AI Value Creators Beyound the Generative-AI-User Mindset (Rob Thomas, Paul Zikopoulos, Kate Soule) (Z-Library)

Praise for AI Value Creators A handbook for the AI Renaissance to help entrepreneurs and innovators drive AI value creation at the next level. —will.i.am, founder and CEO, FYI.AI Rob Thomas brings insight, common sense, and his long experience at IBM to bear on the greatest technological transformations of our lifetime. On the subject of AI, there are few people whose perspective I would value more. —Malcolm Gladwell, host of the Revisionist History podcast With AI reshaping industries, this handbook provides actionable insights that can help you drive innovation and navigate the next wave of AI advancements, positioning your business for long-term success. —Jessica Sibley, CEO, TIME

AI Value Creators Beyond the Generative AI User Mindset Rob Thomas, Paul Zikopoulos, and Kate Soule

AI Value Creators by Rob Thomas, Paul Zikopoulos, and Kate Soule Copyright © 2025 O’Reilly Media, Inc. All rights reserved. Printed in the United States of America. Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472. O’Reilly books may be purchased for educational, business, or sales promotional use. Online editions are also available for most titles (http://oreilly.com). For more information, contact our corporate/institutional sales department: 800-998-9938 or corporate@oreilly.com. Acquisitions Editor: David Michelson Development Editor: Gary O’Brien Production Editor: Kristen Brown Copyeditors: Doug McNair and nSight, Inc. Proofreader: Sonia Saruba Indexer: Potomac Indexing, LLC

Interior Designer: David Futato Cover Designer: Karen Montgomery Illustrator: Kate Dullea April 2025: First Edition Revision History for the First Edition 2025-04-01: First Release See http://oreilly.com/catalog/errata.csp?isbn=9781098168346 for release details. The O’Reilly logo is a registered trademark of O’Reilly Media, Inc. AI Value Creators, the cover image, and related trade dress are trademarks of O’Reilly Media, Inc. The views expressed in this work are those of the authors and do not represent the publisher’s views. While the publisher and the authors have used good faith efforts to ensure that the information and instructions contained in this work are accurate, the publisher and the authors disclaim all responsibility for errors or omissions, including without limitation responsibility for damages resulting from the use of

or reliance on this work. Use of the information and instructions contained in this work is at your own risk. If any code samples or other technology this work contains or describes is subject to open source licenses or the intellectual property rights of others, it is your responsibility to ensure that your use thereof complies with such licenses and/or rights. This work is part of a collaboration between O’Reilly and IBM. See our statement of editorial independence. 978-1-098-16834-6 [LSI]

Preface Thrilling—the one word we use to describe the possibilities and eventualities that generative AI (GenAI) and agents will enable. From the boiler room to the board room, we think GenAI will truly impact every industry. As technologists, this level of excitement comes about only once, maybe every other decade, and this is why more and more, every day, the possibilities of GenAI are being recognized by people all over the world. Indeed, GenAI has the promise of a revolution, but this one will affect the high-status brainwork that the Industrial Revolution never touched. As many know, Steve Jobs dropped out of college and went on to lead one of the most successful companies in the history of the world. Many know Jobs loved simplistic elegance and the beauty of things (like fonts on which he spent a great deal of time studying), and you can see that in Apple’s products to this very day. Many don’t know he was also fascinated with the efficiency of locomotion. He took particular interest in a study that looked at the least amount of energy a species would use to race one kilometer (0.62 miles). It might surprise you (it surprised us) to note that the winner in this category was a condor! Humans? About one-third down the list. But when

humankind got on a bicycle, they blew everyone off the efficiency charts. He concluded that humankind could build tools (computers then, and now GenAI and agents) that can make us better—while other species must adapt (which takes a long time, if they can at all). In 1990, he likened a computer to a bicycle and noted, “What a computer is to me is the most remarkable tool that we have ever come up with. It’s the equivalent of a bicycle for our minds.” We’re pretty sure if he was around today, he’d note that if a computer is a bicycle for our minds, then GenAI and agents combine to become the bicycle for your business. And just as people have mastered the art of using Excel to manage their finances, track expenses, and create intricate color-coded charts (often with more complexity than necessary), these technologies are poised to become the standard tool for automating tasks, sparking creative ideas, and making it seem like you’ve worked tirelessly to achieve them. While this all sounds promising, remember that AI is not a promise about prosperity, and a better world isn’t guaranteed simply because you use GenAI or agents. Why? It can also have a dark side to it, which you will also learn about in this book.

TIMES CHANGE, TECHNOLOGY CHANGES FASTER Let’s get this out of the way: stuff is going to be out of date by the time you read this book. Today, writing about AI is like giving you stats on the number of images uploaded to the internet every second. We tried our best to keep stuff up-to-date (for example, DeepSeek came out, and we harassed our editor to open the manuscript to make some updates). And for sure, there will be new benchmarks, new papers, new state-of-the-art (SOTA) frontier models, GPUs, other accelerators, and more. We reference in this book how technology years used to be like dog years (1:7) and now they’ve become more like mouse years (1:30). So, bear with us when something new shows up that could be at odds with something in the book. That said, we think you’ll find the main point of this book is to give you a mental model of the things to think about on your GenAI and agentic journeys, the things that deserve your attention, the questions to ask (yourselves and your vendors), and more. We think you’ll agree that what we teach you in this book has a way longer shelf life than what GenAI model is today’s SOTA and tomorrow’s “Wait, people actually used that?” With that in mind, we encourage you to look beyond the stats or a model version and ensure you really absorb the advice that lies within —it comes from a place of success and failure, and a large

corpus of real-world experiences and observations of what works and what doesn’t. For example, the buzz around DeepSeek in early 2025 (shortly before we went to print) thunderously demonstrated the very timeless points we are making in this book. Let us explain. The market mostly assumed that training cutting-edge models requires millions (or hundreds of millions) in investment with the latest, fastest, and greatest chips. That it had to be proprietary, and that trade secrecy was essential. DeepSeek proved otherwise, using the very things we’ve written about or mentioned in this book (Mixture of Experts, distillation, and reinforcement learning, among others) and some clever new optimizations (which we cover in this book too). It was released with a very permissive MIT license—that speaks to the open community we talk about in this book. So, while it may not show up in a benchmark comparison chart in this book, that was never the point because benchmarks are like a Whac-A- Mole game. But you will see the very concepts we lay out in this book go to work in the market and spur on new movements and innovations. And make no mistake about it: new innovations and techniques will arise, but we think those will tuck nicely into the playbook we’re giving you in this book to put it to work.

GenAI Is a Lift, Shift, Rift, or Cliff Moment We want you to think back to the first time you heard about GenAI. It’s a phrase that really became part of the public conversation in, maybe, late 2022. We have seen new models, evolved models, and an explosion of open models. In a matter of months, GenAI transformed from an intriguing curiosity into a fundamental force driving business innovation, with a fresh wave of use cases and applications emerging each day. There is such rapid growth that we can’t predict exactly where we will all be 10 years from now—or even 10 months from when we finished writing this book. But one thing we’re certain about: you’re going to want to be actively engaged in shaping that journey—and hopefully that’s why you’re reading this book right now. We’re at a moment in time here: one that is moving from a world of processes run by humans supported by technologies to one where processes will be run by technology that are supported by (or assisting) humans. This truly is a lift (good), shift (opportunity), rift (not so good), or cliff (what you’re running toward if you don’t upskill) moment for you as a

leader, you as an individual, and for the companies you work for. The future of AI is not one amazing model to do everything for everyone (you will hear us tell you time and time again in this book: one model will not rule them all). AI’s future will not just be multimodal (seeing, hearing, writing, and so on); it will also most certainly be multimodel (in the same way cloud became hybrid). AI needs to be democratized—and that can only happen if we collectively leverage the energy and the transparency of open source and open science—this will give everyone a voice in what AI is, what it does, how it’s used, and how it impacts society. It will ensure that the advancements in AI are not driven by the privileged few, but empowered by the many. Indeed, the DeepSeek hoopla raises a bigger question: Who will shape the future of AI? Again, we think AI development cannot be controlled by a handful of players, especially when some may not share the same fundamental values, such as protection of enterprise data, privacy, transparency, and more. We can’t let AI leadership slip to those with different values and priorities. That would mean ceding control of a technology that will reshape every industry and every part of society. And this is why we’ll keep saying that innovation and true progress can

only come by democratizing AI. We think that 2025 must be the year when generative AI gets unlocked from its confines within a few players; and into 2026, we hope that a broad swath of society won’t just be using AI—many will be building it too. As you read this book, you will see why we think a huge part of an enterprise’s GenAI toolkit will be smaller open source models—this is how the future will be built. For too long, AI has been seen as a game of scale—where bigger models meant better outcomes. But the real breakthrough is as much about size as it is about efficiency. In our work at IBM, we’ve seen that fit-for-purpose models have already led to up to thirtyfold reductions in AI inference costs, and made training more efficient and AI more accessible. There is no law of physics that dictates AI must remain expensive. The cost of training and inference isn’t fixed—it is an engineering challenge to be solved. Businesses, both incumbents and upstarts, have the ingenuity to push these costs down and make AI more practical and widespread. There’s an old Chinese proverb about when is the best time to plant a tree. Whatever that time is (it varies depending on who tells you the story), it’s in the past. But there is no argument about the next best time: today. We want to thank you for taking the initiative to read our book. We’re hoping you’ll be

thanking us when you’re done reading it because we’re going to give you a framework so you can start making your GenAI and agentic plans and how you can effectively, safely, and responsibly put AI to work for business. As You Journey into the Book In this book, we are going to demystify AI—generative AI and agents. We’ll explore a bit on how we got here, how it works, and many of the ways they are poised (and will) transform businesses and societies at unprecedented scale. We often refer to this point in time as a “Netscape moment” (Netscape being the world’s first internet browser) because that’s just how profound of an effect we think this technology will have on us all. Before you dive in, here’s a quick summary to give you the highlights. Consider it your cheat sheet—without the guilt and with all the good stuff. Use it as a trailer for the book ahead, or to jump into a section that really catches your attention. Chapter 1, “AI to AI+: Generative AI and the ‘Netscape Moment’”

This is a business moment that rarely comes around; in fact, the last time we saw something this big, it was 1993 when a web browser (called Netscape) freed up the internet from the hands of the privileged few and democratized it for the many. Don’t miss it. Things are going to change in the same way they did with the internet. In this moment, you ultimately won’t compete against AI, but you will be competing against other companies using AI. Think about it. If you’re a company spending 25% of your budget on customer service and another company shifts two-thirds of this same spend to have AI do most of the work...well. GenAI and agents will become a dividing line between which businesses will prosper, and which will struggle to keep pace. But always remember, AI is not magic. A couple of years from now, you will think back to this chapter and literally see which companies were the thrivers, which were the divers, and which ones showed up out of nowhere to become the new arrivers. Which one will you be? Chapter 2, “Oh, to Be an AI Value Creator” There are many ways to use AI. Be an AI Value Creator not just an AI User! Start your AI journey with the notion that your data is important and you shouldn’t give it away.

You’ll need an AI platform to become an AI Value Creator. AI Value Creators accrue and create much more value than AI Users. Chapter 3, “Equations for AI Persuasion” We give you a productivity paradox. Today there are many factors working against business success (decreasing productivity, declining population rates, and more friction and cost when accessing debt). You have a unique opportunity to put AI to work against these forces, especially with AI-fueled productivity and digital labor. Chapter 4, “The Use Case Chapter” This is not about pet projects; this is about use cases that drive real value. When you step back, you start to realize that if you master the horizontal use cases of AI (the patterns and the things they can do, such as see, hear, analyze, and more), you will more masterfully choose the right vertical use cases for your business. Remember, computer vision is computer vision. Writing is writing. After all, to a computer, everything is just a bunch of numbers—even your Taylor Swift Spotify playlist. And don’t forget, ensure you take your company over the Value Tipping Point.

Chapter 5, “Live, Die, Buy, or Try—Much Will Be Decided by AI” AI that people trust is AI that people will use. You must decide up front if you are going to be a good actor or a bad actor—you’ve seen both around social media and other innovations. Make this decision up front! Why? The world needs regulation at the speed of right and too often, governments move at the speed of molasses. In the end, governments (and hopefully customers) are going to demand that your AI is both explainable and accountable. Stuff your AI journey backpack with fairness, robustness, explainability, and lineage—it’s the ultimate gear for scaling new heights and taking in some truly breathtaking views. Ensure that these are forethoughts and not afterthoughts because, like we named in the chapter, “Live, die, buy, and try—much will get decided by AI.” Chapter 6, “Skills That Thrill” Because the half-life of technology skills is so short, know this: you will miss out on the amazing potential GenAI and agents can deliver to your business if you don’t constantly upskill the many. Your teams need to know what AI can do, what it can’t do, what to look out for, and more. No, you don’t need everyone to have computer

science degrees, but the only way to democratize AI for the many is to upskill the many. After all, how can you stop walking by problems every day that you could solve or make better with technology if you don’t know that they are even fixable (using technology) in the first place? Chapter 7, “Where This Technology Is Headed—One Model Will Not Rule Them All!” Remember, one model won’t rule them all. A carpenter’s tool belt doesn’t have one tool; it has many tools. What’s more, smaller, more nimble models are showcasing incredible results while addressing some major challenges the world is facing with the traditional large language model (LLM) approach used by many today. But in the end, open access to multiple transparent and open models is going to give you the best chance at success. Chapter 8, “Using Your Data as a Differentiator” The title says it all: leverage your data as a differentiator. This pairs with being an AI Value Creator. When you step back and realize that perhaps 1% (at most) of enterprise data is in the commonplace LLMs you’re likely using today, you realize there is value to be had. Data is like a gym membership; if you don’t use it, you get nothing from it, but you also can’t just give it away.

Chapter 9, “Generative Computing—A New Style of Computing” A glimpse into the future where generative AI takes its rightful place alongside classical computing and quantum computing as a new building block for applications called generative computing. This implies that the way we use LLMs is going to leverage software development methodologies that will broaden their applicability, safety, scale, performance, and more. We also expect (it’s happening today) our LLMs to reason more, take their time when warranted, and be thoughtful in their responses. This gives rise to a new area of optimization and “magic”—inference time. This change will pull forward new hardware and accelerators, creating a new compute stack, perhaps even a new generative computer. We think—with the help of this book—when you look back at this moment in history, you will be able to do so fondly, as someone who embraced data as a true resource and used GenAI and agents as a utility to create value. Here we are—the start of your journey. Let’s get into it.

Conventions Used in This Book The following typographical conventions are used in this book: Italic Indicates new terms, URLs, email addresses, filenames, and file extensions. Constant width Used for program listings, as well as within paragraphs to refer to program elements such as variable or function names, databases, data types, environment variables, statements, and keywords. Constant width bold Shows commands or other text that should be typed literally by the user. Constant width italic Shows text that should be replaced with user-supplied values or by values determined by context.