From Meta ad to viewing request, in Farsi, in one WhatsApp thread.
NestM8 is our AI lead qualification system for Farsi-speaking real estate buyers in Greater Vancouver. It runs on Meta click-to-WhatsApp ads, qualifies leads through structured Farsi conversations, and hands warm prospects to realtors. We built it, we run it, and three realtors are engaged for handoffs.
Vancouver's Farsi-speaking buyers were underserved.
Greater Vancouver has 60,000+ Farsi speakers and a significant real estate buying population — but most listings, calculators, and realtor sites are English-first. The tools buyers use to evaluate affordability, neighborhoods, and schools assume English fluency, even when the real conversation at the kitchen table happens in Farsi.
Realtors were paying $100+ per lead from shared lead services, getting mostly unqualified contacts, and spending hours manually qualifying Farsi-speaking prospects who wanted to discuss affordability in their native language. Qualification was the bottleneck — not deal flow, not listings, not closing skill.
What we built.
Meta ad → WhatsApp entry
Farsi-language Meta click-to-WhatsApp carousel ads, ~$2/lead, self-selecting on buying intent. No demographic targeting needed due to Meta's housing-ad restrictions — the Farsi copy does the targeting.
Sorena — the AI agent
A LangGraph multi-agent system with persona Sorena (سورنا). Holds structured Farsi conversations across five journey stages: orientation → financial → narrowing → decision → handoff.
Affordability gating
The bot calculates real purchasing power (income, down payment, debt) before showing any listings. Unaffordable homes are never shown. This is the difference between a lead and a qualified lead.
Handoff to realtor
When a lead requests a viewing or reaches handoff-trigger signals, the system routes them to one of three engaged realtors with full conversation context — in English — so the realtor doesn't re-qualify.
What happened when we turned it on.
A Farsi-speaking lead in the Greater Toronto area, planning a move to Vancouver, clicked a Meta ad at 9:50 PM. Entered WhatsApp. Had an 82-message qualification conversation with Sorena in Farsi — shared $75k income, $30k down payment, $600k–$800k budget, and Coquitlam / Port Moody as target areas. The bot expanded the search when inventory was thin, sent 9 matching listings, and the lead picked a $699,999 two-bedroom condo on Como Lake Avenue.
Thirty minutes after the first message, the bot routed the viewing request to Mohammad, one of three Greater Vancouver realtors we're engaged with. Mohammad called back the same day. Confirmed the income the AI captured was correct. Advised pre-approval as the next step. Sent a voice note: "The lead was legit, based in Toronto, thinking of moving to Vancouver, works remotely. Salary was correctly captured."
Total human effort from the realtor before the handoff: zero.
"The lead was legit — based in Toronto, thinking of moving to Vancouver, works remotely. The salary was correctly captured by the AI. I advised pre-approval first."
— Mohammad, Greater Vancouver realtor
"I don't need more AI tools. I need leads. I'd pay $200+ per qualified lead — that's what they're worth."
— Hamid, top 10% realtor by volume
Why this matters if you're not a realtor.
NestM8 is real estate, but the pattern is universal. Every service business with expensive-to-acquire leads and a qualification bottleneck has the same shape: too many unqualified conversations, too few qualified handoffs, too much manual triage.
The funnel we built for Farsi-speaking real estate buyers is the same funnel we build for mortgage brokers, immigration lawyers, or med spas — different language, different qualification logic, same architecture.
- Meta / Google / organic inbound → structured conversation
- AI-driven qualification before human involvement
- Route only handoff-ready leads to the closer
- Measure cost per qualified lead, not cost per click
Under the hood.
Built with LangGraph for multi-agent orchestration, Claude for conversation, WhatsApp Business API for delivery, Meta Ads for entry, and a Python FastAPI backend for state management. The full system runs on ~$50/month in infrastructure and can handle hundreds of concurrent conversations without human intervention.
Run the same play in your business.
If you're a service business with inbound leads and a qualification bottleneck, the audit will quantify what the leak is costing you and show you what the same architecture would look like for your vertical.