CASE STUDY · REAL ESTATE

A Voice Agent That Books Property Viewings — in Urdu and English

A real estate agency was losing enquiries to voicemail after hours, and losing Urdu-speaking callers entirely. We built a voice agent that answers in either language, checks the live calendar, and books the viewing before the call ends.

Google Workspace · Automated Booking Confirmation
Real Estate Booking Confirmation Email
🏢 Industry
Real Estate & Property
🎙️ Service
🗣️ Languages
Urdu & English (Native Switch)
⚙️ Platforms
VAPI & Make.com
🧠 AI Models
GPT-4o, Deepgram, ElevenLabs
📅 Calendar
📋 Handles
Inbound, Qualify & Bookings
🟢 Status
Live in Production

Property Enquiries Don't Arrive During Office Hours

Someone deciding to look at a property does it in the evening, on the sofa, after work. They call. The office is closed.

Property is a high-value, low-urgency purchase, which means the buyer is patient about the decision and impatient about the response. If nobody picks up, they don't leave a voicemail. They call the next agency on the list, and that agency books the viewing.

There was a second problem underneath the first. A large share of callers were more comfortable speaking Urdu. They were getting an English-only experience, or no experience at all. That is not a translation problem. It is a lost customer problem.

What We Built

A native conversational architecture built for instant scheduling:

📞

Answers Every Call 24/7

Any hour, any number of concurrent calls at once, with zero hold queue.

🗣️

Natural Language Switching

The agent opens in conversational Urdu-English and follows the caller's lead — spoken naturally, not formal literary textbook Urdu.

📅

Live Calendar Connection

The agent queries live calendar availability during the call and offers slots that genuinely exist.

Automatic Confirmation Flow

The moment a slot is agreed, the calendar event is created and confirmation dispatches instantly without human intervention.

VAPI · Assistant System Prompt & Bilingual Voice Config
VAPI Agent Configuration
ANNOTATION: First message set to conversational Urdu-English greeting ANNOTATION: System prompt configured for native language switching

What Happens on a Call

Seven automated steps from first ring to confirmed calendar event:

01 STEP 1

The Agent Answers

Picks up immediately, greets in conversational Urdu-English, and asks how it can help.

02 STEP 2

Identifies Intent

Works out if they need a viewing, consultation, or general property details.

03 STEP 3

Qualifies Lead

Captures budget range, preferred location, property type, and purchasing timeline.

04 STEP 4

Checks Live Calendar

Queries the live calendar while using natural conversational filler rather than awkward silence.

05 STEP 5

Offers Real Openings

Presents specific available slots conversationally instead of vague callback promises.

06 STEP 6

Repeats Out Loud

Repeats the confirmed date, time, and property details before ending the call.

07 STEP 7

Books & Confirms

Creates the Google Calendar event and triggers automated email confirmation.

ZERO ADMIN

No Staff Needed

Everything happens autonomously without anyone opening a spreadsheet or tab.

Make.com · Automated Booking & Calendar Webhook Architecture
Make.com Booking Scenario
1. Agent sends booking request 2. OpenAI structures appointment details 3. Writes event to live Google Calendar 4. Router handles booked vs not-booked

The Appointment It Booked

Here is what that produces: a viewing consultation, booked by the agent during the call, written into the agency's Google Calendar with caller details, with confirmation email dispatched automatically and a 30-minute reminder set.

Confirmation Email Dispatched
Confirmation Email Evidence
✓ Sent automatically upon call completion
Google Calendar Event Created
Google Calendar Event Evidence
Created by the system, not a person Reminder set automatically

⚡ No staff member touched any part of this booking.

How It's Wired

A resilient production pipeline tuned for low latency and conversational coherence:

Speech In
Deepgram handles transcription, tuned for accented and code-switched speech — a harder problem than single-language transcription, and the part most off-the-shelf setups get wrong.
Understanding
GPT-4o runs the conversation against a system prompt that defines the agent's identity, qualifying questions, booking rules, and language register. The prompt explicitly steers the agent toward everyday spoken Urdu mixed with English rather than formal literary vocabulary.
Voice Out
ElevenLabs Turbo v2.5, chosen for ultra-low latency. On a phone call, response delay is what makes a system feel artificial — more than voice quality does.
Booking Layer
When the agent has what it needs, it fires a webhook into a Make.com scenario. The scenario structures appointment data, parses it into clean JSON, and writes the event into Google Calendar.
Two-Branch Router
Booked and not-booked branches are handled separately, each returning a response the agent can speak naturally. This keeps the conversation coherent when a slot is unavailable.

What This Means for Your Business

How this architecture translates to service businesses across the US:

🗣️

Answered in Their Own Language

If a meaningful share of your callers speak Spanish, this is the exact same build. The agent switches seamlessly to match the caller rather than forcing them into English.

📅

Books Instead of Taking Messages

The difference between a generic answering service and this is a live calendar connection. One creates a callback chore; the other creates a booked customer.

✉️

Confirmations Happen Without Staff

The calendar entry, the confirmation email, and the 30-minute reminder all fire automatically. Nobody has to remember.

🌙

After-Hours Stops Being a Gap

Evenings, weekends, and holidays are handled with the same speed and accuracy as a Tuesday morning.

Where This Works, and Where It Doesn't

We are transparent about where voice automation excels and where it requires a hybrid approach:

STRONG FIT

This Works Well For

  • Businesses where booking an appointment is the main job of an inbound call
  • Standardized appointment types with consistent information requirements
  • Clear availability rules the system can follow
  • Any business losing high-value evening and weekend calls
  • Any business whose customers span two languages
WEAKER FIT

This Is a Weaker Fit For

  • Bookings requiring extensive bespoke intake or specialist pre-qualification
  • Businesses with no consistent booking process to encode
  • Situations where the first call is a complex price negotiation

💡 Hybrid Architecture: For the middle ground, the agent handles the call and captures details, and a human team member picks up complex custom scoping.

BUSINESS ON AUTOPILOT

Want One Built for Your Business?

On the strategy call we'll go through how calls currently reach you, what your booking rules actually are, and what an agent would handle. If it's not a fit, we'll tell you.

EXAMPLE
Inbound Property Call Log
18:47
(555) 219-8840 Answered in 6s · viewing booked
BOOKED
19:02
J. Reyes Answered in Urdu · qualified & booked
BOOKED
21:38
M. Okafor After-hours call · booked Thu 09:00
BOOKED
22:15
T. Vance 24/7 AI Voice · consultation scheduled
BOOKED
14:10
D. Chen Viewing booked · Google Cal synced
BOOKED

Ready to collaborate with us?

Tell us about your business and we'll design your tailored AI blueprint.