网络自动化与自治第三阶段：自主移动网络的代理人工智能

AGENTIC AI FORAUTONOMOUSMOBILE NETWORKS V1.0ngmn.org Network Automation and Autonomy Phase III:AGENTIC AI FOR AUTONOMOUSMOBILE NETWORKS by NGMN Alliance Public documents (P): © 2026 Next Generation Mobile Networks Alliance e.V. All rights reserved. No part of this documentmay be reproduced or transmitted in any form or by any means without prior written permission from NGMN Alliance e.V.The information contained in this document represents the current view held by NGMN Alliance e.V. on the issues discussed as ofthe date of publication. This document is provided “as is” with no warranties whatsoever including any warranty of merchantability,non-infringement, or fitness for any particular purpose. All liability (including liability for infringement of any property rights)relating to the use of information in this document is disclaimed. No license, express or implied, to any intellectual property rightsare granted herein. This document is distributed for informational purposes only and is subject to change without notice. Readersshould not design products based on this document. Any links to external websites are provided for informational purposesonly - NGMN do not adopt, endorse, or assume responsibility for the content of such external websites. CONTENTS EXECUTIVE SUMMARY........................4 05CONSIDERATIONSFOR THE ECOSYSTEM.............23 01INTRODUCTION.............................5 5.1Purpose and Positioningof Considerations...........................235.2Considerations forOpen-source Communities,Standards and IndustryOrganisations..................................235.3Considerations for MNOs.............245.4Considerations for Vendors,Solution Providers andHyperscalers....................................24 1.1Overview.............................................5 1.2Core Concepts: Agentic AI,AI Agents and AgenticCapabilities.........................................5 1.3Document Structure andAnalytical Framework......................6 1.4Target Audience.................................7 02PROBLEM STATEMENTAND BUSINESS DRIVERS.......8 06CONCLUSION................................25 03SELECTED SCENARIOSDEFINITIONAND ANALYSIS...............................9 APPENDIX A:GLOSSARY...................................................27 3.1Fault Management............................93.2Service Assurance..........................113.3RAN Optimisation..........................13 APPENDIX B:EXISTING STANDARDS,OPEN SROUCEPROJECTS ANDINDUSTRY INITIATIVES...................30 04REFERENCE FRAMEWORKFOR AGENTIC AICOLLABORATION.....................15 REFERENCES...............................................35 4.1Core Components...........................154.2Agent CollaborationPatterns............................................174.3Interface Requirements...............174.4Security and TrustMechanisms.....................................184.5Alignment and Gap Analysiswith Existing Standards................21 FIGURES.........................................................40 ACKNOWLEDGEMENTS...................41 EXECUTIVE SUMMARY Mobile network operators (MNOs) now face a decisivenetwork automation inflection point. Rule-basedautomation and machine learning (ML) tools haveimproved efficiency and enabled networks to becomepartially or conditionally autonomous, characterised byautomated tasks in isolated domains, AI/ML assistinghuman decision making and humans in the loop atleast for approvals. However, the current situationremains insufficient to effectively and efficientlyautomate dynamic, cross-domain high-value use cases,given the growing complexity of mobile networks andhigh expectations regarding customer experience.The next major target in our industry is Level 4autonomy as per TM Forum [7] where networksare highly autonomous through proactive, within-domain and cross-domain intent-driven operationsrequiring minimal human intervention while enablingautomated decision making, self-optimisation,self-healing and self-management. assurance, security and cost control, the vision is neitherachievable nor sustainable for network operators. Inaddition, automation with Agentic AI is only feasible atscale if interoperability is supported by suitable standardsfor agentic communication. In this publication, the vision is motivated and thenexemplified using three illustrative high-value scenarios.For decision makers, this complexity is reduced throughan operator-centred reference framework that identifiesthe core capabilities, functions and enablers required fornetwork automation with Agentic AI. From an exemplarygap analysis across the industry the publication derivesimportant considerations for the whole ecosystem includingstandards organisations. The strategic conclusion is clear: commercial-scaleadoption will depend less on aspects like individualAI model capability than on ecosystem alignment, reductionof fragmentation risks and readiness of support systems forAgentic AI. To name a few, operato