Case Study: AT&T achieves more accurate, efficient AI agents with NVIDIA NeMo and NIM

AT&T improves AI agent accuracy by up to 40% with NVIDIA

AT&T, one of the world's largest telecommunications companies, faced challenges in scaling its AI-powered customer service agents. Issues with model accuracy drift from frequently updated documents, high computational demands, and the need to reduce latency and operational costs led them to seek a solution. They turned to vendor NVIDIA and its NVIDIA AI Enterprise platform, including NVIDIA NeMo microservices and NVIDIA NIM.

NVIDIA's solution involved implementing a data flywheel using NeMo microservices for continuous data curation, model fine-tuning, and evaluation. This approach allowed AT&T to deploy optimized, lightweight models. The results included a 40% improvement in AI agent response accuracy and an 84% decrease in call center analytics costs, achieved while maintaining high performance and stringent compliance standards.

AT&T