EstateFlow AI

The Problem

Inbound leads arrived on WhatsApp unqualified and unenriched. Agents were manually reading each thread to work out budget, timeline, and property interest before following up, and in real estate, speed-to-lead, how fast a lead gets a first response, is one of the strongest predictors of whether it converts at all.

That manual triage step was where leads went cold. What Velyx needed was an agent that could hold the qualifying conversation itself, on the channel leads were already using, and only escalate once there was something for a human to close.

Constraints

• Had to run natively on WhatsApp Business API, not redirect leads to a separate web chat or form.

• Qualification had to feel like a conversation, not a rigid form disguised as one, or leads drop off mid-thread.

• Site-visit booking had to check real calendar availability in the moment, not just collect a preferred time to confirm later.

• Scoring and qualifying questions had to be editable by the Velyx team directly, since their criteria shift with the market.

• The system had to hold up under US business-hours traffic while the build team worked from Pune, across the time-zone handoff.

How Nexolve Is Building It

EstateFlow AI is an agentic system, not a chatbot script: an OpenAI-powered agent reads intent, qualifies, books, and hands off to the CRM, all inside the WhatsApp thread the lead already opened.

1. 1

   WhatsApp-native intake

   Every inbound WhatsApp message routes through the WhatsApp Business API into an OpenAI-powered agent that classifies buyer intent, buy, rent, or browsing, before a human ever sees the thread.

2. 2

   Conversational qualification

   The agent asks the qualifying questions a human agent would, budget band, preferred locality, timeline, property type, writing structured answers to Supabase as the conversation unfolds.

3. 3

   In-chat site visit booking

   Once a lead is qualified, the agent checks live calendar availability and books a site visit directly inside WhatsApp, no phone tag, no separate scheduling link.

4. 4

   Real-time CRM sync

   Every conversation, qualification score, and booked visit syncs to the team's CRM as it happens, so a human agent picks up with full context instead of a cold handoff.

5. 5

   Self-serve tuning

   An admin dashboard gives the Velyx Labs team visibility into conversation volume, qualification rates, and booked visits, and lets them retune the agent's qualifying questions without opening an engineering ticket.

Tech Stack

Where It Stands

• EstateFlow AI is in active development — the agent, booking flow, and CRM sync described above are being built now, not yet running in production.

• Zero duplicate outreach is a design constraint from day one: a lead can only be claimed by one agent, enforced at the data layer.

• The Velyx team will retune scoring and qualifying criteria themselves as the market shifts, without opening an engineering ticket.

• Built as a direct engineering collaboration between Nexolve and the Velyx Labs team, not a handoff.