For brokerage owners, team leads and channel partners

AI agents for real estate brokerages: every portal lead answered in a minute, every viewing booked.

An agent that replies to portal, ad and WhatsApp leads within a minute, qualifies budget and timing, matches listings, books viewings with the owner and the agent, reminds everyone, and writes the summary into the CRM. Agents keep every conversation that negotiates. Nothing steers, nothing invents a feature.

Thirty minutes, free. You leave knowing which workflow to start with, what a pilot covers and what it costs.

PilotAI Agent Pilot: one workflow, 2–3 weeks, $4,500 (₹3,50,000) fixed. What the pilot includes.

Who this is for: Brokerages and channel partners with three to fifty agents, leads from portals (99acres, MagicBricks, Housing, Zillow, Rightmove), ads and walk-ins, and a CRM that is only as true as whoever typed into it last.

The week this fixes

If this is your Monday, keep reading

The lead that enquired on three listings goes to whoever replied first

Portal leads go to several brokerages at once. A Sunday-evening enquiry sits until Monday; by then the buyer has seen two flats with someone else. Speed to lead decides more deals than the pitch does.

Viewings are a four-way scheduling job

Buyer, agent, owner or tenant, and sometimes a society gate pass. Twenty minutes of calls per viewing, a third of them rescheduled, and no-shows because nobody sent a reminder.

The CRM says "interested" for half the pipeline

Budget, areas, timeline and loan status are in agents’ heads and WhatsApp threads. Managers cannot see response times or who is actually warm, and leads leave with the agents who leave.

One workflow, before and after

From portal lead to confirmed viewing

Trigger: A 99acres enquiry lands at 9:50 pm on Sunday: "Interested in 2BHK, Wakad, 85L budget."

Today

  1. LeadEnquiry sits in the portal inbox. Two other brokerages message within ten minutes.
  2. AgentCalls at 11 am Monday. No answer. WhatsApps a brochure PDF.
  3. AgentBuyer replies at 6 pm asking about a different project. Agent is at a site visit.
  4. AgentWednesday: viewing proposed. Owner unavailable. Rescheduled twice over four calls.
  5. AgentCRM entry: "interested". Manager’s pipeline report is a guess.

With the agent

  1. AgentReplies on WhatsApp within a minute, in Hindi, Marathi or English: thanks, confirms buy or rent, budget band, preferred areas, timeline and whether a loan is pre-approved.
  2. AgentMatches two or three listings from the live inventory that fit, with the facts from the listing record only.
  3. AgentChecks the assigned agent’s calendar and the owner’s availability window, offers three slots, confirms the viewing, sends directions and the gate-pass note.
  4. AgentWrites the summary and a hot/warm/nurture tag into the CRM; reminds buyer, agent and owner the evening before; follows up after the viewing.
  5. AgentOpens Monday with a qualified, scheduled buyer and a CRM that is true. Spends the day on viewings and negotiations.

What stays with a person

  • Negotiation, offers and anything about price beyond the listed figure. The agent never suggests what to offer.
  • Any question about neighbourhood "type", community, schools or safety: answered by a person, within fair-housing and local rules, never by the agent.
  • Listing facts: the agent repeats the listing record and says "I will check" for anything not in it. No invented views, renovations or distances.
What it saves

Speed to lead and the hours behind viewings: a worked example

MeasureTodayWith the agent
First reply to a new leadNext working day; longer on weekendsUnder a minute, including Sunday night
Agent hours on first contact and qualification, per month~40 hours~12 hours
Agent hours coordinating viewings, per month~40 hours~12 hours
Leads with budget, areas, timeline and loan status recordedWhatever got typed inEvery one, in the agent’s own words

Worked example. The assumptions:

  • 400 portal, ad and WhatsApp leads a month across a ten-agent brokerage
  • About 6 minutes of agent time to make first contact and qualify a lead, averaged over the ones who never pick up
  • 120 viewings a month, about 20 minutes of calls to coordinate each
  • The agent completes qualification and scheduling for about 70%; the rest reach a person with context

Hours are the small number. The one that matters is how many of the 400 reach a viewing and an offer, which is why a pilot measures lead-to-viewing rate and no-shows against the previous quarter rather than quoting an uplift here.

Put your own numbers in the automation savings calculator, or get a price range from the AI agent cost estimator.

The agents

Usually one to start, three when it is working

Each agent has a small set of tools it is allowed to call and nothing else. That list is the guardrail, and it is written down here so you can see it.

Lead response agent

On WhatsApp and the website, in the language the buyer writes in. Qualifies, answers from listing records and the brokerage’s approved FAQ (RERA number, possession dates, payment plans as published), matches listings, and hands over the moment a question needs a person.

  • lookup_lead
  • qualify
  • match_listings
  • answer_from_listing
  • update_crm
  • handover_to_agent

Viewing coordinator

Reads the agent’s calendar and the owner’s window, offers slots, books, sends directions and gate-pass instructions, reminds everyone the evening before, reschedules within a limit, and asks for feedback after the visit.

  • check_agent_calendar
  • check_owner_window
  • book_viewing
  • send_reminder
  • reschedule_viewing
  • collect_feedback

Pipeline hygiene agent

Nightly: flags leads with no touch in seven days, drafts the specific re-opener from the last conversation, merges duplicates, and gives the manager response times by agent. Drafts; the agent sends.

  • scan_stale_leads
  • draft_followup
  • merge_duplicates
  • report_response_times

Guardrails

  • The agent never answers questions that could steer a buyer by community, religion, caste or "kind of people". Those route to a person, with the brokerage’s policy as the rule.
  • Every listing fact comes from the listing record; nothing is inferred from photos or filled in from general knowledge.
  • No price guidance, no "they will accept less", no offer drafting. Negotiation is a licensed person’s job.
  • Messaging only to leads with consent on record; DND, WhatsApp template and opt-out rules enforced, and the agent stops the moment someone asks.
  • Every action is logged against the lead, so a manager can audit a week in ten minutes and a dispute has a transcript.

Systems it talks to

  • Portal lead feeds (99acres, MagicBricks, Housing.com; Zillow, Realtor.com, Rightmove outside India) by email parsing or API
  • WhatsApp Business API and the website enquiry form
  • CRM: Sell.Do, LeadSquared, Zoho CRM, HubSpot, Follow Up Boss, kvCORE, or the sheet you have outgrown
  • Google Calendar or Outlook for agents; a shared sheet for owner availability; the listing inventory in your CRM or a sheet
Proof, and its limits

What we have built, and what we have not measured yet

The lead-response agent on the demos page is the same design as this one, built for a software studio instead of a brokerage: it qualifies, scores and drafts the handover. We have not yet published a brokerage result and will not quote one until a brokerage agrees to publish theirs. The closest shipped proof is The Raising Club, a marketplace in production in the US where listings, bookings and availability live on one schema.

FAQ

Questions real estate brokerages ask us

Will the agent say something that breaks fair-housing or anti-discrimination rules?

It is designed not to be able to: questions about neighbourhood "type", community, religion or safety are recognised and handed to a person rather than answered, listing copy it drafts goes through a banned-phrase check and a human review, and it never characterises an area. In the US that is the Fair Housing Act; in India and the UK the rules differ but the design is the same.

Which portals and CRMs does it work with?

Any portal that emails leads or offers an API; we parse the email where there is no API. CRMs with APIs (Sell.Do, LeadSquared, Zoho, HubSpot, Follow Up Boss, kvCORE) are straightforward. A brokerage running on WhatsApp and a sheet can start with the sheet; integration follows once the pilot proves it.

Our agents will not use another tool. Does this need them to?

No. The agent works inside WhatsApp and the CRM you already have, and the agents keep doing what they do: viewings and negotiations. What changes is that the lead arrives qualified and scheduled, with a summary, instead of as a cold name.

Is it legal to message leads automatically in India?

With consent, on the WhatsApp Business API with approved templates, respecting DND and opt-outs, yes. The agent enforces those rules rather than relying on an agent to remember them, and the consent you capture on the portal and the form is what it checks before the first message.

What does it cost?

The AI Agent Pilot covers the lead-response agent on WhatsApp for one team or one project for a fixed fee in two to three weeks, measured on lead-to-viewing rate against your last quarter. Viewing coordination and pipeline hygiene are scoped after. The cost estimator gives you a range for your setup.

Next step

Tell us about your brokerage

How many agents, where the leads come from, which CRM. We reply with what a pilot on your lead flow would cover.

Or start with a number: the AI agent readiness quiz takes three minutes and gives you a score and a 30-day plan.

Or send the details

Name and email are all we need. The rest helps us reply with something specific.