The Problem Ringg Set Out to Solve
Ringg builds voice and chat agents for large consumer businesses in India. The company says it saw the scaling problem firsthand while working with these clients: as call volumes rose, the standard response was to add more human agents, which increased both cost and complexity with every new hire. On top of that, existing customer service teams were often working around fragmented, manual systems just to complete basic tasks, from processing an insurance purchase to booking a healthcare appointment. That experience pushed Ringg to build an enterprise agent platform rather than a simple chatbot layer.
How the Platform Works
Ringg's agents don't just answer questions, they act. The platform connects a conversation to account records and business software, so an agent can complete a booking, update a system, or escalate to a human with a full summary attached. Underneath, a knowledge system combines structured filtering with semantic retrieval across datasets, PDFs, CSVs and other business documents, so agents can pull relevant information rather than relying purely on a general model's training. For longer conversations, the system compresses history into a structured summary of roughly 80,000 tokens, preserving key details across multi-step workflows without resending an entire transcript each time.
Why OpenAI, and What It Costs
Ringg evaluated OpenAI's models against alternatives, including Gemini 2.5 Flash, across conversational quality, latency, instruction following, tool calling, multilingual performance, reliability and cost. It concluded OpenAI offered the strongest overall balance for production workloads. In one internal test, GPT-5.6 Terra reportedly outperformed Gemini 2.5 Flash for post-call analysis, reaching up to 97% accuracy on common regional languages, including conversations that blend English with local-language phrases. Migrating real-time workloads from GPT-4.1 to GPT-5.6 is credited with roughly a 90% reduction in model costs for those workloads.
Results Across Ringg's Customers
Three named case studies anchor the claim:
Policybazaar, one of India's largest online insurance platforms, routes more than 57,000 customer requests through Ringg, with 67% resolved without a human. Average response time fell from 8–12 minutes to under 60 seconds, roughly an 88% improvement.
Practo, a healthcare platform, uses Ringg for appointment booking and service requests, reporting an 85% first-call resolution rate, response times under three seconds, a 70% drop in operating costs versus its previous human-led workflow, and more than 1,000 daily appointment bookings.
Groww, an investment platform, resolves 72% of inbound queries about IPOs, futures and options entirely through self-service.
What Comes Next
Ringg is developing browser agents that use OpenAI's computer-use capabilities for tasks like onboarding, KYC verification, IT troubleshooting, incident support and claims processing. It's also building a context layer intended to let a customer start a request on a phone call, continue it on WhatsApp, and finish it in a browser, without repeating information at each step. Both efforts are still in development, not yet live at scale.
A Necessary Caveat
Independent coverage has pushed back gently on the headline figure. Analysts note that 65% describes routine inquiries in Ringg's best-performing deployments, not an average across every call type, and Ringg hasn't published how it defines "routine." Commentators also distinguish resolution from containment, the older industry metric that only tracks whether a caller avoided reaching a human, a number that can look strong even when the underlying problem was never solved. The customer-specific figures, such as Practo's 85% first-call resolution, are more concrete, but readers should treat the 65% topline as an upper bound rather than a guaranteed average.