According to a case study published by OpenAI, Ringg — a voice and chat agent platform serving large consumer businesses in India — says its AI agents resolve up to 65% of incoming requests on their own. The agents handle more than 7 million connected calls each month, and customers give them an average CSAT score of 4.8.
Ringg routes different tasks to different OpenAI models: GPT-4.1 handles most real-time voice and chat traffic, GPT-5.6 Luna is used when its price-performance profile fits better, GPT-5.6 Terra runs post-call summaries and sentiment classification, and GPT-5.6 Sol supports evaluation and prompt improvement. The company says migrating suitable real-time workloads from GPT-4.1 to GPT-5.6 cut model costs by roughly 90% while maintaining required quality and latency.
Ringg tests models on historical conversations and simulated flows before production. In one evaluation, GPT-5.6 Terra beat Gemini 2.5 Flash for post-call analysis and achieved up to 97% accuracy on common regional languages, which led Ringg to adopt Terra for those tasks. Because the source is OpenAI's own customer story, the numbers are vendor-reported rather than independently verified.