Skip to content

Home

Automated Network Incident Management with AI Agents and Knowledge.
– Incident Manager -
A solution that achieves full automation of network failure response operations through data utilization. It automatically structures unorganized data and knowledge into AI-friendly formats, dynamically generating failure analysis and resolution workflows using this data. It enables expert-level rapid analysis and resolution even for unknown failure events.

Challenges in Network Operations Automation

The introduction of AI agents is accelerating the automation of network operation by autonomously integrating tasks previously performed by people or tools. However, two major challenges remain for advancing further automation: how to organize diverse data on past and current incidents, along with analysis and response procedures, into knowledge that AI can effectively utilize; and how to derive analysis and response methods for unknown incidents without precedents.

With Incident Manager

ネットワーク運用自動化における課題
From various types and formats of information—such as emails and chat exchanges between operators, and established analysis procedure manuals—we extract meaningful data like fault classifications and root causes. This structured knowledge is then organized into a format easily interpreted and utilized by generative AI. Furthermore, based on the causal relationships between this structured knowledge, it dynamically constructs analysis and response workflows. Then, the agent autonomously repeats event verification and judgment while executing analysis and response workflows for unknown events. Incident Manager further accelerates the automation of network incident response.

Values Brought by Incident Manager

  1. Contributing to the further enhancement of customer organization's capabilities.
    • AI-driven automation reduces response time during network failures.
    • By having AI perform these tasks, the psychological stress and physical burden on operators are reduced.
  2. Achieving data-driven operational management contributes to well-being.
    • By automatically converting tacit operational knowledge and experience into data, it eliminates reliance on individual expertise.
    • Continuously improve operational methods and systems through feedback from AI.

Technical Summary

Target Industry & Users

  • Telecommunications carriers providing public communication services, including 5G, and the operators managing their networks.
  • Businesses that build and operate their own enterprise networks, such as financial and power networks, and the operational administrators of those networks.

Challenges in Target Industry & Tasks

  • Diversification and expansion of services, along with the resulting high SLAs
    • Technological advancements have led to increasingly large-scale and complex networks, while the services they provide have become more diverse and sophisticated. Networks are now required to deliver high stability and robustness as critical social infrastructure, making rapid response and recovery during failures an urgent priority.
  • Limitations of Operational Automation
    • While automation has been implemented for analysis and response procedures established based on past cases, unknown events still require human intervention, leading to time-consuming resolution and recovery.
  • Increased maintenance costs for automation rules and training costs for operators
    • As networks become larger and more complex, the costs of updating and creating new automation rules are rising, while developing new experts capable of handling these challenges has become a major issue.

Technical Challenges

To further advance AI-driven operational automation, we believe two key challenges must be addressed:

  • Technology to transform information about past and current events —documented in various formats and styles— into knowledge that AI can interpret and utilize.
  • Technology to rapidly derive appropriate analysis and response procedures even for unknown events.

Solution

Automatic metadata generation technology

Fujitsu and 1Finity leverage generative AI technology distilled from their accumulated network operation experience and knowledge. This technology automatically extracts the metadata (such as alarm type, location of occurrence, cause, etc.) necessary for analysis and response from information and procedures documented in various types and formats, structuring it into AI-friendly knowledge.

Dynamic extracted knowledge graph technology

This technology dynamically extracts and generates a knowledge graph for the event that has occurred by utilizing the aforementioned structured knowledge, enabling the accurate and rapid derivation of analysis procedures and countermeasures.

Fujitsu's Technical Advantage

Generative AI utilizes knowledge structured in an AI-friendly manner to generate analysis and response procedures. By dynamically extracting knowledge graphs relevant to the occurring event, it eliminates potentially noisy, unnecessary information. This enables the rapid derivation of accurate analysis and response procedures for various events.

Use Cases

  • End Users:
    • A solution used by network operators to execute tasks from root cause analysis to resolution during network outages.

Examples & Case Studies

  • About the Proof of Concept
    • We conducted effectiveness verification targeting "root cause analysis" and "countermeasure planning" with communication service providers, utilizing the "incident analysis function" and "document search function." The analysis results, response quality, and usability achieved evaluations meeting practical levels.
    • Moving forward, we plan to advance more practical verification by expanding the scope of automation in incident response operations, alongside planning to commercialize the Incident Manager product in the second half of fiscal year 2026.

Trial of Technology