Case Study: AT&T achieves 90% cost savings in call center operations with H2O.ai

AT&T Achieves 90% Cost Savings with Fine-Tuned Small Language Models

AT&T, a broadband connectivity provider, faced the challenge of analyzing 15 million customer calls each year to improve service, efficiency, and business insights. To better extract value from these recorded, transcribed, and summarized interactions, AT&T partnered with H2O.ai and used AI and language models to support its call center operations.

H2O.ai helped AT&T distill larger models like GPT-4 into three smaller fine-tuned open-source models, including H2O.ai’s Danube 1.8B built with H2O LLM Studio. The solution achieved 91% accuracy while cutting costs to 35% of the previous approach, with Danube accounting for just 10% of the total, and also reduced processing time while increasing transcript processing capacity.

AT&T