Ringg uses OpenAI models to automate up to 65% of customer requests
Ringg, a voice and chat agent platform serving consumer businesses in India and other markets, says its AI agents resolve up to 65% of routine customer requests without human involvement. The company handles more than 7 million connected calls per month and reports an average customer satisfaction score of 4.8.
Its platform supports voice, chat, WhatsApp and web interactions. Ringg uses OpenAI models to interpret requests, select tools and guide customers through multi-step workflows such as checking a policy, retrieving account data, scheduling an appointment, updating a CRM record or transferring a case to a specialist.
Model routing for customer operations
Ringg’s orchestration layer connects agents to CRMs, ticketing platforms, payment systems, scheduling tools and internal APIs. When an escalation is necessary, the system passes a conversation summary to a human agent. Its knowledge system combines structured filtering and semantic retrieval across datasets, PDFs, CSVs and business documents.
The company assigns models according to task requirements. GPT-4.1 handles most real-time voice and chat traffic, while GPT-5.6 Luna remains in production for requests where its performance, latency or price-performance profile is more suitable. GPT-5.6 Terra is used for post-call summaries and sentiment classification, and GPT-5.6 Sol supports evaluations, prompt improvement and model-as-judge workflows.
For longer conversations, Ringg creates a structured summary as context approaches about 80,000 tokens. This lets an interaction continue while preserving relevant information rather than repeatedly transmitting its complete history. The approach complements enterprise adoption trends reflected in OpenAI’s ChatGPT Work user growth as organizations expand AI tools into operational workflows.
Testing, resilience and economics
Before deployment, Ringg tests models with historical conversations and simulated customer flows. It also evaluates language and regional variation, including exchanges that mix English with local-language phrases. Ringg says GPT-5.6 Terra outperformed Gemini 2.5 Flash in its tests, reaching up to 97% accuracy on common regional languages.
Models that pass offline evaluation first receive a small share of production traffic. Ringg monitors latency and endpoint health across regions, shifting traffic when an endpoint is unavailable or exceeds a latency threshold. Specialized nodes, alerts and versioned deployments are intended to isolate issues and limit their effect.
Ringg says migrating selected real-time workloads from GPT-4.1 to GPT-5.6 Luna reduced model costs by approximately 90% while meeting its required quality and latency. The company also reports customer outcomes at Policybazaar, Practo and Groww, including faster response times, appointment bookings and self-service handling of investment queries.
Browser agents are the next extension
Using OpenAI computer-use capabilities, Ringg is developing browser agents for onboarding, Know Your Customer processes, IT troubleshooting, on-call incident support and claims processing. It is also building a context layer intended to retain information as a customer moves between voice, WhatsApp and a browser session.
The practical implication for businesses is to assess service automation in terms of completed customer tasks, reliable integrations, response quality and effective human escalation. Model selection, production evaluation and channel continuity are operational components of that assessment, not separate implementation details.