Swisscom leverages Amazon Bedrock AI to automate network operations

Swisscom, Switzerland’s leading telecommunications provider, has developed an AI-powered Network Assistant built on Amazon Bedrock, aiming to automate and streamline complex network operations. The solution addresses inefficiencies in manual data retrieval and analysis, significantly reducing engineers' workload while improving accuracy.

The Challenge

Network engineers at Swisscom previously spent over 10% of their time manually processing data from multiple sources, leading to potential errors and inefficiencies. Key challenges included:

• Fragmented data retrieval
• Complex KPI calculations
• Scalability issues

The Solution

The Network Assistant leverages Amazon Bedrock’s Retrieval Augmented Generation (RAG) framework:

1. Retrieval – Matches queries with relevant data using embedding models
2. Augmentation – Enriches context with retrieved information
3. Generation – Uses LLMs to provide informed responses

Enhanced Architecture

A multi-agent system improves precision:

• Supervisor Agent – Orchestrates interactions
• Documentation Management Agent – Provides insights from large datasets
• Calculator Agent – Performs complex calculations using natural language

Data security was paramount, with guardrails implemented to filter sensitive information and prevent malicious inputs.

Results

• 10% reduction in manual tasks for engineers (~200 hours saved annually)
• High accuracy in KPI calculations
• Scalable & cost-effective serverless architecture (AWS Lambda, Glue, Athena)

Future Plans

Swisscom aims to expand the Network Assistant with:

• Proactive monitoring using a network health tracker agent
• Automated alerts via Amazon SNS
• Integration with additional network systems

Conclusion

Swisscom’s implementation showcases how AI-powered automation can revolutionize network operations. By combining with robust data pipelines, they have unlocked efficiency, accuracy, and scalability.