变革推动者：将人工智能集成到客户体验运营中

Agents of change:integrating AI intocustomer experience Contents 3The big picture 7Key findings 8Section 1:Customer engagement evolves with generative AI (GenAI)14Section 2:How will agentic AI transform customer operations?19Section 3:What are the roles for Model Context Protocol (MCP) and Agent2Agent Protocol (A2A)?23Section 4:Where’s the value in agentic AI for customer experience? The big In September 2024 we published a comprehensive report aboutAI in customer experience (CX), but nowhere in it is the phrase “agenticAI” used. What a difference a year makes. Today, it’s impossible to Agentic AI stands to revolutionize customer experience,going well beyond the capabilities of generative AI(GenAI) chatbots. As TM Forum Chief Analyst MarkNewman explains in our recent agentic AI Benchmarkreport: “Rather than putting GenAI into the hands ofcontact center agents, agentic AI can move beyond the‘chat interface’ trigger point to one that is activated by, But telcos have two important competitive advantages:their mobile networks and devices, and their long-standing relationships with customers built over yearsof providing highly reliable services. Unlike generalAI platforms that operate through apps or webinterfaces, telcos can embed AI assistants directly intosmartphones and SIM cards. This allows for real-time, Read the reports: Identifying the challengesAgents, in the context of agentic AI, are essentially Leading communications service providers (CSPs)like Deutsche Telekom and SK Telecom are goingbeyond chatbots in the consumer market to developsophisticated AI personal assistants that aim toaccomplish a wide range of tasks for customers,such as planning and booking travel autonomously.They aren’t the only companies doing this, of course. pieces of software that can act autonomously – inessence taking the place of a human agent makingdecisions and acting upon them. Full autonomy incustomer experience operations is possible only ifCSPs can securely integrate customer-facing Legacy OSS/BSS weren’t designed to work with modernAI tools. Many of the systems are hard to integrate,lack real-time access to data and don’t support cloud-native technologies. Access to high-quality data isone of the biggest obstacles, if not the biggest. AIrequires unified, data-as-a-product, but within telco On top of the technical challenges, CSPs’ limited AIexpertise and cultural resistance to change makesystems integration even more difficult. And thenthere is the cost. The business value of generative and The graphic oppostite shows how CSP respondents toour agentic AI Benchmark survey ranked these and other “There are some challenges that we need to addressbefore we are ready to scale [agentic AI], because weare talking about agents – effectively software – thattake decisions,” says Dr. Nektaria Efthymiou, NetworkPlatform & Security Director, BT Group. “We need tomake sure that there are guardrails. We need to make What’s in the report?But even in the face of difficult challenges, operators Read the report to understand: are optimistic they will be successful in deployingagentic AI. This report looks at how CSPs intend to •How customer-facing systems are evolving with AI•How CSPs believe agentic AI will transform customeroperations•Why standards are needed to solve the biggestintegration challenges as operators deploy agentic AI•How CSPs expect to use Model Context Protocol “We need to make surethat there are guardrails...Because it’s not about AIsitting over the top of the Dr. Nektaria Efthymiou, BT In 2024 CSPs spent$90 billion on labor relatedto customer experience,making automation Legacy OSS/BSS systems,fragmented data andcultural resistance are major High-quality, unifieddata and cloud-nativeinfrastructure are Section 1 Customerengagementevolves withgenerative AI Improving customer experience is the top GenAI and agentic AIuse case. The reason is simple: Satisfied customers are more likely Reducing costs is also a major driver for using AI toimprove CX. “Customer care is an enormous cost to thebusiness,” says Eoin Coughlan, Global CTO for Telco, Indeed, in 2024 CSPs collectively spent about$90 billion on labor related to customer experience,according to research firm Omdia.This includes staff Implementing GenAI chatbotsTM Forum’s Generative AI Maturity Interactive Tool (GAMIT) survey, conducted at the end of 2024 andearly in 2025, shows that CSPs are putting AI chatbotsinto production at a faster pace than other types ofGenAI applications. A full 80% of CSPs said they are IBM's recent research supports these findings.The company’s Institute for Business Value surveyed106 CSP executives globally in 2025 about their AIstrategy, initiatives and outcomes and found that 69% Agentic AI in CX Software agents are not new in telecoms, butgenerative and agentic AI systems are. In ourBenchmark report, we define agentic AI as an AIservice that interacts with a large language model(LLM) and is capable of