Havells Launches Agentic AI Voice Mode Based on ElevenLabs… Supporting 8 Languages
Key point
Havells has launched an Agentic AI Voice Mode in the Havells One app, powered by ElevenLabs technology, enabling users to control home appliances in 8 languages.
Details
Major Indian electrical goods company Havells has launched Agentic AI Voice Mode in its app Havells One. This feature was developed based on ElevenLabs' real-time speech recognition and synthesis technology, supporting a total of 8 languages, including English and 7 Indian languages.
Natural Language-Based Intelligent Control
While existing voice interfaces were command-based, requiring users to memorize specific phrases, the new Voice Mode is designed to understand user intent and context. By integrating Scribe v2 Realtime (Speech to Text) and Eleven v3 Conversational (Text to Speech) via ElevenAPI, it provides low-latency responses and a natural conversational experience.
Users can use natural language instead of fixed commands, and the system can handle code-mixed speech and colloquialisms. For example, saying "Yaar thoda thanda lag raha hai" (Friend, it's a bit cold) will automatically adjust fan speed and air conditioning without mentioning device names.
Multi-Device Orchestration and Context Memory
Users can control 7 device categories, including air conditioners, fans, water purifiers, and washing machines, with a single voice command. Even devices using different connection protocols can be executed simultaneously with commands like "Geyser on karo aur fan bhi chalu karo" (Turn on the geyser and the fan too).
The agent maintains conversation context to understand follow-up commands such as "Fan slow karo... thoda aur" (Slow the fan... a bit more). It also provides bidirectional feedback, such as asking which room is meant instead of guessing when commands are ambiguous, or offering plain explanations and troubleshooting steps when error codes occur.
Development Process and Future Plans
The Havells Centre for Research and Innovation (CRI) took approximately 6 weeks from prototype to production. The first two weeks were dedicated to validating the path from voice command to device actuation, while the remaining period focused on multilingual device catalogs and resolving edge cases. ElevenLabs' engineering team supported pronunciation tuning for Indian regional languages and latency optimization.
The feature is currently being rolled out to all Havells One users, with the app having surpassed 2.7 million downloads. Havells plans to expand functionality in the future with deeper personalization, expanded language coverage, a voice-first gateway, and routines for problem-solving and automation that go beyond simple control.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.