AI Agent Execution Plans for Telecom Network Troubleshooting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud networks with thousands or millions of nodes face challenges in maintaining node and function health, leading to processing delays and increased costs due to inefficiencies in updating and troubleshooting, which can result in revenue loss and customer dissatisfaction.
Innovation Solution
AI agents are used to automate the generation and execution of executable plans from troubleshooting guides (TSGs) and method of operation (MOPs), verifying each operation's correctness to ensure efficient and accurate execution without operator intervention, reducing the need for manual action and LLM usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If manual operator action is used to execute troubleshooting guides and operational procedures, then flexibility and complex decision-making are maintained, but execution time and labor costs increase significantly
Solution Approach 1:
The system enables self-service automation where AI agents autonomously execute operational procedures and troubleshooting guides without requiring human operator intervention. The agent independently parses procedural documents, generates executable plans, performs actions, and verifies outcomes, allowing the system to service itself and reducing dependency on manual operator actions while maintaining high execution speed
Solution Approach 2:
The patent replaces the mechanical system of manual operator execution with an automated AI agent system. The agent uses natural language processing to parse procedural documents, transforms them into structured executable plans, and automatically performs the specified actions. This substitution eliminates the need for human operators to manually follow troubleshooting guides, significantly increasing execution speed while achieving high-level automation
2Adaptability or versatility
If LLMs are used during execution to handle complex decisions, then adaptability is improved, but cost and execution time increase
Solution Approach 1:
The system performs preliminary action by pre-parsing operational procedures and troubleshooting guides into structured executable plans before execution begins. The AI agent analyzes the procedural document, identifies actionable steps, extracts parameters, and creates a detailed execution plan in advance. This preliminary structuring allows the system to handle complex decisions through pre-planned logic rather than requiring real-time LLM invocation during execution, reducing both cost and processing time while maintaining adaptability
Solution Approach 2:
The patent segments the operational procedure into distinct actionable steps with clear decision points. Each step is parsed and structured independently, allowing the system to execute straightforward steps automatically without LLM intervention. Only at specific decision points where complex judgment is needed does the system invoke LLM capabilities, thereby reducing overall LLM usage and associated costs while maintaining the necessary adaptability for complex decisions
3Measurement precision
If comprehensive verification of each operation is performed, then accuracy is improved, but processing time increases
Solution Approach 1:
The system implements feedback mechanisms where the AI agent verifies the output of each operation against expected criteria before proceeding to the next step. The verification process checks whether the actual outcome matches the anticipated result defined in the executable plan. This structured feedback approach ensures high verification accuracy by systematically validating each step while maintaining efficient processing time through automated validation logic that doesn't require exhaustive checking of every possible outcome
Data Source
AI summary
Conditions are identified in a telecommunications network. Based on the condition, a data store is accessed to identify an associated executable plan for responding to the detected anomaly. The executable plan is generated by an AI agent based on a structured operator-readable document comprising operator-executable procedures. The executable plan comprises a series of operations that are executable by an execution component.