Register here
In collaboration with ITU, GSMA, ETSI, IEEE, ETI GenAINet, and TM Forum, we are pleased to announce the launch of the first global challenge dedicated to the design of AI Agents for telecom troubleshooting.
The challenge will leverage a dedicated, multi-vendor Agent Sandbox capable of modeling realistic, real-world telecom scenarios, together with a central orchestration framework that:
- Exposes structured questions to participating agents
- Connects agents to domain-specific Tools APIs
- Coordinates interactions between agents and the sandbox environment
Participants will compete for significant monetary prizes and will be invited to present their solutions at MWC Shanghai 2026 (24–26 June 2026).
Throughout the challenge, GSMA, ETSI, ITU, IEEE, and TM Forum will provide continuous support, offer a collaborative platform for discussion and interaction, ensure a smooth and transparent evaluation process, and disseminate the competition results within their respective AI and standardization initiatives.
About the Challenge:
Modern telecom networks are evolving toward highly dynamic, software-driven infrastructures, where failures can rapidly cascade across layers and services. Traditional network operations—largely reactive and manual—are increasingly inadequate to manage this complexity, resulting in extended outages and rising operational costs.
Recent advances in AI Agents enable a paradigm shift in network operations: from reactive troubleshooting to proactive, self-healing, and continuously optimized networks. By leveraging multi-turn tool use and complex reasoning, AI Agents can provide 24/7 monitoring, rapidly diagnose issues, and autonomously trigger recovery actions—enabling truly intelligent network management.
The AI Telco Agentic Challenge focuses on IP Networks and targets core operational tasks essential for autonomous networking, including:
- Topology Restoration
- Path Query
- Fault Localization
Participants will design agentic systems capable of reasoning over network state, interacting with tools, and adapting to unseen fault conditions in real time.
To rigorously evaluate progress, the challenge relies on a dedicated, robust, interactive, and industry-specific benchmark, enabling systematic assessment of:
- Accuracy in problem diagnosis and resolution
- Efficiency in multi-step reasoning and tool usage
- Robustness across diverse and evolving network scenarios
By addressing these challenges, participants will contribute to:
- Reducing network downtime and operational costs through autonomous, self-healing operations
- Advancing agentic AI systems that generalize across real-world IP network environments
- Enabling scalable, resilient, and always-on network operations powered by reasoning-driven AI
- Exploring efficient and deployable models that lower barriers to real-world adoption
Learning Objectives:
By the end of this challenge, participants will be able to:
- Identify the roles of international standardization bodies (ITU, GSMA, ETSI, IEEE, TM Forum) in fostering AI-driven telecom innovation
- Design agent workflows that enhance multi-turn tool usage and complex reasoning capabilities
- Apply fine-tuning techniques to off-the-shelf Large Language Models (LLMs) for telecom troubleshooting tasks
- Analyze agent performance using accuracy-, efficiency-, and robustness-based evaluation criteria defined by the organizing bodies
- Design and demonstrate a prototype or workflow showing how LLM-powered agents can enhance automated fault detection and resolution in telecom networks
Recommended Mastery Level / Prerequisites:
This challenge is recommended for participants with intermediate to advanced knowledge of AI agents and LLMs, along with a basic understanding of telecom networks. Prior experience with Python programming and collaborative development platforms such as GitHub is desirable.
