智能自治手册：如何用智能自治重新定义零售业

Executive Summary Agentic AI (artificial intelligence) systems are capable of solving complex problems and taking action both as directed by humans and autonomously, andcan even work and communicate with other AI agents. Retail is a highly competitive industry, and retailers that embrace AI, including AI agents, will have anadvantage in 2025 and beyond. ThisPlaybookis a comprehensive guide for companies searching for the right approach to adopting agentic AI and seizing the benefits of this new technology. Coresight Research Analysis 4. Risk and Resilience—Agentic AI Systems BringAdditional Security Concerns 1. Strategy—Cost Is a Topmost Concern When DecidingWhether or Not to Implement Agentic AI •Retailers need to analyze whether agentic AI technology fitstheir goals and brings an acceptable return on investment (ROI),as well as whether the implementation costs are manageable. •Deploying agentic AI systems introduces complex risksto organizations—with challenges ranging from securityvulnerabilities to ethical dilemmas—necessitating architecturaland operational safeguards. 2. Technology—An Option for Every Company, fromPre-Built Platforms to In-House Development 5. Ethics—Greater Power Requires Greater Oversight •Autonomous agentic AI systems require greater oversight thanpure GenAI (generative AI) platforms, especially when they areconnected to company systems or machines and processes inthe physical world. •There are four main pathways for agentic AI development,ranging from code-based development done in-house tocomplete enterprise applications and targeted startup offerings. 3. Operations—A Structured Approach Is Required,as Is Human Oversight What We Think •Implementing agentic AI in a corporate environment requires astructured approach with clear milestones and success metrics.Execution should be incremental, starting with manageablepilots to validate performance and minimize risk. The goal of AI is to automate the tedious, repetitive and evendangerous tasks that humans face, freeing humans to completethe tasks that we are better suited for, such as customer interactionand intuitive communication. Agentic AI offers powerful tools tohelp retailers do their business better, with examples spanningthe retail chain, including monitoring supply chains, handlingroutine customer inquiries and ensuring that stores and malls arewell-lit and comfortable. Additionally, as retailers face heightenedcompetition for workers from other industries, AI agents can helpfill in some of the gaps in their workforces. Contents Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Market Scale and Opportunity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 1.Strategy—Deciding Whether or Not to Implement Agentic AI..........................102.Technology—A Wide Array of Options..........................................123.Operations—How to Implement an Agentic AI System...............................194.Risk and Resilience—Agentic AI Systems Bring Additional Security Concerns................215.Ethics—Greater Power Requires Greater Oversight. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23 What We Think. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25 The Coresight Research View on AI..............................................25Implications from This Report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26Implications for Brands/Retailers................................................26Implications for Technology Vendors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26 Notes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .28 About AWS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .29 Introduction Agentic AI (artificial intelligence) systems are capable of solving complex problems and takingaction both as directed by humans and autonomously, leveraging their ability to analyzesituations, adapt to new information in real time, and even work and communicate with other AIagents without human input. With these benefits, agentic AI offers new, unparalleled capabilities for enterprises to leveragethe power of multiple types of AI, such as ML (machine learning) and GenAI (generative AI),and benefit from the creation of an agentic workforce that can operate on its own, which isparticularly important during times of (human) worker shortages. •Agentic AI is one of the top AI trends that Coresight Research will watch throughout 2025.We see this year as the year of agentic AI, with many leading global technology companieshaving already made strong commitments to develop AI agents. ThisPlaybookis intended for companies searc