Game-X is Game Changing Autonomous Networks eXperience Game X - Phase II builds on the success of Phase I, where we demonstrated “zero-wait, zero-touch, zero-trouble” operations and a 30% reduction in OpEx through intent-driven autonomy, to address one of the most significant market shifts of the decade: the explosive rise of AI workloads. As AI becomes a top IT investment priority and is projected to dominate global network traffic, CSPs must evolve beyond pure connectivity. Phase II introduces a new vertical, Artificial Intelligence Network Services, positioning service providers as essential enablers in the multi-trillion-dollar AI economy. This phase also introduces an innovative agentic-based orchestration using distributed reasoning that provides an additional level of autonomy for Eligibility & Quoting, Planning & Reservation, Activation, and Assurance. Together with the use of digital twin technologies and advanced closed-loop automation, the Catalyst defines a new category of services optimized for AI-driven applications and agent traffic—particularly across distributed training, inference, and WAN AI flows. We will focus on the “Model-as-a-Service" set of products as described in TMF GB1085. With “Model-as-a-Service", the service provider can become a true gateway for inference traffic, tapping into the large AI Cloud Infrastructure spending from B2B customers. At the heart of Phase II are two dimentions: the Demand side is represented by two actors: the IT department who sets the interconnection intent (sovereignty, budget) and the AI agents themselves run by developers. These AI Agents require a certain level of SLA both from the network (bandwidth, latency, packet drop) and the LLM layers (availability, time to first token, etc.). The Supply dimention represents the CSP’s autonomous network, dynamically provisioning AI-optimized slices and compute resources. A customer needs to consume an LLM service hosted in the operator’s footprint or in a partner data center, with specific constraints on latency, sovereignty, model availability, and SLA. The full-service lifecycle spans four phases: Eligibility & Quoting, Planning & Reservation, Activation, and Assurance. The new “Agentic Assurance” closed loops continuously monitor AI KPIs, adapt configurations, and preserve model performance—delivering superior customer experience and operational efficiency. By integrating autonomous operations with AI workload needs, Game X Phase II shows how CSPs can claim a differentiated role in the AI value chain, offering hyper-personalized, monetizable services instead of remaining best-effort transport providers. The Catalyst will define new AI service ontologies and introduce AI-focused KPIs to measure success, demonstrating how service providers can reuse TM Forum standards and autonomous capabilities to unlock new revenue, deliver predictable performance for critical AI applications, and secure a strategic position in the evolving AI market. TMF Resources: GB1085 Model-as-a-Service (MODaaS) GB1064 Closed Loop and Artificial Intelligence (AI) TMF921 Intent Management IG1253 Intent in Autonomous Networks IG1218 Autonomous Networks Business Requirements and Framework IG1252 Autonomous Network Levels Evaluation Methodology IG1256 Autonomous Network Effectiveness Indicators GB900 ODA Blueprint Proposed assets: Distributed Reasoning Agent: A functional building block that executes reasoning protocols by invoking reasoning primitives exposed by IAPs across multiple domain systems, without absorbing their authority. Standardizing the Distributed Reasoning Agent would cover its role and boundaries (orchestrate, do not decide), its required interfaces to IAPs and to the conversational or intent layer, and the contracts it must honor for observability, governance, and security — OpenTelemetry traces, decision lineage, policy enforcement, and access control. Reasoning Protocol: A structured method that governs how a Distributed Reasoning Agent selects, sequences, and interprets reasoning primitives to fulfil a goal. A reasoning protocol defines the required inputs, the order and conditions under which primitives are invoked, the composition of logic for partial results, and the decision of policy that produces the outcome. Standardizing reasoning protocols would allow CSPs and partners to share, compose, and audit agentic workflows across domains in the same way standardized API choreographies to enable today's integrations.