BACKGROUND In the fiercely competitive telecommunications industry, customer experience has become the core metric for Communications Service Providers (CSPs) to consolidate market share and achieve sustainable development. According to the GSMA 2025 Mobile Economy Report, global individual mobile user penetration will only register marginal growth over the next five years. This compels operators to shift their strategy from scale expansion to experience refinement, retaining users and driving growth by delivering superior service experience. CHALLENGE However, with the rapid rise of AI applications, ultra-high-definition video, live streaming, cloud gaming and other services, user expectations for network experience continue to escalate. Traditional network optimization only focuses on conventional network KPIs, and suffers from obvious limitations in user experience optimization — including difficult analysis workflows, long turnaround time and unsatisfactory improvement results. SOLUTION To address the core demands of telecom operators, this project launches the industry’s first vertical large model solution for network optimization built on the AI-Native Framework. Oriented toward tangible business outcomes and driven by digital twins together with domain-specific professional models, it enables a paradigm shift in operations and maintenance: transforming from humans executing and closing tasks with tools to humans supervising and augmenting AI Agent-driven task execution and closed-loop management, ultimately evolving toward the advanced stage of Agentic Operations. The solution improves key performance indicators including Net Promoter Score (NPS), user churn rate, top-tier OTT experience and complaint handling duration. It also deeply integrates into operators’ daily operational scenarios, builds a digital employee AI assistant matrix empowered by domain-specific large models, and delivers end-to-end intelligent upgrading of O&M operations. Built on China Telecom Qiming Large Model foundation, the solution leverages full-scale experience data collected by intelligent radio network elements, and incorporates a newly upgraded radio network optimization domain-specific large model. Combined with scenario-based capabilities of digital employees, it drives the transformation of network optimization logic from expertise-driven to data-and-AI-driven, shifting optimization priorities from pure network indicators to end-user experience. Timely optimization of dynamically changing wireless networks has long been a persistent industry challenge. As the core engine of the solution, the radio network optimization domain-specific large model relies on full-scale data collection and automatic document aggregation capabilities of intelligent radio network elements. It adopts two Transformer-based domain professional sub-models to form a closed-loop service experience optimization system:The User Experience Diagnosis Large Model (UELM) accurately identifies user experience anomalies through intelligent analysis of full-dimensional data from intelligent wireless networks.The Beam Space Large Model (BSLM) automatically completes problem diagnosis, simulation deduction and global optimization scheme generation. Optimized policies are directly delivered and enforced on live networks via digital employee assistants, achieving immediate network performance improvement and effective mitigation of potential complaint risks. This fully materializes the vision of shifting from users adapting to networks to networks proactively adapting to users. IMPACT Empowered by the disruptive design of the AI-Native Framework, the project delivers remarkable results across three dimensions to form a closed-loop value system: Outcome Delivery: Centering on core user demands, it achieves breakthrough improvement in key experience indicators. The average processing time of user complaint tickets is reduced by 37.6%, quality optimization indicators rise by 25%, and overall customer satisfaction is greatly enhanced. Domain-specific Digital Twin & Large Model Empowerment: Endowed with in-depth understanding of network mechanisms via the radio optimization vertical large model, the solution derives globally optimal overall schemes through simulation. It significantly reduces repetitive manual analysis and on-site testing, cutting on-site verification workload by over 50% in pilot areas, boosting radio optimization efficiency by more than 10 times, and substantially lowering operational costs. Operational Paradigm Upgrade: The entire user complaint handling process is greatly streamlined, enabling VIP complaint resolution within one hour. The automatic processing rate of poor-quality network optimization tickets increases by 11.2%, realizing the upgrade from passive problem response to proactive prediction and autonomous remediation. TM Forum Assets Used & Contributed Aligned with the TM Forum Open Digital Architecture (ODA), this project has deeply contributed to multiple TM Forum AIOps and DT4DI standards through practical implementation. It clarifies the core value and key application path of Domain Digital Twin (DTN) and domain large models within the AI-Native Framework, providing a replicable and scalable practical model for the AI-Native transformation of the global telecommunications industry.