Driving Telco Innovation: Distilling Large Models Into Economical Small Language Models for Agentic Workflow Automation

Discover how the largest telecom company in the United States fine-tuned open-source models to create LLM-driven customer-agent interaction categories, dramatically enhancing customer service and reducing costs. We’ll explore how our approach achieved an 85% cost reduction, while seamlessly transitioning from OpenAI with no compromise in performance or quality. By distilling large language models into efficient small models, we optimized telecom AI without sacrificing effectiveness. These interaction categories now play a critical role in customer churn models, identifying service gaps, enabling proactive outreach, and reinforcing the company's commitment to a customer-first strategy.