Voice AI that calls a property lead while the form is still on screen

An outbound assistant for a residential brokerage. A form submission on a listing page fires a webhook and the call goes out within ten seconds, while the prospect is still looking at the property. The assistant introduces the listing, answers what it can from a knowledge base built for that one property, and works out from the answers whether this is someone ready to view. When they are, it checks the agent's calendar live and books the slot — and a custom date-parsing layer turns 'this Sunday afternoon' into a real timestamp in the property's own timezone, which is the part that is harder than it sounds.

A residential real-estate brokerage

The client

A residential real-estate brokerage

Industry
Real estate & PropTech

Challenge

By the time an agent called back, the lead had moved on

In property, whoever calls first is usually the one who gets the viewing — and agents were taking 2-4 hours to follow up on web enquiries, by which time the prospect had often filled in a form somewhere else as well. Manual scheduling created constant back-and-forth emails, double-booked showings, and no structured record of what a buyer had actually said about their budget or their timeline.

For a high-volume agency handling hundreds of daily enquiries across multiple landing pages and listing platforms, the manual overhead was capping growth at 15% despite 3x lead volume increases. They needed a system that could work at 3am and behave like a senior sales associate — answering questions about the property, working out how urgent the buyer is, and getting a viewing in the calendar without an agent picking up the phone.

Solution

The call goes out while the form is still on screen

We designed a multi-layer automation stack where lead capture triggers (via webhooks from landing pages) launch an outbound Voice AI call within 10 seconds. The AI uses a custom 'Date Normalisation' engine to convert conversational time references like 'this Sunday afternoon' into precise calendar timestamps for instant API-based booking.

The system follows a 'Qualify-then-Book' strategy. The AI introduces the property, gauges interest level, and answers specific questions using a per-listing knowledge base. For hot leads, it performs real-time Free/Busy checks against agent calendars. If the preferred slot is taken, the AI proactively suggests alternatives — mimicking the persuasive flow of a human closer.

Calls we would still defend

  • 10-Second Outbound Trigger — Configured instant call initiation upon form submission to capture prospects at peak interest — before they close the browser tab or call a competitor.

  • NLP Date Normalisation — Built a custom parsing layer handling vague temporal references ('next weekend', 'tomorrow evening') with timezone-aware conversion, ensuring 'tomorrow at 5' correctly maps to 5:00 PM in the property's local timezone.

  • Dual-Path Lead Routing — If a prospect isn't ready to book, the system automatically tags them for a nurture drip campaign in the CRM with full conversation context — converting 'not now' into 'next month' instead of 'lost forever'.

Results

Agents stopped chasing and started confirming

Before

Leads sitting in CRM for 4+ hours. 30% of follow-up calls going to voicemail. Agents spending 15 hours/week on scheduling admin. Double-booked showings causing client frustration.

After

Leads contacted within 10 seconds of the form landing. Viewings booked while the prospect is still looking at the listing. Double-bookings largely gone, because availability is read off the calendar rather than remembered. Agents waking up to appointments already confirmed.

10s

Lead Response Speed

Every enquiry engaged within about ten seconds, day or night

40%

Tour Booking Rate

More property tours booked by replying before the lead cools

5,000$+

Monthly Savings

Coordinator time saved each month, handled without a human in the loop

FAQ

The questions this one raises

What was actually built, the constraints it had to meet, and what we would do differently.

  • Within 10 seconds. The system monitors landing page webhooks in real time and initiates an outbound call the moment a form submission is detected — contacting prospects while they're still actively browsing the listing.

  • Each listing gets its own knowledge base, so the assistant can answer the things a buyer actually asks on a first call: floor area, what the monthly charges cover, which schools the address is zoned for, what the immediate area is like. It is the same ground a senior agent would cover, drawn from the listing rather than from the model.

  • The AI uses 'Graceful Exit' logic — it politely concludes the call, updates the CRM status to 'Not Interested' with the stated reason, and routes the lead into an appropriate nurture sequence. An agent's time then goes to the people who said yes rather than to the ones who did not.

  • It reads the likely location from the dialling code and normalises every scheduling request against the property's own timezone, so a buyer relocating from somewhere else does not turn up an hour out. Where the dialling code and the caller disagree — a mobile kept after a move — the assistant confirms the time out loud before it books.

  • Through webhooks and REST APIs, which is what most CRMs and listing platforms already expose. We build the connectors a deployment actually needs rather than shipping a library of them, so a new lead source is a small piece of integration work rather than a switch to flip — and once it is wired, conversation data and booking confirmations sync back on their own.

Keep reading

Other builds, taken apart the same way

Want to talk about yours?