AI Agents for Real Estate in India | LandlordAgent & Realzent Team

FILE 01 / WHERE REAL ESTATE OPS BLEED

Leads leak. Follow-ups drop off. Rent chasing eats the founder's calendar. Portals go stale.

You spend on MagicBricks and Meta ads. Enquiries come in at 11 pm. The first response goes out at 10:30 am the next day, if at all. By then the buyer has spoken to three other brokers. Site visits get booked over phone, then cancelled without notice. Rent reminders go out on a WhatsApp broadcast and half the tenants ignore them until the 15th. Listings on 99acres show a wrong price because the sales exec who updated the price file went on leave. A promising deal sits at "site visit done" for 12 days because the broker forgot to follow up.

Agents fix this at the workflow layer, not the messaging layer. A lead qualification agent responds inside 90 seconds, tags the buyer's budget and preferred locality, and books a site visit before your team has coffee. A rent collection agent sends the reminder, checks Razorpay, escalates the defaulter, and updates the ledger. You watch outcomes, not tickets.

The Indian real estate ground reality is broker-heavy, WhatsApp-first, and still cash-tolerant in tenancies. A generic SaaS built for a US brokerage does not survive here. Agents built for this market handle mixed cash and UPI rent, respect the informal follow-up patterns brokers actually use, and know that a buyer saying "sunday check kar leta hoon" needs a Saturday nudge, not a Monday CRM report.

PROOF OF WORK

We run these agents inside Realzent and LandlordAgent every day

Realzent is Decipher's property CRM. LandlordAgent handles tenancy and rent management for NCR portfolios. When we deploy an agent for your brokerage or developer sales team, we're shipping something we already run on our own book. Rent reminders, lease renewals, listing sync, lead qualification, all live in production against real tenants and real buyers.

See LandlordAgent product

New to agents? How they plan, call tools, and stay inside guardrails is covered on the AI Agents & Automation service page

FILE 02 / OUR APPROACH

Agents wired into the actual real estate stack, not a demo sandbox

Every agent we ship has three parts: a defined task with a success metric, tools it can call (WhatsApp API, portal feeds, Razorpay, your CRM, Google Calendar), and an escalation path when it hits a case it should not handle alone. Nothing runs open-loop. Every agent action is logged, reviewable, and reversible.

We pull inventory from MagicBricks, 99acres, and Housing.com feeds. We send WhatsApp through the official Business API. Payments reconcile against Razorpay. Site visits sync to Google Calendar. Rent ledgers sit inside LandlordAgent or your existing accounting stack. No screen scraping, no fragile browser automation.

Human-in-the-loop is the default in week one. Every agent action is proposed for a manager to approve. As accuracy stabilises over the first month, you unlock actions category by category. Rent reminders auto-send by week 3. Site visit confirmations by week 4. Lease renewal terms stay owner-approved for as long as you want. You control the trust curve.

DEPLOYMENT MAP A typical multi-agent deployment

✓ Lead intake agent receives portal and Meta ad enquiries, enriches, scores intent.
✓ Property match agent pulls 3-5 matching units, sends WhatsApp shortlist.
✓ Scheduling agent books the site visit across broker, buyer, property in-charge.
✓ Follow-up agent nudges the deal if it stalls beyond 72 hours.
✓ Rent collection agent handles reminders, receipts, escalations.
✓ Listing agent keeps portal inventory fresh and accurate.

FILE 03 / AGENTS WE SHIP REGULARLY

Eight agent patterns for real estate operations

AGT-01

Lead Qualification Agent

Enriches every inbound enquiry from MagicBricks, 99acres, Housing, Meta ads. Scores buyer intent, tags budget bracket, preferred locality, timeline, self-use vs investment. Hot leads get routed to a broker with a briefing note. Cold leads enter nurture.

40-60% FASTER FIRST RESPONSE

AGT-02

Property Matching Agent

Reads the buyer brief. Runs it against your live inventory (owned stock plus feeds from partner brokers). Sends a 3-5 unit WhatsApp shortlist with photos, floor plans, price, RERA number, and a booking link. Learns from what the buyer opens.

3-5X SHORTLIST ENGAGEMENT

AGT-03

Site-Visit Scheduling Agent

Coordinates a 3-way slot across broker, buyer, and property in-charge. Checks Google Calendar, sends confirmations, runs 24-hour and 2-hour reminders. Handles reschedules conversationally. Escalates no-shows to the sales manager same day.

30% VISIT CONVERSION LIFT

AGT-04

Rent Collection Agent

Sends rent reminder templates 5 days and 1 day before due date. Checks Razorpay for payment. Marks the ledger. Escalates on day 5 of default with a firmer message. Sends the owner a weekly rent-status digest. LandlordAgent runs this today.

CUTS OWNER ADMIN BY 70%

AGT-05

Lease Renewal Agent

Watches lease expiry dates 90 days out. Drafts renewal terms inside owner-approved bands (typical 5-10% escalation, notice period, deposit adjustment). Opens the conversation with the tenant. Routes final terms to the owner for a one-tap approval.

ZERO MISSED RENEWALS

AGT-06

Maintenance Ticket Agent

Tenant reports a leaking tap on WhatsApp. Agent classifies, checks the vendor roster for plumbers in that locality, dispatches, tracks resolution against SLA. Escalates if the vendor misses the 48-hour window. Closes the loop with tenant confirmation.

VENDOR SLA COMPLIANCE VISIBLE

AGT-07

Listing Content Agent

Drafts unit descriptions from raw inputs (BHK, carpet, amenities, floor, facing, RERA). Uploads across MagicBricks, 99acres, Housing.com through their APIs. Refreshes pricing weekly. Flags listings that go 30 days without an enquiry so you can reprice.

HOURS PER LISTING TO MINUTES

AGT-08

Broker Follow-Up Agent

Watches deals in your CRM. Nudges the broker when a deal sits at "site visit done" for more than 72 hours. Drafts a hand-off note when a broker goes on leave. Flags deals that have gone silent past 14 days for manager review.

RECOVERS 15-20% STALLED DEALS

Want to see one of these agents running on your data?

15-min call. Tell us your inventory count, current CRM, and the one workflow that eats the most manager time. We come back with a pilot scope and fixed price inside 48 hours.

Chat with Deci

FILE 04 / AGENT STACK

What we build on

STACK-01 Orchestration

LangGraph for stateful workflows. CrewAI for multi-agent coordination. Custom Python runners for high-throughput agents.

STACK-02 Models

Claude Sonnet for reasoning. GPT-4o for extraction and drafting. Smaller open models for high-volume classification. Every agent has a fallback.

STACK-03 Messaging & Calls

WhatsApp Business API (Interakt, Gupshup, direct Meta). Twilio for voice callbacks. SMS OTP fallback through BookMySMS.

STACK-04 Real Estate Integrations

MagicBricks API, 99acres, Housing.com, Google Calendar, Razorpay, Realzent CRM, LandlordAgent ledger, legacy broker CRMs.

FILE 05 / OUR OWN BOOK

What we run inside LandlordAgent and Realzent

PRD-01 LandlordAgent

LandlordAgent. Tenancy & rent management

Rent collection agent handles reminders 5 and 1 day pre-due, checks Razorpay in real time, marks the ledger, escalates on day 5 of default. Lease renewal agent watches expiry dates 90 days out and drafts terms inside owner-approved bands. Maintenance ticket agent routes complaints to vetted vendors in the tenant's locality and tracks SLA to close.

Result: owner admin down 70%. Rent default cases resolved 2x faster.

See LandlordAgent

PRD-02 Realzent

Realzent. Property CRM for brokerages

Lead qualification agent responds to portal enquiries inside 90 seconds. Property matching agent sends WhatsApp shortlists that get 3-5x the click-through of a portal email. Site-visit scheduling agent coordinates broker, buyer, and property in-charge without a phone call. Broker follow-up agent nudges stalled deals and drafts hand-off notes.

Result: 40-60% cut in first response time. 30% site-visit conversion lift.

See Realzent

FILE 06 / GUARDRAILS & ESCALATION

Every agent has a leash. Nothing runs open-loop.

Each agent has a defined action scope. The rent collection agent can send reminders and mark payments, but cannot issue eviction notices. The lease renewal agent can draft terms, but cannot execute the final agreement. The listing agent can update descriptions and photos, but cannot change price beyond a threshold without owner approval.

Escalation is built in. When an agent hits an edge case (a tenant disputing a rent charge, a buyer asking about legal issues on a property, a broker calendar conflict it cannot resolve), the case surfaces to a human dashboard with full context. You never wake up to a mess made by an agent going rogue at 2 am.

AUDIT LOG What we log for every agent action

✓ Input the agent received (message, event, trigger)
✓ Reasoning trace, including model used and cost
✓ Tools called, with parameters and responses
✓ Final action taken, or escalation reason if none
✓ Human reviewer if the action was reviewed
✓ Outcome tag once the workflow closes

DPDP audit trail out of the box. Roll back any action inside the review window.

FILE 07 / PROCESS & PRICING

Pilot to production in 6-10 weeks. Fixed pricing per phase.

WEEK 1

Discovery

Map your inbound flows, CRM, portal contracts, rent book if applicable. Written feasibility with recommended first agent.

FREE

WEEKS 2-3

Agent Pilot

One agent shipped end-to-end. Lead qualification or rent collection or listing sync. Runs on your live data with human review before every action.

₹1.5-3 L

WEEKS 4-10

Multi-Agent Build

3-6 agents wired together. CRM, WhatsApp API, portal feeds, Razorpay, calendar sync. Escalation paths, audit logs, dashboards.

₹5-15 L

ONGOING

Managed Ops

Model API costs, weekly evals, prompt tuning, escalation review, WhatsApp template management, quality monitoring.

FROM ₹40K/MO

REF-A Parent & Sibling Capability

REF-B Agents by Industry

REF-C Related Products & Services

FILE 08 / FIELD MANUAL — FAQ

Questions we get on every founder call

What is an AI agent versus a chatbot for real estate?

A chatbot replies to messages. An agent takes actions. A rent collection agent sends the reminder, checks Razorpay for payment, marks the ledger, and escalates to the owner if 5 days pass without payment. A lead qualification agent scores the lead, tags budget and location, creates the CRM record, and books a site-visit slot. Chatbots answer. Agents do the work.

Is agentic AI DPDP compliant for tenant and buyer data?

Yes if built correctly. We keep buyer and tenant PII in Indian regions, log every agent action for audit, use purpose-limited data access per agent, and support consent withdrawal. Payment data stays with Razorpay or your existing rails. We never train external models on your customer data.

Will these agents work with our existing broker CRM?

Yes. We integrate with Sell.Do, LeadSquared, Salesforce, Zoho, HubSpot, and older custom CRMs common in NCR and Mumbai brokerages. If the CRM lacks an API, we run agents against Google Sheets or a nightly export. Same pattern we use in Realzent for our own operations.

How do agents handle WhatsApp Business API rules?

All outbound uses approved template messages inside the 24-hour session window rules. Rent reminders, visit confirmations, and listing shortlists run as pre-approved templates. Two-way conversations happen during active sessions. We manage template submissions, quality ratings, and provider selection across Interakt, Gupshup, and direct Meta.

Do you handle RERA and legal document workflows?

Yes. Agents can reference RERA registration numbers when drafting listing content, cite approved plans in buyer conversations, and flag missing compliance docs before publishing. For rental agreements and lease renewals, agents draft within owner-approved templates and route final review to a human before execution. Nothing legal goes out without a human tap.

What is the ROI horizon on an AI agent deployment?

Lead qualification and rent collection agents pay back inside 3-4 months for portfolios above 100 leads a month or 200 rental units. Listing sync agents pay back in 6-8 weeks by removing manual portal upload work. Site-visit scheduling and broker follow-up agents show up as pipeline lift within 60 days, harder to attribute cleanly but visible in your conversion numbers.

Which models do you use and why?

We route by task. Claude Sonnet for reasoning-heavy work like lead scoring and negotiation drafts. GPT-4o for structured extraction and portal listing generation. Smaller open models for high-volume classification jobs like intent tagging. Every agent has a fallback model. We monitor cost per successful action, not tokens.

How much does a full AI agent build cost?

Single agent pilot (2-3 weeks): 1.5 to 3 lakh. Multi-agent build with 3-6 agents wired to your CRM, WhatsApp, portals, and payment rails (6-10 weeks): 5 to 15 lakh. Managed operations from 40,000 per month including model API costs, weekly evals, escalation review, and prompt tuning. Discovery is free.

Stop paying managers to chase rent and follow up on stalled deals

Book a 15-min call. Tell us your inventory count, your CRM, and the one workflow that eats the most manager time. We come back with a pilot scope and fixed price inside 48 hours.

Chat with Deci Send Requirements →