How Frontier Communications uses Amazon Bedrock to transform customer insights from call transcripts

This article describes how Frontier Communications built Giga-T, a generative AI application using Amazon Bedrock to analyze 125,000 monthly call center transcripts and extract customer insights.

• Processes 6,000 daily calls in six hours versus ten months manually; 416x faster analysis
• Classifies calls into 50 mutually exclusive categories to identify customer intent and issues
• Achieved 70%+ precision and recall rates on call intent classification in testing
• Scaled from proof-of-concept to full production analyzing all Retention & Loyalty calls daily
• Uses Amazon Bedrock with Claude models, S3, Step Functions, Lambda, and EventBridge
• Helped achieve record 83% save rate in 2024 within Retention & Loyalty group
• Trained 14 prompt engineers company-wide to extend Giga-T across other call centers
• Key success factors: startup mentality, human-in-the-loop approach, AWS expert partnership

Frontier's Giga-T demonstrates how generative AI can transform unstructured call data into actionable business insights at scale, improving customer retention and agent performance measurement.