Case studies / Plotane
AI Sales AgentOur product · Real estate6 days, Sep 2026

Plotane

A WhatsApp and phone agent for estate agencies that finds the right property, shares photos and a map pin, and books the site visit.

Built for
Estate agencies in India and the US handling buyer and renter enquiries
Platform
WhatsApp, phone and WhatsApp calls, web CRM
Status
Live MVP
The Plotane India landing page, with a WhatsApp chat where the Maya assistant finds a Kochi home and books a site visit.
6 daysfrom first commit to live MVP
3languages, including voice notes
24h · 4h · 1hautomatic visit reminders
0.6sreply time for photo requests
Overview

The short version.

Plotane is a real-estate CRM with an AI agent called Maya built in. Maya lives on the agency's WhatsApp number and phone line. She answers buyers and renters in English, Malayalam or Hindi, finds listings that match what they're looking for, sends photo cards and location pins, and books a site visit with a human agent who is actually free.

It began as a live conference demo for Kerala agencies and was rebuilt into a product with separate India and US editions within the same week. The US edition runs on a real Arizona inventory of 50 listings across six cities.

The challenge

What was broken.

Property enquiries arrive on WhatsApp at 11pm, in Malayalam, as a voice note. By the time an agent replies the next morning, the buyer has messaged three other agencies. Agencies were losing warm leads to response time alone, and the enquiries they did answer took five back-and-forth messages to find a matching listing and a free viewing slot.

  • Late-night and weekend enquiries go unanswered until the next working day.
  • Matching a buyer to listings and a free agent's calendar takes several rounds of messages.
  • Site-visit no-shows waste agents' afternoons.
How it came together

From idea to production.

  1. 23 Sep

    Conference build

    CRM, property inventory and the WhatsApp agent working end to end for a live demo.

  2. 24–25 Sep

    Becomes Plotane

    Rebranded, with India and US landing pages chosen by the visitor's country, pricing for both markets and a real Arizona inventory.

  3. 26 Sep

    Voice calls

    Maya started answering phone and WhatsApp calls live, with transcripts in the CRM.

  4. 28 Sep

    Open to every caller

    Voice opened to all numbers after rate limits and call caps were in place.

The solution

What we built.

Maya handles the first conversation from hello to booked visit. She understands what the buyer wants, searches live inventory, sends rich WhatsApp replies with photos, prices and map pins, and proposes a visit only in slots where a real agent is free. The buyer taps Confirm, the visit appears on the team's board, and reminders go out automatically. Any agent can take over a conversation with one click.

How it works
  1. Step 1Buyer messages on WhatsApp
  2. Step 2Maya matches live listings
  3. Step 3Photos, price and map pin sent
  4. Step 4Visit proposed in a free slot
  5. Step 5Buyer confirms, agent notified
Key features

What we delivered.

Speaks the buyer's language

English, Malayalam and Hindi, by text or voice note. Voice notes are transcribed and answered in the same language.

WhatsApp that looks like an app

Image cards, lists, quick replies, and Confirm / Not now buttons, so buyers tap instead of typing.

Only real availability

Agent rosters with working hours and time off. Visits book in 15-minute slots up to 90 days ahead, and double bookings are impossible.

Reminders that cut no-shows

Automatic reminders 24 hours, 4 hours and 1 hour before each visit, with the location pin attached.

A lead board for the team

Kanban and table views, inquiry history, lead ownership and instant human takeover.

Voice calls

Maya answers phone and WhatsApp calls live, and every call's transcript is saved to the lead.

Follow-ups with consent

Maya follows up only with permission and stops the moment a buyer replies STOP.

Agent studio

Test Maya's behaviour in a sandbox that runs the same engine without messaging anyone.

Inside the product

A closer look.

Where the AI does the work

Intelligent, and trustworthy.

Tools, not guesses

Maya can only act through tools that are locked to the current customer, so she can never see or book for someone else.

Propose, then confirm

Maya proposes a visit; the buyer confirms. Unconfirmed proposals expire after 15 minutes.

A fast lane for simple asks

A small, quick model handles "send photos" and "where is it?" before the main agent runs, cutting those replies from about 3 seconds to under one.

Long memory, short context

Maya keeps a rolling summary of each buyer plus the latest messages, and can search older history when needed.

Built with

The stack.

Platform
Cloudflare Workers, D1 database, R2 media storage, Queues
AI
Vercel AI SDK agent on OpenAI, Gemini Live for voice
Channels
Meta WhatsApp Cloud API, Twilio voice and WhatsApp calling
Web
React 19 CRM with Google sign-in and team roles
Results

The numbers.

Build proof

6 days
idea to live MVP
Verified
193
automated tests passing
Verified
50
live Arizona listings, 150 photos
Verified
−77%
reply time on photo requests
Verified

Expected impact for users

<30s
first reply, any hour
Projected
3×
more enquiries reaching a site visit
Projected
40%
fewer no-shows with reminders
Projected
0
enquiries left unanswered overnight
Projected

Verified from project repositories, test runs and release records. Projected modelled estimates of user impact, not measured client results.

Your product next

Got an idea? Let's ship it.

Tell us the workflow that's costing you customers or hours. We'll come back with a prototype plan, a timeline in weeks, and a fixed quote.