book

AI Value Creators

by Rob Thomas, Paul Zikopoulos, Kate Soule

April 2025

Intermediate to advanced

300 pages

10h 21m

English

O'Reilly Media, Inc.

Read now

Unlock full access

GenAI Is a Lift, Shift, Rift, or Cliff MomentAs You Journey into the BookConventions Used in This BookO’Reilly Online LearningHow to Contact UsThe Collective Thank-YousOur Personal Dedications
What Is a “Netscape Moment”?AI and the Magical MomentBut...AI Is Not MagicMoving Your Business from +AI to AI+Before You Do Anything, Change Your Mental Model from +AI to AI+The AI Ladder, Rebooted for GenAIBefore You Start Your Journey, Classify the Budget and Identify How AI Is Going to HelpDimension One: Spend Money to Save Money, or Spend Money to Make Money? How Will AI Help Your Business?Dimension Two: Categorize How the AI Helps Your BusinessUse an Acumen Curve to Visualize How AI Helps Your BusinessWhere to Start? Here’s Our Helpful AdviceBecome a Shifty Business: Shift Left, and Then, You Can Shift Right!Every Day, We Walk by Problems That Can Be Solved or Made Better with TechnologyTips for Harnessing Foundation Models and GenAI for Your BusinessTip 1: Act with UrgencyTip 2: Be an AI Value Creator, Not Just an Occasional AI UserTip 3: One Model Will Not Rule Them All, So Make a Bet on CommunityTip 4: Run Everywhere, EfficientlyTip 5: Be Responsible Because Trust Is the Ultimate License to OperateAnd with That, Let’s Focus on the AI Part
AI Through the Years: The AI “Time Lapse” SectionA Quick Bit on Foundation ModelsGoing a Little Deeper: The Evolution of Large Language Models and Comparing Supervised Learning with Self-Supervised LearningAI Value Creation Should Be Your DestinationHow Do You Consume AI: Be Ye a Value Creator or a Value User?Planning Your AI Future: A Future with Many GenAI ModelsIt’s Time to Demystify and Apply AIThe Future of AILet’s Get into It
Some Things Are TimelessTension Has Always Existed with Technology—AlwaysNo Calculators Needed! Our Three Persuasion EquationsEquation 1: How to Grow GDPEquation 2: What Makes for AI Success?Equation 3: Find Your Balance—Navigate the ParadoxOne Last Piece of Advice: See AI as a Value Generator, Not a Cost CenterWrapping Up
The Use Case Value Creation CurveGoing Horizontal Gets You the Most VerticalExperimentationPutting Your Data to WorkIT AutomationCode—The Language of ComputersDigital Labor and AI AssistantsAgentsThe Business Lens: Use Cases—Horizontally SpeakingThe Bonus (Horizontal) Use Case—Synthetic DataA Smattering of Use Cases—Vertically SpeakingAgricultureAccountingEducationHealthcareInsuranceLegalManufacturing and ProductionPharmaEndless Possibilities: More Industries Where GenAI ShinesThe Building Blocks of AI
LLMs—The Stuff People Forget to Tell YouThe Knowledge Cut-Off DateLLMs Can Be Masters of Making It Up as They GoFootprints in the Carbon: The Climate Cost of Your AI BFFCopyright and LawsuitsWhat About Digital Essence?Your Expanding Surface Area of AttackData PrivacySteal Now, Crack LaterGood Actor Levers for All Things AIFairness—Playing Fair in the Age of AIBias Here, Bias There, Data Bias Is EverywhereRobustness—Ensuring Artificial Intelligence Is Unbreakable IntelligenceExplainability—Explain the Almost UnexplainableLineage—Tracing the Trail: Let Good Data PrevailRegulations—The Section That Wasn’t Supposed to BeWhat to Regulate—Our Point of ViewManaging the AI LifecycleWrapping It Up
Let the Skilling BeginThe Path to AI+ Requires Scaling Skills Across a Broad Spectrum of RolesAI—Job Destroyer or Job Creator?You’re Only Going to Get Checkmated if You Don’t Up Your SkillsDemocratized Technology: The Job CreatorLevers of Clever: Unlocking a Skills Program That Lasts ForeverLever 1: Start at the Beginning—Hire Employees Who Want to Know the “Why”Lever 2: Recruit Digitally Minded TalentLever 3: Take Count—Inventory Your SkillsLever 4: Plan for Everyone—A Plan Without Action Is a SpeechLever 5: Embrace the Learning (and Forgetting) CurvesLever 6: Combine Instruction + Imitation + CollaborationLever 7: Culture Matters—Be a Skills Verb, Not a NounLever 8: Set the Organizational Tone for AICase Study: IBM’s Skills Challenge—the CEO Asked; We All RespondedThe Final Word
The Bigger the Better, Right? Perhaps at the Start, But That Was a Long Time AgoThe Rise of the Small Language ModelData Curation Results in AI SalvationThink About This When It Comes to Data CurationModel Distillation—Using AI to Improve AIThink About This When It Comes to Model DistillationWhere Are We Going Next? Small Language Models...Assemble!Model RoutingThink About This When It Comes to Model RoutingMixture of Experts (MoE) ArchitectureThink About This When It Comes to MoEsAgentic SystemsWhat’s Your Reaction to This Agent in Action?A Little More on AgentsHow Agents Are BuiltRisks and Limitations of Agentic SystemsThree Tips to Get You Started: Our Agentic Best PracticesThink About This When It Comes to AI AgentsWrapping It Up
Customizing Open Source for the Enterprise: A New Way of Looking at Enterprise DataThe Original Eras Tour: Looking Back a Few Decades on Data RepresentationsStand Up and Represent!...Your DataStep 1: It All Starts with TrustStep 2: Representing your Enterprise Data within an LLMStep 3: The Grand Finale: Deployment and ExperimentationThe Future Is Open, Collaborative, and Customizable
The Building Blocks of ComputingTransformers—More Than Meets the AINot Back to the Future; Back to Computer ScienceDoors Wide Open—Reimagining the PossibleHow Models Are Built in Generative Computing“Libraries” for Adding Capabilities to a Generative Computing SystemThe Quick Compare Summary—How You Use LLMs Today Versus Generative ComputingA Generative Computing Runtime—What Can We Program It to Do?OpenAI’s Strawberry—A Berry Sweet InnovationFrom Generative Computing to a Generative Computer—What Does All of This Mean for Hardware?Experimenting with the Acceleration of AI at the IBM NorthPoleThe Final Prompt: Wrapping It All Up

Content preview from AI Value Creators

Chapter 2. Oh, to Be an AI Value Creator

In the previous chapter, we gave a set of imperatives that can instantly improve the odds of success for any AI journey. This all comes from our countless collective experiences, which range from thousands of customer engagements to TV shows such as 60 Minutes, to the US White House, NATO, senior management, and even the Vatican! (The Vatican houses priceless artifacts in nitrogen vaults, and we—well, the broader IBM team—helped open those artifacts up to scholars to safely scale knowledge and history. While we can’t share with you the details of that deal, we’re confident in the afterlife.)

You also learned about the Netscape moment of today and how it’s a tsunami of change that will wash across your personal and professional shores. You now understand that just as electricity was once deemed magical even though it wasn’t, AI is not magic either. We nudged (OK, two-hand shoved) you into a +AI to AI+ mindset and gave you a rebooted for this moment AI Ladder to climb for AI success. Finally, we gave you some operational frameworks to classify AI budgets, pick use cases, and envision outcomes that either shift left or shift right your business.

We think Chapter 1 and this chapter are important because they are both about defining the right details to pay attention to on your AI journeys. Why? Details will matter, details will differentiate, and details will earn (or keep) trust. We’ll use the history of the Statue of Liberty as an analogy ...