and so. 0:10:0.952 –> 0:10:3.592 Brandon Dey They need a demand planning system and. 0:10:5.132 –> 0:10:9.332 Brandon Dey They they want to address this demand planning system up to be sort of next generation. 0:10:9.332 –> 0:10:12.732 Brandon Dey So they kinda wrap it in a a large language model, right to to. 0:10:14.412 –> 0:10:20.692 Brandon Dey Allow interfacing with that system via a chat chatbot interface, right? 0:10:20.692 –> 0:10:27.772 Brandon Dey So you kinda have generative AI being married with sort of legacy AI in this system now. 0:10:30.262 –> 0:10:35.542 Brandon Dey The question is, should you build this internally? 0:10:36.22 –> 0:10:46.942 Brandon Dey Should you partner with somebody to to build it for you? Or should you just buy something off the shelf and this map is gonna walk us through how to answer those questions. 0:10:47.262 –> 0:10:50.582 Brandon Dey So in basically plotted on this graph. 0:10:52.132 –> 0:11:0.132 Brandon Dey Our system components that are needed in order to create this generative AI demand planning system you need. 0:11:0.802 –> 0:11:2.402 Brandon Dey Forecasts of the widgets, right? 0:11:2.402 –> 0:11:5.802 Brandon Dey How many widgets are am I going to need in the next week? 0:11:7.372 –> 0:11:15.292 Brandon Dey You need what’s called explainability, which is allowing users to understand why those forecasts are the forecasts. 0:11:15.292 –> 0:11:35.12 Brandon Dey Why did the system make those decisions? You need, you know, certain business logic, right to to filter to the right result set and just do regular processing, business, business logic processing. And then of course you need, you need data to go into it. That data requires ETL and. 0:11:36.142 –> 0:11:43.262 Brandon Dey The the the forecasting model that produces the forecasts that requires ML OPS, right? 0:11:43.262 –> 0:11:44.782 Brandon Dey Just automated governance. 0:11:45.462 –> 0:11:47.142 Brandon Dey All that requires a platform. 0:11:48.692 –> 0:11:51.212 Brandon Dey And you know, there’s and then there’s more data that you need, right? 0:11:51.212 –> 0:11:56.452 Brandon Dey So what’s highlighted in red are the the AI components of this now. 0:11:56.452 –> 0:11:57.572 Brandon Dey Very faintly. 0:11:59.292 –> 0:12:1.772 Brandon Dey And sort of masked is all of the other. 0:12:3.372 –> 0:12:7.12 Brandon Dey All the other components that are not AI, but I’m pointing these out. 0:12:7.532 –> 0:12:14.212 Brandon Dey Just highlight how they’re dependent on each other and where they’re at in the map or on the map. 0:12:14.212 –> 0:12:17.292 Brandon Dey So forecasts those are very visible. 0:12:18.852 –> 0:12:21.12 Brandon Dey Users interact with those forecasts. 0:12:22.732 –> 0:12:26.172 Brandon Dey Those forecasts are and and also. 0:12:27.812 –> 0:12:30.852 Brandon Dey They’re new, right? They are a forecast. 0:12:32.532 –> 0:12:35.812 Brandon Dey Are need to be created with every new system, right? 0:12:35.812 –> 0:12:36.772 Brandon Dey And so for that reason. 0:12:37.692 –> 0:12:39.732 Brandon Dey They’re a sort of low maturity. 0:12:41.292 –> 0:12:45.772 Brandon Dey Sort of component to this system and that produces some risk, right? 0:12:45.772 –> 0:12:55.612 Brandon Dey ‘Cause it’s you know, if there were a utility, it’s it’s less risky. It’s tried and true, but these forecasts you don’t know. You don’t know how good they’re gonna be until you until you make them. 0:12:57.372 –> 0:13:1.292 Brandon Dey Explain you don’t get those forecasts without explainability. 0:13:2.212 –> 0:13:8.492 Brandon Dey Explainability is less less visible, but it’s a little more mature business logic. 0:13:8.942 –> 0:13:11.502 Brandon Dey This is usually behind the wall. 0:13:13.132 –> 0:13:18.892 Brandon Dey But still influences the experience and the data that the users see. 0:13:19.732 –> 0:13:21.612 Brandon Dey And so it kind of slides down the visibility. 0:13:23.292 –> 0:13:28.212 Brandon Dey Access and also this is typically custom built right? 0:13:28.212 –> 0:13:33.772 Brandon Dey You can’t, like, go buy, you know, a business’s business logic sort of module or something. 0:13:35.772 –> 0:13:38.12 Brandon Dey The data there’s really. 0:13:39.112 –> 0:13:49.832 Brandon Dey There’s really five different data inputs to to any any demand planning system and that’s pricing data, historical demand data. 0:13:49.912 –> 0:13:52.952 Brandon Dey So think of how many widgets were purchased. 0:13:54.292 –> 0:13:57.452 Brandon Dey You know, historically that’s the thing you’re trying to predict. 0:13:57.452 –> 0:14:5.452 Brandon Dey You have taxonomy data where those widgets fit in the sort of product catalog you know. Is it. If your furniture manufacturer. 0:14:5.452 –> 0:14:9.372 Brandon Dey Is it a couch or is it a, you know, loveseat? Or is it a chair? 0:14:9.532 –> 0:14:10.12 Brandon Dey Is it a? 0:14:10.12 –> 0:14:11.772 Brandon Dey It is it a bed, that kind of stuff? 0:14:11.772 –> 0:14:18.252 Brandon Dey And then you have weather data and calendar data. Now each of these data sets occupy a different space on the map. 0:14:19.792 –> 0:14:27.232 Brandon Dey This data down here is relatively invisible, and it’s usually custom built. The taxonomy, historic data and price data because this is unique to. 0:14:28.162 –> 0:14:37.642 Brandon Dey An organization. And so the the code and the logic of what’s called the feature engineering required to make sense of this data is all. It’s all bespoke. 0:14:38.842 –> 0:14:41.402 Brandon Dey That’s why it’s in this custom built section. 0:14:42.952 –> 0:14:44.472 Brandon Dey Weather data and calendar data. 0:14:44.472 –> 0:14:48.632 Brandon Dey This is much, much more commoditized and it’s also more visible. 0:14:50.432 –> 0:15:3.472 Brandon Dey In my experience having having built and implemented these demand planning systems with like an LLM sort of interface, everybody wants to know how whether influences the demand for their business. Right. Imagine your. 0:15:4.42 –> 0:15:17.2 Brandon Dey Manufacturer, but you’re a restaurant chain and you wanna know how many people are going to show up to your your restaurants every hour for every day of the week that you’re open so that you can do things like schedule your staff. 0:15:18.552 –> 0:15:19.752 Brandon Dey And you know. 0:15:21.472 –> 0:15:27.912 Brandon Dey Stock your kitchens and so on. The weather drives consumer behavior and so it’s a very it’s a very visible thing. 0:15:28.552 –> 0:15:29.512 Brandon Dey That’s why it’s where it’s at. 0:15:29.512 –> 0:15:30.672 Brandon Dey But weather data is also. 0:15:32.592 –> 0:15:32.992 Brandon Dey Everywhere. 0:15:33.842 –> 0:15:38.522 Brandon Dey So and everybody always thinks of of weather being an important input to these systems. 0:15:38.522 –> 0:15:40.362 Brandon Dey And so it’s relatively commoditized, right? 0:15:40.362 –> 0:15:42.322 Brandon Dey So you can basically kind of just configure that. 0:15:43.952 –> 0:15:50.432 Brandon Dey Through a sort of publicly available API, from from, whether underground or or or elsewhere. 0:15:51.232 –> 0:15:52.872 Brandon Dey Calendar data is even more commoditized. 0:15:52.872 –> 0:15:58.912 Brandon Dey That’s why it’s, you know, pushed to the right a bit and it’s less visible. 0:16:1.372 –> 0:16:5.332 Brandon Dey OK, so I wanna highlight just a couple of things on why. 0:16:6.872 –> 0:16:7.832 Brandon Dey On how the map works, right? 0:16:7.832 –> 0:16:11.592 Brandon Dey So each of these components are placed. 0:16:13.272 –> 0:16:15.192 Brandon Dey And that shows two important things. 0:16:15.192 –> 0:16:25.32 Brandon Dey The first is position and the 2nd is is movement and the reason this is a map and not like a graph is because it shows movement. 0:16:26.712 –> 0:16:29.512 Brandon Dey And what I mean by movement is how each of these components. 0:16:31.232 –> 0:16:31.952 Brandon Dey Evolve over time. 0:16:32.582 –> 0:16:34.702 Brandon Dey And how they influence each other and so. 0:16:36.272 –> 0:16:43.592 Brandon Dey Due to natural forces in a competitive marketplace and supply and demand and competition and so on. 0:16:43.952 –> 0:16:47.552 Brandon Dey Each of these components tend to move to the right, right? 0:16:47.552 –> 0:16:50.592 Brandon Dey So things, once they’re invented over here on the left. 0:16:52.232 –> 0:16:55.352 Brandon Dey If they’re useful, they tend to get pushed into. 0:16:56.952 –> 0:17:0.592 Brandon Dey Basically this middle zone, which is more refined. This is kind of like a product space. 0:17:0.592 –> 0:17:3.432 Brandon Dey So think of your SAS market and then. 0:17:4.742 –> 0:17:11.142 Brandon Dey And then as they are used more, they get pushed more and more towards the utility and the prices drop and so on. 0:17:12.102 –> 0:17:19.22 Brandon Dey But these things are always moving from left to right at different paces, depending on specifics of the component. 0:17:20.632 –> 0:17:21.792 Brandon Dey Position really matters, right? 0:17:21.792 –> 0:17:27.232 Brandon Dey This is what we were talking about on the previous slide just with with where it is, where it is vertically and horizontally. 0:17:29.312 –> 0:17:33.712 Brandon Dey And it’s important to talk about the mechanics of this, because it lets us. 0:17:35.162 –> 0:17:39.722 Brandon Dey The question we care about ultimately, which is what should we build? 0:17:39.762 –> 0:17:41.2 Brandon Dey What do we configure? 0:17:41.2 –> 0:17:43.642 Brandon Dey What do we buy off the shelf and? 0:17:45.152 –> 0:17:47.712 Brandon Dey That’s why I’ve created this graph to to kind of highlight those zones. 0:17:49.432 –> 0:18:0.72 Brandon Dey Basically, the further to the left and and it’s like altitude on the map, the more likely you are to to consider buying it. 0:18:0.432 –> 0:18:2.792 Brandon Dey The middle zone is really the configure or. 0:18:4.872 –> 0:18:6.392 Brandon Dey The configure space and then. 0:18:7.42 –> 0:18:13.322 Brandon Dey The further to the right you go, the more likely you you would be to to buy these things off the shelf because. 0:18:14.872 –> 0:18:15.872 Brandon Dey Their commodities. 0:18:15.872 –> 0:18:18.312 Brandon Dey There’s really no sense reinventing the wheel. 0:18:19.872 –> 0:18:20.952 Brandon Dey And so you should buy those. 0:18:23.682 –> 0:18:28.802 Brandon Dey OK, so one of the things I that’s important to talk about is why? 0:18:29.882 –> 0:18:33.642 Brandon Dey Why this system exists in the first place? 0:18:34.442 –> 0:18:35.402 Brandon Dey Why are we building? 0:18:36.952 –> 0:18:38.592 Brandon Dey This system could be any other system though. 0:18:38.592 –> 0:18:45.752 Brandon Dey Pick your favorite solution could be a, you know image processing tool. 0:18:45.752 –> 0:18:55.192 Brandon Dey It could be you know something that scans PDFs or receipts or something, and then you know, uploads it to some database so that. 0:18:55.722 –> 0:18:59.482 Brandon Dey A downstream process can pick it up and and do something useful with it. 0:18:59.682 –> 0:19:0.362 Brandon Dey It doesn’t matter. 0:19:0.362 –> 0:19:1.362 Brandon Dey But there is something. 0:19:2.912 –> 0:19:5.992 Brandon Dey It exists to do in this case with the demand planning. 0:19:7.672 –> 0:19:10.432 Brandon Dey System that’s to produce healthy working capital. 0:19:12.32 –> 0:19:18.32 Brandon Dey It’s not to produce good forecasts. It is to produce healthy cash flow, healthy working capital. 0:19:19.792 –> 0:19:24.712 Brandon Dey Manufacturers, restaurants, retailers with with complicated supply chains. 0:19:26.2 –> 0:19:33.402 Brandon Dey Invest a lot of their assets into their inventory and that locks it up right your it makes them less liquid. 0:19:33.722 –> 0:19:49.832 Brandon Dey And so the less you can invest in your inventory, obviously the more free cash flow you have, the more you can invest in other things and the, the the better off you are, so long as you’re not compromising the customer experience because you don’t have enough of what? 0:19:49.842 –> 0:19:50.842 Brandon Dey They want stocked. 0:19:51.632 –> 0:19:52.72 Brandon Dey And so on. 0:19:52.72 –> 0:19:53.872 Brandon Dey And so you put this at the top of the list. 0:19:53.952 –> 0:19:56.752 Brandon Dey Healthy working capital or you put it at the top of the map. Excuse me. 0:19:58.332 –> 0:20:2.92 Brandon Dey Now, in order to get there, you need all of these system components. 0:20:2.92 –> 0:20:13.52 Brandon Dey Now what’s highlighted on this. This map is all of the non AI components right in the first graph I showed all the AI components that were highlighted in this graph. 0:20:13.52 –> 0:20:14.452 Brandon Dey I’m just emphasizing the. 0:20:15.992 –> 0:20:20.952 Brandon Dey The non AI components right that are responsible for producing this healthy working capital. 0:20:21.592 –> 0:20:24.872 Brandon Dey The first thing that you need is the ability to to track. 0:20:26.592 –> 0:20:28.512 Brandon Dey And benchmark your current working capital. 0:20:30.102 –> 0:20:32.342 Brandon Dey To see if the system is actually useful, right? 0:20:32.622 –> 0:20:33.942 Brandon Dey And so in order to. 0:20:35.512 –> 0:20:39.232 Brandon Dey Do that. You need an ERP integration which itself needs business logic. 0:20:40.952 –> 0:20:41.272 Brandon Dey You need. 0:20:43.152 –> 0:20:47.752 Brandon Dey A. Basically a very simple crud UI or create, read, update, delete. 0:20:47.752 –> 0:20:51.112 Brandon Dey Just a very simple application that allows users to actually take action. 0:20:52.832 –> 0:20:54.352 Brandon Dey On in this case, inventory, right? 0:20:54.352 –> 0:20:55.592 Brandon Dey We need 7. 0:20:57.272 –> 0:20:58.952 Brandon Dey Widgets 3 weeks from now. 0:20:59.602 –> 0:21:2.2 Brandon Dey And so somebody needs to actually click a button to order. 0:21:3.992 –> 0:21:5.392 Brandon Dey Seven, let’s say couches. 0:21:5.712 –> 0:21:10.912 Brandon Dey So that in three weeks when consumers come in and and buy it, they can actually buy one of those seven. 0:21:11.72 –> 0:21:20.192 Brandon Dey So you need a a UI to actually change the inventory that itself needs ERP integration. ERP integration needs a platform. 0:21:20.192 –> 0:21:24.952 Brandon Dey A platform needs compute and and compute needs electricity or energy, right? 0:21:25.432 –> 0:21:26.832 Brandon Dey It might be funny to put. 0:21:28.432 –> 0:21:28.512 Brandon Dey Energy. 0:21:29.702 –> 0:21:45.102 Brandon Dey So, like laughably commoditized on this map, but it’s important because it illustrates the really what drives these technology choices, which is, are you going to build it? 0:21:45.102 –> 0:21:47.422 Brandon Dey Of course, nobody’s going to build their own. 0:21:48.992 –> 0:21:53.392 Brandon Dey Like generators to produce electricity to run their compute, to run their platform. 0:21:55.112 –> 0:21:57.432 Brandon Dey And similarly, you wouldn’t do that for compute. 0:21:57.432 –> 0:21:58.632 Brandon Dey You wouldn’t do that for the platform. 0:21:59.582 –> 0:22:1.462 Brandon Dey But you would do that for ETL, right? 0:22:1.462 –> 0:22:6.142 Brandon Dey And you would do that for your ERP integration. And so following this. 0:22:7.712 –> 0:22:15.952 Brandon Dey Sequence allows you to see more clearly and more explicitly what you could build versus. 0:22:17.552 –> 0:22:19.392 Brandon Dey Versus versus buyer configure so. 0:22:22.822 –> 0:22:26.22 Brandon Dey Let’s jump into. 0:22:28.152 –> 0:22:29.32 Brandon Dey This again. 0:22:30.872 –> 0:22:33.432 Brandon Dey Does anybody have any questions so far? 0:22:35.232 –> 0:22:38.232 Brandon Dey If so, feel free to dump them into the chat and I’ll. 0:22:39.952 –> 0:22:40.512 Brandon Dey Respond there. 0:22:42.592 –> 0:22:45.592 Brandon Dey This is kind of an overwhelming Mac. 0:22:45.592 –> 0:22:48.192 Brandon Dey Or it can be as soon as you understand the mechanics though. 0:22:49.872 –> 0:22:56.872 Brandon Dey It can be a really useful tool to figure out where you should invest, right and and importantly where you shouldn’t invest. 0:22:59.432 –> 0:23:4.592 Brandon Dey And so it’s still not all the way, though it doesn’t answer that question entirely. 0:23:4.792 –> 0:23:23.632 Brandon Dey And for that I’ve tried to codify how I think about this in in this chart, because it it I think it very succinctly distills how to arrive at those decisions based on a a map that you made. 0:23:24.432 –> 0:23:25.712 Brandon Dey And really, once you make the map. 0:23:27.672 –> 0:23:29.312 Brandon Dey I think you’re in a better position to. 0:23:30.572 –> 0:23:40.892 Brandon Dey The sort of reason either with yourself or with with your team, with other leaders, to figure out where you should put your investment, and so. 0:23:42.392 –> 0:23:43.672 Brandon Dey I’ll walk through this, but basically. 0:23:45.272 –> 0:23:47.72 Brandon Dey You have three options, right? 0:23:47.72 –> 0:23:54.192 Brandon Dey The first preference would be to buy things off the shelf if it satisfies some conditions and it’s aligned with your Technology Strategy. 0:23:54.352 –> 0:23:56.792 Brandon Dey But really, the preference if you don’t. 0:23:58.272 –> 0:23:58.392 Brandon Dey Umm. 0:24:0.2 –> 0:24:7.162 Brandon Dey If it’s not like super custom and it’s kinda just whatever technology but it’s needed, buy it off the shelf, it’s cheaper. 0:24:8.712 –> 0:24:13.432 Brandon Dey It’s it’s reliable and there are some constraints, but that’s preference. Number one is to buy off the shelf. 0:24:13.432 –> 0:24:15.672 Brandon Dey Preference two would be to build it internally. 0:24:17.392 –> 0:24:25.672 Brandon Dey For reasons related to ownership and control and the last preference would be to build with a partner if if certain conditions are true. 0:24:25.672 –> 0:24:30.112 Brandon Dey So those are the three three options you have with with. 0:24:30.922 –> 0:24:31.802 Brandon Dey Technology. 0:24:33.912 –> 0:24:36.552 Brandon Dey And let’s walk through how to get there though. 0:24:37.712 –> 0:24:41.232 Brandon Dey So the first thing that I like to guide clients through. 0:24:43.352 –> 0:24:51.352 Brandon Dey Is, is is this component I should say too that each this map is used at the component level, right? 0:24:51.352 –> 0:24:54.272 Brandon Dey So you know, if we take one of these. 0:24:56.32 –> 0:25:0.832 Brandon Dey One of these things, so let’s say like this this UI to change the inventory or. 0:25:1.742 –> 0:25:3.302 Brandon Dey That will take this UI to change the inventory. 0:25:4.832 –> 0:25:6.712 Brandon Dey Should you buy that off the shelf? 0:25:6.712 –> 0:25:8.352 Brandon Dey Build it internally or build it with a partner. 0:25:8.832 –> 0:25:18.512 Brandon Dey The first question to ask yourself in service of that is, is it a competitive advantage for you to have aui that allows you to change inventory? 0:25:20.312 –> 0:25:21.312 Brandon Dey The answer depends. 0:25:21.312 –> 0:25:22.392 Brandon Dey There’s no right or wrong answer. 0:25:22.392 –> 0:25:26.152 Brandon Dey It just depends on the kind of organization you are and and your relationship with technology. 0:25:27.832 –> 0:25:30.632 Brandon Dey For a lot of manufacturers, this can be a competitive advantage. 0:25:30.632 –> 0:25:35.232 Brandon Dey The supply chain and how you manage it, very important part of. 0:25:35.892 –> 0:25:48.332 Brandon Dey Profitability in that in that sector and so you could make a case that it is in fact a competitive advantage. And so you move, you move to the next decision point which is can can I build this in House. 0:25:48.332 –> 0:25:54.652 Brandon Dey Do I have the staff and skill set to build aui in house if you don’t? 0:25:56.192 –> 0:25:57.72 Brandon Dey Build with a partner. 0:25:58.672 –> 0:26:0.472 Brandon Dey Why versus buying off the shelf? 0:26:0.592 –> 0:26:7.712 Brandon Dey Because it’s a competitive advantage and you just can’t build it in house and so build with a partner. Now let’s say you can build it in house. 0:26:9.692 –> 0:26:17.52 Brandon Dey The next question to ask yourself is if if you need it fast, right, do do you have a very short time to market? 0:26:18.592 –> 0:26:22.992 Brandon Dey And if not and you have the skills in house to build it well, then build internally. 0:26:23.672 –> 0:26:33.352 Brandon Dey But if you do have a tight timeline, then move into the next decision point which is related to cost and how you prefer to structure your costs. 0:26:33.352 –> 0:26:38.832 Brandon Dey So do you prefer to save long term by investing upfront? 0:26:40.782 –> 0:26:43.782 Brandon Dey If not, build internally right? 0:26:44.862 –> 0:26:49.622 Brandon Dey What that means is it’s gonna be cheaper to build it in house, and it’s going to. 0:26:51.752 –> 0:26:55.32 Brandon Dey There’s like lower upfront capital costs because you’re not partnering with somebody. 0:26:56.632 –> 0:27:1.952 Brandon Dey Those folks are already on staff and so build it internally. However, if you do want to save. 0:27:4.192 –> 0:27:9.552 Brandon Dey By investing upfront and you have a culture of innovation that is, it’s important to be. 0:27:11.312 –> 0:27:14.752 Brandon Dey To be bringing to market innovative technology, right? 0:27:14.752 –> 0:27:19.352 Brandon Dey And this happens time and time again in, in our line of work. And what I see with with customers is. 0:27:20.2 –> 0:27:30.722 Brandon Dey Often the motivation for a project is we want to demonstrate to our customers, to our investors, to our stakeholders that we in fact can deliver on this innovative technology. 0:27:32.272 –> 0:27:35.72 Brandon Dey And so we’re going to partner with somebody to to help do that. 0:27:36.792 –> 0:27:39.992 Brandon Dey That doesn’t mean they don’t participate at all in the build. 0:27:39.992 –> 0:27:45.672 Brandon Dey They they can or they don’t need to, just depending on what they prefer. 0:27:46.152 –> 0:27:48.752 Brandon Dey But if if all of these questions are yes, right. 0:27:48.752 –> 0:27:57.272 Brandon Dey If you if you can build it in house but you need it fast and you prefer to save long term costs by investing upfront and you have a culture of innovation. 0:27:57.882 –> 0:27:58.762 Brandon Dey Build with a partner. 0:28:0.232 –> 0:28:2.232 Brandon Dey Let’s say you don’t have a culture of innovation though. 0:28:2.232 –> 0:28:4.592 Brandon Dey And you still prefer to save long term cost by upfront you. 0:28:4.592 –> 0:28:6.272 Brandon Dey Can you can buy it off the shelf? 0:28:9.392 –> 0:28:21.352 Brandon Dey If we kind of rewind and go back to this first decision point we started with, is it a competitive advantage? If it’s not a competitive advantage? 0:28:24.312 –> 0:28:26.792 Brandon Dey The two following decisions. 0:28:28.392 –> 0:28:28.432 Brandon Dey Are. 0:28:29.152 –> 0:28:35.352 Brandon Dey Basically, if you need customization and flexibility so it might not be a competitive advantage. But if it needs to be really flexible. 0:28:36.872 –> 0:28:37.872 Brandon Dey You know it’s gonna be. 0:28:37.872 –> 0:28:45.152 Brandon Dey You probably shouldn’t buy it off the shelf because they’re a little more rigid and constrained in what they can accommodate, and so. 0:28:46.872 –> 0:28:49.632 Brandon Dey You know, buy off the shelf, but if? 0:28:51.832 –> 0:29:2.912 Brandon Dey If you do need that customization and flexibility, the next question to ask yourself is if if it impacts the customer experience or something that you think of as being strategic, right. 0:29:5.432 –> 0:29:10.512 Brandon Dey That set of of activities that that you consider strategic depends. 0:29:12.72 –> 0:29:23.152 Brandon Dey Typically they affect the customer experience or things really core to what you do as a business. And if so, then you move up into this decision sequence of like can you build it in house? 0:29:24.952 –> 0:29:25.552 Brandon Dey And so on. 0:29:25.792 –> 0:29:27.752 Brandon Dey But if it doesn’t impact the customer experience. 0:29:29.472 –> 0:29:30.552 Brandon Dey You know, buy it. 0:29:30.552 –> 0:29:31.592 Brandon Dey Buy it off the shelf. 0:29:34.592 –> 0:29:40.312 Brandon Dey You could also build it internally depending on the the degree of customization that you need. 0:29:42.152 –> 0:29:44.232 Brandon Dey And really, the degree to which it impacts. 0:29:46.612 –> 0:29:49.332 Brandon Dey Your customer experience or or what you consider strategically. 0:29:49.932 –> 0:29:54.12 Brandon Dey So I know this is a lot, might be fire housing, you guys. 0:29:54.772 –> 0:29:57.812 Brandon Dey This is not a Tri it’s a, it’s not. 0:29:59.752 –> 0:30:4.72 Brandon Dey A like hyper prescriptive way to make these decisions. 0:30:4.112 –> 0:30:6.952 Brandon Dey It’s a great starting point though, and. 0:30:8.592 –> 0:30:9.752 Brandon Dey It’s useful. 0:30:11.812 –> 0:30:24.52 Brandon Dey It’s useful in in how it makes assumptions explicit and helps you, at least in my experience, helps folks see sort of how they’re typically making these really important decisions because. 0:30:26.192 –> 0:30:30.152 Brandon Dey As common as the question is, you know, do I build this? 0:30:30.152 –> 0:30:30.992 Brandon Dey Do I buy this? 0:30:30.992 –> 0:30:32.152 Brandon Dey Like should I invest in this? 0:30:32.152 –> 0:30:35.432 Brandon Dey Should I not invest in this as common as that is? 0:30:37.842 –> 0:30:51.722 Brandon Dey People often don’t have a reliable way of navigating that, and this is something that we’ve arrived at after a lot of iteration with a lot of companies building generative AI technologies. 0:30:53.232 –> 0:30:54.512 Brandon Dey Or and really implementing them. 0:30:56.552 –> 0:31:1.272 Brandon Dey To to set really to to, you know, deliver goals that they care about. 0:31:5.2 –> 0:31:12.562 Brandon Dey This is the last slide and I’m curious if anybody has any questions on everything or anything that we talked about so far. 0:31:23.582 –> 0:31:32.342 Brandon Dey If I jump back to this slide, this one might make sense more now with the context of of the full map and so. 0:31:33.912 –> 0:31:40.872 Brandon Dey I’ll speak to this for a second. Over here in utility zone, regardless of how visible the technology is. 0:31:42.632 –> 0:31:43.752 Brandon Dey Don’t build it. 0:31:43.752 –> 0:31:45.272 Brandon Dey It’s a utility, right? 0:31:45.872 –> 0:31:50.952 Brandon Dey These things are commoditized and it’s often not worth expensive. 0:31:50.992 –> 0:31:53.752 Brandon Dey You know, engineering hours to to build a utility. 0:31:55.392 –> 0:31:56.912 Brandon Dey As you move to the left though. 0:31:58.552 –> 0:31:59.912 Brandon Dey That calculus changes and so. 0:32:0.642 –> 0:32:2.122 Brandon Dey If it’s very visible. 0:32:3.872 –> 0:32:9.232 Brandon Dey And well above this minimum like accepted return that you need. 0:32:10.952 –> 0:32:15.112 Brandon Dey And it’s a relatively stable technology. 0:32:15.192 –> 0:32:16.672 Brandon Dey Build it if you can, right? 0:32:16.672 –> 0:32:23.32 Brandon Dey If you have the skills in house, build it if you can as you move to the left more, but you’re still up top if you have the budget. 0:32:26.292 –> 0:32:31.932 Brandon Dey You can more seriously consider building more nascent technologies, technologies that. 0:32:33.472 –> 0:32:36.512 Brandon Dey You don’t hear a lot about. There aren’t options. 0:32:36.512 –> 0:32:45.552 Brandon Dey Maybe in in the Saa’s marketplace, there’s maybe there aren’t many peers that are peers of yours that are doing this. If you have the budget for that. 0:32:47.152 –> 0:32:49.392 Brandon Dey You know you can be more likely to consider to to build it. 0:32:51.32 –> 0:32:51.352 Brandon Dey In house. 0:32:52.952 –> 0:32:57.552 Brandon Dey Now, if it’s brand new technology, it needs to basically be invented. 0:32:57.552 –> 0:33:3.552 Brandon Dey We did this a lot at Kroger and R&D where we were doing things that that there weren’t known solutions for and so. 0:33:4.862 –> 0:33:7.862 Brandon Dey You know, that’s why you have an R&D department is to invent new tech. 0:33:8.622 –> 0:33:11.582 Brandon Dey We built a lot of stuff because they had a very high tolerance for failure. 0:33:12.662 –> 0:33:15.142 Brandon Dey You know it is well resourced organization. 0:33:16.712 –> 0:33:26.392 Brandon Dey And that’s where they wanted to put their their resources. And so the the cost of failure wasn’t prohibitive now. 0:33:28.352 –> 0:33:30.992 Brandon Dey If if the technology. 0:33:32.552 –> 0:33:34.392 Brandon Dey Is more buried and more invisible. 0:33:36.112 –> 0:33:39.232 Brandon Dey You’re less likely to say yes to to building it right. 0:33:40.952 –> 0:33:47.272 Brandon Dey And as you move down to this threshold of investment, this return that you need. 0:33:48.992 –> 0:33:59.152 Brandon Dey You you get into this zone where you can sometimes find a lot of pet projects by various executives or leaders, and that’s not necessarily bad. 0:33:59.662 –> 0:34:8.782 Brandon Dey It’s still right around the return point and as you move to the right though, you should be less likely to say yes to building it. 0:34:9.732 –> 0:34:23.92 Brandon Dey And that’s why this who’s to say, you know, it’s kind of depends on depends on your organization, the really bad pet projects or the utility projects that, you know, somebody’s excited about and is choosing to prioritize for some reason. 0:34:24.622 –> 0:34:25.582 Brandon Dey Anything below this line? 0:34:27.302 –> 0:34:31.622 Brandon Dey It’s should be scrutinized the the choice to to build it. 0:34:34.252 –> 0:34:34.452 Brandon Dey OK. 0:34:34.452 –> 0:34:40.812 Brandon Dey So I’m gonna. I’m gonna wrap up there all these slides we made available. Afterwards, we’ll send them out. 0:34:41.412 –> 0:34:42.612 Brandon Dey But I hope this. 0:34:42.932 –> 0:34:53.92 Brandon Dey I hope you come away with something somewhat useful and you have a a framework that you can consider to use when you are deciding to invest or not in Gen. AI. 0:34:53.92 –> 0:34:56.532 Brandon Dey Appreciate the time everybody and I hope you have a great rest of your day.