释放代理型AI之力：人工智能下一次演进高管指南

An executive guide to next evolution in AI Table of contents Introduction The agentic AIopportunity 3 Chapter 1 The enterprise'sagentic AI challenge Chapter 2 Adopt an openplatform approach foragentic success Chapter 3 8 Charting youragentic AI strategywith Red Hat AI Learn more The future ofenterprise automationis autonomous 14 Introduction The agenticAI opportunity Today’s executives are facing a pivotal moment in theevolution of technology. While AI has already redefined howwe interact with data, automate tasks, and serve customers,the next phase in AI is emerging, pushing beyond contentgeneration and predictive analytics. interfaces (APIs), and improving over time. This canlook like responding to customer service issues acrossmultiple platforms, automating IT remediation steps,or managing supply chain operations in real time. Agentic AIis an evolution from generative models toautonomous systems with greater adaptability and thecapacity for continuous learning. Agentic AI can be aphysical structure, a software program, or a combinationof both. This allows agents to perceive, decide, andact within predefined parameters across virtuallyevery part of the business. Instead of responding toprompts, agentic AI is capable of initiating multisteptasks, accessing tools and application programming For senior leaders, understanding this shift is astrategic imperative for remaining competitive inthe future. How agentic AI works Agentic AI is most useful for tasks that require continuous monitoring or rapiddecision making. It can be thought of as a way of combining automation withreasoning, decision making, and the creative abilities of alarge languagemodel (LLM). To adopt agentic AI into everyday operations, organizations mustbegin by creating a system to provide an LLM with access to external toolsand algorithms that supply instructions for AI agents, which sit on top of othersoftware tools and operate them. AI agents can then communicate with tools involving orchestration of workflows,depending on the framework being used. This approach allows the LLM toreason and determine the best way to answer a question. By 2029, agentic AI will autonomouslyresolve 80% of common customer serviceissues without human intervention, leadingto a 30% reduction in operational costs.¹ The benefits of agentic AI include: Accelerated decisionmaking and executionacross business units,informed by real-time dataand contextual insight. Competitive differentiation, Greater operationalefficiencythroughautonomous workflowsthat reduce the need formanual intervention. Reduced costsas as organizations are able toorchestrate and govern AIagents to increase return oninvestment (ROI). a result of minimizinghuman intervention andimproving productivity. In an era defined by constant disruption, agentic AI helps organizations anticipate, adapt, and lead. The adoption challenge From orchestrating complex agent workflows, to maintaining trust and reliability, toscaling these innovations efficiently across the enterprise—incorporating agentic AIinto your organization will come with challenges. However, with business environmentsevolving more quickly than ever, adopting agentic AI is no longer optional fororganizations seeking to sustain their competitive edge. This e-book explores these challenges and how to overcome them, the businessadvantage of agentic AI, and actionable strategies to reshape efficiency andinnovation in your organization using agentic AI. Chapter 1 The enterprise'sagentic AI challenge Enterprise knowledge and use of AI is growing quickly, andwith that growth has come a level of maturity where moreorganizations are recognizing its long-term business value. In fact, AI-enabled workflows—many powered by agentic AI—arepoised to expand from 3% in 2024to 25% by 2026, as investment in AIcontinues to rise.² Understanding these challengesearly can empower leaders to askthe right questions, design pilotprograms effectively, and choosetechnologies that will set them upfor success. But transformation is never easy.There are real challenges that everyorganization must navigate beforethey can deploy intelligent agentsat scale and with confidence. Challenge 1: Orchestrating complex agent workflows A key differentiator of agentic AI systems, compared to other forms of AIsuch as gen AI, is that it isn’t meant to be a single-point solution. AI systems can comprehend tools, APIs, communication protocols, datasources, machine learning (ML) models, LLMs, and AI agents. The AI agentis the final decision, but this requires orchestration of the other parts. This level of orchestration introduces new layersof complexity, including the need to: Provide teams the right tools Coordinate across agents Manage across systems Developers and IT teams need newtools to design, test, and iteratethese workflows quickly so they’re notbuilding from scratch every time. Agentic AI must manage dependenciesacross systems that ca