Our Work/Case Study/Real Estate SaaS
Jul – Oct 2026 · LiveRealTex CRM.
One multi-company platform for an Indian real-estate sales team's leads, WhatsApp, AI calls and inventory. Before it, a single lead could pass through a Facebook form, a spreadsheet, a personal phone and a broker's memory — and nobody saw the whole story.
(01) — Summary
At a glance.
Developers, brokerages and their field teams
Sold as SaaS with plans and Razorpay billing, so each company runs its own isolated workspace.
Leads leak, follow-up is slow, managers are blind
Leads arrive from many channels, follow-up is manual, and nobody can see what happened to each lead.
One lead record, with automation around it
AI qualification calls, WhatsApp automations, drip campaigns, property matchmaking and field reporting.
(02) — Background
Speed wins the site visit.
Indian real-estate sales runs on speed and WhatsApp. RealTex was built for the teams that run that cycle every day.
A buyer fills a Facebook form for a 2 BHK and hears from three projects within an hour. The one that calls first and follows up best usually gets the visit.
Owner / sales head
Buys ads, sets pipelines, watches conversion and agent performance.
Sales agent
Calls leads, chats on WhatsApp, books site visits, logs follow-ups.
Site staff
Marks attendance, reports site visits with GPS, shares brochures.
Platform operator
Onboards companies, manages plans, billing and support.
(03) — The Problems
Six leaks costing leads and money.
Before RealTex, a typical team ran on Leads Center exports, Excel sheets, agents' personal WhatsApp and handwritten call notes. Inventory lived in yet another sheet.
Lead leakage across channels
Facebook ads, website forms, portals, walk-ins and WhatsApp each lived in their own place. Leads went missing, were duplicated or never assigned.
Slow first response
Agents couldn't call every new lead within minutes — especially at night and on weekends — so the hottest leads went cold.
No consistent follow-up
Nurturing depended on memory. Most leads got one or two touches, then were forgotten.
WhatsApp outside company control
Chats lived on agents' personal phones. When an agent left, the history and relationship left too.
No visibility for managers
Owners couldn't tell which ad, agent or project produced visits and bookings, so ad spend was tuned blind.
Manual matching & field reporting
Agents matched budget and BHK to flats from memory. Visits and attendance were reported on paper or in chat.
(04) — Goals
What success looks like.
One source of truth for every lead
Every channel lands in one lead record, with no duplicate phone numbers.
Respond in minutes, around the clock
New leads get an AI call or WhatsApp message without waiting for an agent.
Follow-up that never stops
Drips and automations nurture leads until they convert or opt out.
Company-owned WhatsApp
Every message, manual or automated, shows in a shared inbox the company controls.
Visibility from ad to booking
Managers see source, stage, calls and visits — and stage events flow back to Meta.
(05) — The Solution
Built around one lead record.
Each problem gets a dedicated part of the product — all reading from and writing to the same lead.
AI voice agents call every new lead
Bolna AI agents call new leads automatically, qualify them and write the outcome back to the lead. Interested buyers are passed live to an available salesperson.
- Three-step campaign setup with lead filters
- DND numbers skipped before a single dial
- Retries, concurrency and a 9 AM – 9 PM window
Warm hand-off, with full context
When the AI finds a hot lead, the agent's phone rings with a screen-pop explaining why — budget, BHK, location, timeline — and the lead record already open.
- AI lead brief summarises every past touch
- Call outcomes map to pipeline stages and follow-ups
- Remarks, requirements, matches and history in one place
Every channel into one lead list
Facebook Lead Ads (webhook plus polling fallback), website forms, a drag-and-drop form builder, webhook triggers and bulk import all feed the same list.
- Duplicate phone numbers blocked at entry
- Form fields map straight to CRM properties
- Round-robin pools assign agents instantly
Inventory, reports and field work — in your pocket
Inventory Radar ranks towers, flats and plots against each lead's budget, BHK and facing. Owners get live dashboards; field staff get attendance, GPS visit reports and a call-reporter app.
- Live unit status: available, on hold, booked, sold
- Conversion funnel from lead to deal
- Capacitor app for Android/iOS + Expo call reporter
WhatsApp control
Official WhatsApp Business API with templates, broadcasts and a health dashboard. Personal numbers connect through managed sessions with daily send caps. A chatbot builder and shared inbox keep every message on record.
Drip & automations
A visual flow builder runs drip campaigns and event-triggered automations — new lead, stage change, overdue follow-up, overdue visit. Every run is logged.
Closing the ad loop
Dashboards, reports and call reporting, plus an AI brief for each lead. The Meta Conversions API sends stage events back to Facebook to improve ad targeting.
(06) — Architecture
Sources in, channels out.
A pnpm monorepo. One API serves the web app, the mobile apps and every integration.
Leads enter from the left and are stored in PostgreSQL. Slow work — imports, broadcasts, drips, dial queues — runs on BullMQ workers backed by Redis. Messages and calls leave through the channels on the right.
- Web app
- React 19, Vite 7, Tailwind 4, TanStack Query, Radix UI — installable as a PWA
- Mobile
- Capacitor 8 wrapper for Android/iOS, plus an Expo Android call-reporter app
- API
- Node.js, Express 5, WebSockets for live lead & chat updates, OpenAPI-generated typed client
- Data
- PostgreSQL 17, Drizzle ORM, company-scoped tenancy
- Background jobs
- Redis 7, BullMQ queues for drip, automations, broadcasts, calling, imports
- AI
- OpenAI (gpt-4o-mini, Whisper), Bolna voice agents
- Deployment
- Docker Compose, Nginx, automated database migrations
(07) — Challenges
The hard parts were integrations.
Not the CRM screens — the messaging and calling pipes behind them.
01WhatsApp number bans
+
Personal numbers that send too much get banned, and the daily counter never reset for new sessions.
A rolling 24-hour send window per session. Every send path — broadcast, automation, drip — checks its budget first.
02Automated messages missing from chats
+
Messages sent by automations and staff alerts were never saved, so agents couldn't see them.
Every automated send now writes to conversation history, so the inbox shows the full thread.
03Turning AI calls into CRM actions
+
A call's result sat in a transcript nobody read.
Outcomes map to pipeline stages and set follow-ups within business hours. Transcript fields fill the lead, and drips advance on the stage change.
04Calling compliance
+
Automated dialling could hit numbers at night or numbers on the DND list.
Calls only between 9 AM and 9 PM, DND/DNC scrubbing, and numbers normalised to E.164 before any dial.
05Data isolation between companies
+
Many companies share one database — a single leak would be serious.
Company-scoped tables and access checks, with automated tests for lead access and Facebook data isolation.
06Meta API changes
+
Facebook webhooks sometimes miss leads, and Graph API v24 moved the WhatsApp messaging-limit field.
Polling backs up the webhook, sync-health alerts fire when leads stop arriving, and the limit is read from its new location.
07Heavy work blocking the app
+
Large imports, broadcasts and calling campaigns slowed every user's requests.
Ten BullMQ queues on Redis run this work in background workers.
(08) — Results
A full platform in three months.
From first commit to production at app.realtex.in.
Product modules
From leads and deals to plot management, attendance and brochures.
Apps
Web (PWA), Android/iOS wrapper, Android call reporter, API server.
Database tables
Company-scoped, with automated isolation tests.
Job queues
Drips, automations, broadcasts, calling and imports off the main thread.
(09) — Lessons
What we'd tell the next team.
Integrations are the product
Most hard bugs lived in WhatsApp, Meta and Bolna handling. Retries, fallbacks and health alerts mattered more than new features.
Every automated action leaves a trace
Agents trust automation only when they can see what it sent and why — in the inbox and timeline they already use.
Rate limits protect the customer
Send caps and calling windows feel restrictive, but they keep numbers from being banned and companies from being penalised.
Close the loop with ad platforms
Sending stage events back to Meta turns the CRM from a record of leads into a way to buy better leads.
Focus the interface
Features that confuse daily users can be hidden from main screens without removing them — the drip-campaign shortcut was recently taken off the leads page this way.
(10) — Conclusion
From phone-and-spreadsheet chaos to one system.
Every lead captured, contacted within minutes, nurtured automatically and tracked from ad click to site visit — without losing control of data, numbers or compliance.