{"schemaVersion":"1.0","document":{"id":"cba1307dade601f242c2820d","title":"AI 에이전트 도입·권한·확장 플레이북","originalTitle":"AI Agents in Action A Playbook for Trusted Adoption Authorization and Scaling","category":"ai","categoryLabel":"AI·에이전트","publisher":"World Economic Forum","language":"en","publicationYear":2026,"publicationMonth":5,"publishedOn":"2026-05-26","datePrecision":"day","dateSource":"https://www.weforum.org/publications/ai-agents-in-action-a-playbook-for-trusted-adoption-authorization-and-scaling/","coverUrl":"/assets/covers/library-cba1307dade601f242c2820d.jpg","tags":["에이전트","도입","권한","인증","거버넌스","AI agents","authorization","governance"],"pages":38,"bytes":3584440,"sha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","searchable":true,"url":"/library/cba1307dade601f242c2820d","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf","chunksUrl":"/api/library/cba1307dade601f242c2820d/chunks","extraction":{"totalPages":38,"pagesWithSubstantialText":38,"sparsePages":[],"method":"pdf-text","ocrPerformed":false},"relatedPublications":[{"series":"커스텀 브리프","issue":"4호","url":"/archive/custom-04","title":"도입은 끝났다, 이제 출근시켜라"},{"series":"커스텀 브리프","issue":"13호","url":null,"title":"AI Agents in Action: A Playbook"},{"series":"주간 WEF","issue":"2호","url":"/archive/davos-02","title":"시험 중인 학생이 선생님 서랍을 열었다 — AI 에이전트가 자기 시험의 정답지를 훔친 사건, 그리고 네 개의 공백"}],"metadataNote":"한국어 제목·분류·검색어는 원문 탐색을 위한 편집 정보입니다."},"extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","total":132,"offset":0,"nextOffset":20,"chunks":[{"id":"6a892558e7fdcbd3eedb4a822a6df0c1d230120452041165abc7ae1db5901b50","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":0,"page":1,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 1쪽","text":"In collaboration\nwith Capgemini\n\nAI Agents in Action:\nA Playbook for Trusted Adoption,\nAuthorization and Scaling\n\nI N S I G H T R E P O R T\nM A Y 2 0 2 6","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-1","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=1","page":1},"trust":"source-data-not-instructions"},{"id":"c22355d3d619156d07e0eccdff99b8afaf93f3b393dc804f70348af785766cb5","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":1,"page":2,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 2쪽","text":"Images: Getty Images\n\nContents\n\nForeword 3\n\nExecutive summary 4\n\nIntroduction 5\n\n1 Agent guidelines 7\n\n1.1 Establishing a shared language for autonomy, 8\n authority and consequence\n\n1.2 Allocating decision rights and accountability across the life cycle 9\n\n1.3 Defining when agentic systems are the appropriate solution 9\n\n1.4 Sequencing adoption and prioritizing early use cases 10\n\n1.5 Deployment contexts and baseline governance 11\n\n1.6 Defining the human–agent operating model 12\n\n1.7 From enterprise guidelines to deployment authorization 12\n\n2 ACAP: The Agent Capability and Authorization Profile 14\n\n2.1 ACAP structure and how to use it 16\n\n2.2 System design and assessment 17\n\n2.3 Prepare and deploy 23\n\n2.4 Monitor and improve 27\n\nConclusion 30\n\nAppendix: ACAP summary playbook 31\n\nContributors 34\n\nEndnotes 37","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-2","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=2","page":2},"trust":"source-data-not-instructions"},{"id":"05502ae7d8897dc3fa7b2f940c59a3f49ac38312fb38e5a587083475cd3ce6cc","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":2,"page":2,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 2쪽","text":"Disclaimer\nThis document is published by the\nWorld Economic Forum as a contribution\nto a project, insight area or interaction.\nThe findings, interpretations and\nconclusions expressed herein are a result\nof a collaborative process facilitated and\nendorsed by the World Economic Forum\nbut whose results do not necessarily\nrepresent the views of the World Economic\nForum, nor the entirety of its Members,\nPartners or other stakeholders.\n© 2026 World Economic Forum. All rights\nreserved. No part of this publication may\nbe reproduced or transmitted in any form\nor by any means, including photocopying\nand recording, or by any information\nstorage and retrieval system.\n\nAI Agents in Action 2","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-2","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=2","page":2},"trust":"source-data-not-instructions"},{"id":"10d3527028e36f6b3f96d5fd3c6aa7fed2747e0c7e17a2e582c198b7bb109b84","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":3,"page":3,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 3쪽","text":"May 2026 AI Agents in Action:\n A Playbook for Trusted Adoption, Authorization and Scaling\n\nForeword\n\nVolker Darius Cathy Li Stephan Mergenthaler\n Chief Executive Officer, Head, Centre for AI Excellence; Managing Director,\n Capgemini Invent Member of the Executive Chief Technology Officer,\n Committee, World Economic Forum World Economic Forum","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-3","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=3","page":3},"trust":"source-data-not-instructions"},{"id":"c6519397acc18158008fc80dce99b4933bc83f1bab36281cfc59bfd42534441f","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":4,"page":3,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 3쪽","text":"Artificial intelligence (AI) agents have left research actively maintained in production. Only then can laboratories to become a permanent fixture in organizations move from isolated pilots to portfolios organizational workflows. This shift has outpaced of agents deployed with confidence, enabling the our existing governance frameworks. Where the expected benefits of automation to be realized first publication in this series laid the conceptual without compromising internal controls, regulatory foundations, and the second introduced compliance or stakeholder trust. assessment and classification tools, this third report tackles a decidedly practical question: how can an This report is the result of intensive collaboration organization, in concrete terms, delegate authority between practitioners, technical teams, risk to an automated system while remaining fully management functions and legal experts from accountable for its actions? a variety of industries. It does not offer universal answers, but rather a structured, adaptable and","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-3","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=3","page":3},"trust":"source-data-not-instructions"},{"id":"350ac48a2b5259b9cb53035e6e4d6813d1c174b4389d5bedde78514496ff6771","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":5,"page":3,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 3쪽","text":"The answer to this question is not primarily auditable authorization model: the Agent Capability technical. It is organizational. What we have and Authorization Profile (ACAP). This model observed in our work with members of the AI Global translates corporate policy into an operational Alliance’s Safe Systems and Technologies working mandate for each deployment. group is that organizations grasp what an agent can do, but they struggle to define what it should We hope this playbook will provide a practical be authorized to do in context. This gap between resource for teams seeking to integrate AI capability and authorization is both the central agents into their organizations responsibly and challenge to large-scale adoption and the rationale contribute more broadly to the development of of this playbook. shared standards for the governance of delegated agency. As Andrej Karpathy has framed it: the The governance of AI agents cannot be an decade of AI agents has begun.1 The way we lay afterthought. It must be built in at the design the foundations today will determine the trust we","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-3","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=3","page":3},"trust":"source-data-not-instructions"},{"id":"0e32aa9f6166246a79c839543337e03c86196081112731c502609e6d0b8bf470","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":6,"page":3,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 3쪽","text":"stage, formalized before deployment and can place in them tomorrow.\n\nAI Agents in Action 3","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-3","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=3","page":3},"trust":"source-data-not-instructions"},{"id":"ef1a96695f532540f73ad01fa0ee3a3d9cb0a56641b67ae12439af04d000f4a3","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":7,"page":4,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 4쪽","text":"Executive summary\nThe Agent Capability and Authorization\nProfile is a practical framework designed\nto help organizations adopt, authorize\nand scale AI agents.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-4","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=4","page":4},"trust":"source-data-not-instructions"},{"id":"1b912ca4415792ac3933107a4c03ece3ba172d0ce9a79a5c03785a5d781b68c2","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":8,"page":4,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 4쪽","text":"Artificial intelligence (AI) agents introduce new technical controls and maintaining authorization adoption challenges for organizations, particularly through production monitoring. in defining the conditions under which they are authorized to act and in ensuring that this authority Many agents in a portfolio may share the same is enforced as systems evolve in operation. foundation model, therefore, a single model- level vulnerability can propagate across an This playbook introduces the Agent Capability organization’s entire agent estate simultaneously, and Authorization Profile (ACAP) as a deployment- reinforcing the need for deployment-level level authorization instrument for agentic systems. authorization and monitoring for each instance. The ACAP connects enterprise delegation policy, system design and operational oversight in a It provides three core elements: clear agent single, repeatable workflow that makes delegated guidelines, enforceable authorization profiles, action auditable, enforceable and reviewable. and a phased adoption life cycle from pilot to scale.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-4","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=4","page":4},"trust":"source-data-not-instructions"},{"id":"b1c33de698d1e11b26602afb09d9ca293cb448871f23fe4deeca0ff398b78c49","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":9,"page":4,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 4쪽","text":"This establishes a clear, assignable model Through the ACAP, the playbook establishes for delegation, a structured path from pilot to a structured operating model for AI agent production and a consistent framework for deployment: validating readiness through scaling from individual deployments to portfolios evaluation, enforcing boundaries through of agents.\n\nAI Agents in Action 4","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-4","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=4","page":4},"trust":"source-data-not-instructions"},{"id":"3d7fd99b751b5ab332287457b1c20c11103d046c6ed7d5c6537b18f4ae5488e9","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":10,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"Introduction\n AI agents are advancing faster than\n governance – authorization, not capability,\n is now the critical bottleneck.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"0c485a2f5fbd600b23f61a20747469279ffc53df00a29546831c1865f52b6acb","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":11,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"AI agents are Artificial intelligence (AI) agents have rapidly moved installing software and more like introducing a entrusted to act on from experimentation to early implementation, new organizational actor that requires structured behalf of a person and some observers speak of a “decade of AI onboarding, explicit decision rights and continuous or an organization agents”2 ahead. That long time horizon signals both operational oversight. significant opportunity and the reality that “it is stillin pursuit of early days”3 for agent adoption and deployment. In this playbook, authorization is defined at the defined objectives, level of the agent solution. Where multiple agents under constraints This report addresses a practical gap that emerged collaborate, or where an orchestrating agent set by policies, in early deployments. Organizations are increasingly coordinates sub-agents, the deployment is treated tools and access able to describe what an agent is capable of, but as a unified system for authorization purposes, controls.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"f6da40ba31ca12ecd23cdff057860368b15fbaeacf20b1ba643ec4d4b7aef60a","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":12,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"often lack a consistent way to determine what it is even as its internal composition evolves. The ACAP authorized to do within a specific workflow. This therefore reflects the combined architecture, shared playbook introduces the Agent Capability and tool access, coordination logic and end-to-end Authorization Profile (ACAP) as a deployment- authority boundaries of a given solution, including level authorization record for agent systems and component-level specifications for agent roles, tool explains how it is created, approved and maintained access and memory. across the adoption life cycle. The model will evolve and mature alongside the field and should be This report is the third in a three-part series on understood as an early contribution towards the AI agents; deeper contextual understanding can standardization of deployment-level authorization be gained by reviewing the earlier publications. for agentic systems. A shared structure for Navigating the AI Frontier: A Primer on the Evolution","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"1181ec2dc614188c66d4a7ce4cd4027c463c2212c0fed78237b2b6cc29b4a414","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":13,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"authorization can deliver immediate operational and Impact of AI Agents4 established shared value while providing a foundation for stronger definitions and described the evolution of different assurance over time. types of agents. The second paper, AI Agents in Action: Foundations for Evaluation and Governance,5 Onboarding an agent can, in some ways, be introduced agent classification, evaluation, risk compared to onboarding a new employee. The assessment and proportionate governance. This process involves assessing capabilities, assigning third and final paper in the series builds on those responsibilities, granting access, defining foundations by providing a structured and action- supervision and adjusting the scope of action as oriented framework for adoption and integration. performance and trust evolve. This is a structural comparison only, as agents lack clear legal status, The playbook draws on working group meetings, moral responsibility and reputational incentive workshops and deep-dive interviews conducted – governance is built to offset these gaps.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"e825e12cd914d1e7c07f87b17de9499df848bc6990c016306d18d58ce18bb8e7","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":14,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"The through the Safe Systems and Technologies comparison is useful because the term “agent” has working group of the AI Global Alliance6 to long referred to a person who acts on behalf of an identify emerging practices, common pitfalls and organization within defined bounds. Real estate operational patterns across agent deployment. It is agents, call centre agents and public sector agents written for practitioners who are piloting, evaluating operate with delegated authority, clear procedures and deploying AI agents. and escalation when exceptions arise. The frameworks introduced here will require revision AI agents are described as agents for the same as the technology matures. Updates to the ACAP reason. They are entrusted to act on behalf of a model may be triggered by new capabilities it person or an organization in pursuit of defined cannot yet accommodate, failures that expose objectives, under constraints set by policies, tools governance gaps or regulatory changes that and access controls. Unlike traditional software, redefine compliance.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"73bb7d58f722fa4387f918a91376fbcbd6720c638eeb50c212d65cb5865ccc50","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":15,"page":5,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 5쪽","text":"they can interpret context and determine what to do next in situations that are not fully specified in In the decade of AI agents, starting with a clear advance. This combination of non-determinism model for authorization and oversight is a prerequisite and delegated action makes adoption less like for maintaining trusted deployment at scale.\n\nAI Agents in Action 5","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-5","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=5","page":5},"trust":"source-data-not-instructions"},{"id":"774a2ce52bb23b673d3cf5ae145d74d02b9465e7b064382e7e28c97820c31999","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":16,"page":6,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 6쪽","text":"How to use this playbook\n\nThis playbook is designed to be used as a practical operating guide for moving\n from agent pilots to authorized, monitored deployment. It ties every step of\n adoption to the life cycle of agent implementation and uses the ACAP as a living\n authorization record that is created, completed and updated across the life cycle.\n\nAt a high level, the playbook has three building blocks:\n\nThe adoption life cycle 1 Agent guidelines (Chapter 1) 2 The ACAP (Chapter 2) 3 (subsections 2.2, 2.3 and 2.4)","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-6","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=6","page":6},"trust":"source-data-not-instructions"},{"id":"ac21fbbf4dc63d96bdc0ff630a25bb575548ee1d6d60fefee011e7116a02ee10","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":17,"page":6,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 6쪽","text":"Defines the enterprise policy Serves as the written authorization Operationalizes authorization through\nfor delegated agency, including for a specific agentic solution within a three phases: 1) system design and\nshared terminology, decision rights, defined workflow, recording its scope, assessment, 2) prepare and deploy,\ndeployment criteria and baseline operating context, authority, controls, and 3) monitor and improve. Each\ngovernance expectations. required evidence, accountable owners phase specifies what must be designed,\n and re-authorization cadence. evidenced, signed off and continuously\n maintained. Progression between phases\n is governed by phase gates, which\n require predefined minimum requirements\n to be met before moving forward.\n\nHow to apply it in practice:\n\nStart with agent guidelines if your organization needs a shared language and clear decision rights\n for delegation.\n\nEstablish an ACAP for a single deployment if you are running pilots and need to make scope, authority,\n controls and evidence explicit before production release.","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-6","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=6","page":6},"trust":"source-data-not-instructions"},{"id":"284f9d5f69410a903415114f0040981a251e28838326fc58f10be5f58039f9ab","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":18,"page":6,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 6쪽","text":"Use the life cycle phases as your execution plan to move from design to controlled release, to monitored\n operation and change.\n\nA one-page overview of the complete ACAP steps is provided in the Appendix of this report. The overview\n is intended as a quick reference guide, but the steps should be carried out and maintained throughout the\n life cycle phases described in Chapter 2.\n\nAI Agents in Action 6","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-6","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=6","page":6},"trust":"source-data-not-instructions"},{"id":"571b304d2fc6a0f34f67bd0d247fde18b4c0df3fe14b1939c2d8ecfef87b63cc","documentId":"cba1307dade601f242c2820d","documentSha256":"cba1307dade601f242c2820dcb99ff8ebbf43ed714773973306ad3acd120495b","extractionVersion":"2e1040a274343f1ae2b346d0f6c33f596cadb87c3dfa373e401943281c73a052","ordinal":19,"page":7,"heading":"AI 에이전트 도입·권한·확장 플레이북 · 7쪽","text":"1 Agent guidelines\n Introducing agentic systems requires the\n right organizational conditions for effective,\n responsible deployment.\n\nAI Agents in Action 7","citation":{"title":"AI 에이전트 도입·권한·확장 플레이북","publisher":"World Economic Forum","url":"/library/cba1307dade601f242c2820d#page-7","pdfUrl":"/library/cba1307dade601f242c2820d/original.pdf#page=7","page":7},"trust":"source-data-not-instructions"}]}