Ava, Lead Qualification Agent.

A real estate team had stopped spending on paid lead generation because more than 10,000 site registrations had cluttered the CRM. Agents spent hours cold-dialing and still missed prime buyers. The diagnosis mattered more than the fix: lead qualification bandwidth was the bottleneck, not lead supply. Buying more leads would have made it worse.

Automate qualification before enrichment or nurture. Every instinct on that team was to add more top-of-funnel, but a database that already held 10,000 unqualified records did not need more records. It needed the existing ones sorted. I also gave the agent a persona and a conversation rather than a form, because the qualifying data they needed (timeline, budget, motivation) is data people volunteer in a conversation and abandon in a form.

• Mapped the qualifying data that actually predicts a buyer: location, timeline, budget, down payment, ownership status, motivation.
• Chatbot embedded on the Buy and Sell CTAs, running GPT-4o via custom API with Google Maps and an MLS scraper.
• Collects answers to variables, scrapes live listings matching the criteria, and displays them as cards in-chat with map lookups and showing scheduling.
• Posts the full JSON payload (answers plus listing clicks) to the CRM lead record and emails the assigned agent instantly.
• The ownership question opens a seller flow, surfacing listing opportunities from buyer traffic.

80%

less manual prospecting, about 4 hours to under 1 per day

2×

increase in hot leads within 30 days

0

additional headcount to reactivate a stale database