To serve customers effectively and efficiently, CSPs’ IT departments need an accurate real-time view of how their networks are performing. But the network and service resource inventory in a CSP’s operations support system (OSS) often does not reflect the current context of the network topology. This can result in costly service provisioning re-runs, impairments to the service and customer experience, and the need to reconcile data between systems and database, which takes considerable time and effort. This Catalyst is developing a generative AI-based toolkit, encompassing an adaptor agent and large language model (LLM), which will essentially mediate between the systems, processes and data in the network/service inventory database and the corresponding elements of the IP network. The solution should facilitate timely discovery of information across the network, associated network and service elements with a high degree of granularity. The overarching goal is to enhance customer experience by lowering the time needed to provision a service, and eventually enable autonomous network and service orchestration. The Catalyst plans to pilot the solution for fiber access MPLS-based IP-VPN services and 5G access services, with a view to replicating it across other services. The LLM, which will be fine-tuned across data, systems and process artifacts in relation to the network and service, will draw on the TM Forum’s business process framework and information reference model principles (such as ETOM/SID). The TM Forum assets will act as guardrails, supplemented by preventive tests for hallucination, following a significant pre-training process. The adaptive LLM will use a closed feedback loop, ingressing to the main generative AI adaptive engine, to further improve efficacy. On the operational process side, the project team will measure the service delivery time in relation to a baseline approach. Where diverse AI models are used for diverse scenarios, the team will employ A/B testing to evaluate each model’s effectiveness.