Ringg, a voice and chat agent platform, uses OpenAI models — including GPT‑5.6 — to power multilingual agents across voice, chat, WhatsApp and web channels. The company says migrating certain real‑time workloads from GPT‑4.1 to GPT‑5.6 reduced model costs by about 90% while preserving required quality and latency.
Why they built the system
While working with large consumer businesses in India, Ringg observed that rising call volume typically led to hiring more staff, increasing per‑interaction cost and complexity. Agents at these companies often had to operate across fragmented manual systems, making tasks like buying insurance or booking appointments slow and error‑prone.
To address this, Ringg developed an enterprise agent platform that pairs efficient models with an orchestration layer to execute actions across CRMs, ticketing systems, payment providers, scheduling tools and internal APIs. The platform can escalate to a human agent with a conversation summary when needed.
How it works
- For real‑time interactions Ringg uses GPT‑5.6 Luna and other models to interpret customer requests, select tools, and guide customers through multi‑step workflows. GPT‑4.1 handles most real‑time voice and chat traffic, while GPT‑5.6 Luna is used when its performance, latency or price‑performance profile is better suited to a request.
- GPT‑5.6 Terra is used for post‑call analysis such as summaries and sentiment classification; GPT‑5.6 Sol supports evaluation, prompt improvement and model‑as‑judge tasks.
- The knowledge system combines structured filtering with semantic retrieval across datasets, PDFs, CSVs and business documents to surface relevant enterprise information for agents.
- Tasks can be split among specialized subagents (qualification, support, verification, scheduling, escalation) while maintaining a consistent conversation with the customer across channels.
- For longer conversations, the system creates a structured summary when context approaches roughly 80,000 tokens so important information is preserved without resending the full history.
Testing, evaluation and reliability
Ringg evaluates models using historical conversations and simulated customer flows before production rollout. Their evaluation platform identifies weaknesses and recommends prompt improvements, creating a continuous improvement loop. Testing covers language and regional variations, including code‑switching between English and local phrases.
In one evaluation, GPT‑5.6 Terra outperformed alternatives such as Gemini 2.5 Flash for post‑call analysis; as a result Ringg moved summaries and sentiment classification to Terra. According to Ringg, Terra maintained high accuracy on summaries and sentiment analysis while materially improving unit economics.
Models that pass offline testing are introduced to a small share of production traffic. The production router monitors latency and endpoint health across regions, shifting traffic when an endpoint becomes unavailable or exceeds latency thresholds. Versioned deployments, specialized nodes and alerts help isolate issues and limit impact.
Results and customer examples
- Ringg’s agents handle more than 7 million connected calls each month, and Ringg customers have an average customer satisfaction (CSAT) score of 4.8.
- The platform resolves up to 65% of routine customer inquiries without human involvement.
Customer examples provided by Ringg:
- Policybazaar: over 57,000 customer requests are connected through Ringg, with 67% of calls handled without human intervention. Average response time fell from 8–12 minutes to under 60 seconds (~88% improvement).
- Practo: Ringg agents assist with healthcare appointment bookings and service requests; after deployment Practo achieved an 85% first‑call resolution rate and response times under three seconds. Operating costs fell by 70% compared with its prior human‑led workflow, and Ringg now completes more than 1,000 appointment bookings per day for Practo.
- Groww: the online investment platform resolves 72% of inbound queries related to IPOs, futures and options via self‑service, with an average handling time of two minutes.
Future work and outlook
Ringg is developing browser agents using OpenAI’s capabilities for platform onboarding, Know Your Customer (KYC) processes, IT troubleshooting, on‑call incident support and claims processing. The company is also building a context layer to preserve information across channels and interactions so a customer can start a request by voice, continue on WhatsApp and finish in a browser without repeating details.
Ringg says the next generation of customer operations will be measured by completed business outcomes and automation depth rather than call volume or headcount.
