Voice-Bot Site-Visit Booking: Multilingual Qualification for Indian Brokerages

The 35% after-hours enquiry leak

Most Indian property brokerages staff their phones from 10am to 7pm. The data from twelve brokerages we worked with in 2024 shows that 35-42% of inbound enquiry calls land outside this window, with a peak between 8:30pm and 10:30pm — the time families finish dinner and start their property research. Every one of those calls that goes to voicemail is a coin flip: the family will either call back the next morning, or they will move on to the brokerage whose voice bot picked up the call.

Language detection in the first three seconds

A voice bot for Indian property cannot assume English. The first three seconds of the call should detect the caller's language preference — usually from the language of the opening 'hello' or the response to the bot's bilingual greeting. The bots we have seen succeed support at minimum Hindi, English, and one regional language per geography (Tamil in Chennai, Telugu in Hyderabad, Marathi in Pune, Bengali in Kolkata). Trying to start every call in English loses the family before they have asked their first question.

• Bilingual greeting that probes language preference within 3 seconds
• Mid-call language switch supported — caller may begin in English and slip into Hindi mid-sentence
• Regional accent variants trained for at least the top 4 South Indian and 3 North Indian dialects
• Number recognition robust across language switches (budget figures often spoken with mixed code)
• Fallback to a designated human agent if the detected language is outside the supported set

The five-question qualification script that converts

A voice bot has a narrow window to qualify a property lead — about 90 seconds before the caller starts feeling interrogated. The five questions that fit that window: budget range (so the bot can immediately surface RERA-matching projects), preferred locality (so the bot routes to the right project portfolio), end-use (self-occupation vs investment, which changes the project filter sharply), timeline (immediate vs 6-month, which determines urgency tier), and family-size (which determines BHK and the kind of amenity to highlight). Anything beyond these five turns the call into an interview.

When the bot must hand off to a human

The single biggest mistake voice-bot implementations make is trying to close the entire transaction. A bot should book the site visit and end the call — not negotiate the price, not discuss home-loan options in depth, not commit to special discounts. The handoff triggers we have locked in: any pricing negotiation, any home-loan eligibility question, any specific legal-title query, and any time the caller explicitly asks to speak to a human. The bot's last-call action should always be a WhatsApp message confirming the site-visit slot and naming the human counsellor who will follow up.

RERA disclosure by voice — done correctly

RERA mandates that the registration number be communicated in every project promotion — including voice calls. A voice bot promoting a project must speak the full RERA registration ID at least once during the call. Brokerages that skip this disclosure are at risk under Section 61. The bot script should also send the same RERA ID by WhatsApp immediately after the call so there is a written record of the disclosure. This is one of the easier RERA compliance wins because it requires a single script update rather than a process change.

Post-call follow-up that closes the loop

A voice-bot site-visit booking is incomplete without three follow-up touches: an immediate WhatsApp confirmation with the address, route, and counsellor name; a same-day evening reminder if the visit is the next morning; and a post-visit recap call (this time from a human) within 24 hours. Brokerages that automate the first two and assign the third to a named counsellor see 28% higher visit-show rates compared to brokerages that rely on a single confirmation SMS.