Gupshup has launched a self-serve Voice AI platform that allows enterprises to design, deploy and monitor AI agents for inbound and outbound calls. The product extends the company’s conversational engagement infrastructure beyond WhatsApp, RCS and SMS, bringing voice interactions onto a common platform for customer support, collections, sales qualification, verification and scheduling.
The platform is designed primarily for repetitive calls that can be resolved without continuous intervention from a human representative. It can answer support questions, qualify prospective customers, issue payment reminders, recover abandoned transactions, schedule appointments and conduct identity-verification workflows. Founder and chief executive Beerud Sheth has positioned the product as a way to move routine conversations away from contact-centre employees while retaining human agents for cases requiring judgement or more complex handling. Launch details published September 1 describe the platform as an extension of Gupshup’s existing enterprise communications stack.
Businesses can configure an agent’s voice, language, knowledge base, system instructions and workflow without building the underlying speech infrastructure. The system supports phone networks, SIP trunks, WhatsApp calling and browser-based voice interactions. Companies may retain their existing carrier or telephony infrastructure, connect services such as Twilio and Plivo, or use managed numbers provided through Gupshup-owned Knowlarity.
The platform supports more than 90 languages and dialects, including dedicated capabilities for 11 Indian languages. A visual workflow builder enables business teams to create branching conversations, connect enterprise tools and configure transfers to human representatives. Agents can retrieve answers from a company’s documentation and invoke external functions or Model Context Protocol servers, with approval controls attached to actions that require additional oversight.
Gupshup is allowing customers to select their preferred speech-to-text, text-to-speech and large-language-model providers rather than binding deployments to a single model vendor. This lets enterprises evaluate the trade-offs among response speed, transcription accuracy and operating cost. The company states that typical end-to-end response time is between two and three seconds.
Enterprise controls include per-agent guardrails, scenario simulations, automated testing and recurring checks designed to detect behavioural changes after prompts or knowledge bases are updated. Every call can produce a transcript, recording, conversation history, debug log and breakdown of latency and cost. Post-call analytics cover customer intent, sentiment, outcome and language use, giving operations teams a basis for reviewing performance and escalating failed conversations.
The product specification also supports warm or cold transfer to human agents during a live call. Gupshup says its broader communications platform serves more than 50,000 customers in over 130 countries and handles more than 120 billion messages annually. Voice AI adds a real-time, higher-complexity interaction layer to that messaging footprint while retaining the company’s existing focus on enterprise marketing, commerce and support automation.
