AI Real Estate Lead Enrichment: Agent Research in 30 Seconds
Anas Yaseen
The client's math: about $4,000 in AI fees to have Claude research 2,000+ real estate agents.
The problem
The client cold calls real estate agents around San Antonio, Texas. Each call needs a personal hook: a military past, the areas an agent serves, a recent sale. Finding it meant searching Google, Zillow, Realtor.com and brokerage sites, one agent at a time.
What I built
An AI research system with a simple dashboard:
Upload the Excel list of agents and click Start
Each name and phone number is searched on Google
AI reads the top 5 sites and writes a 30-second call brief: brokerage, background, areas served, recent listings and icebreakers
A link to every brief is added to the Excel sheet
The team edits the AI instructions themselves
The hard part
Google blocked the first test runs with captchas. Rotating proxies would fix that but add a monthly bill. Instead, each browser runs on its own signed-in Google profile and searches like a normal person, and the final test runs showed no captchas. Agents with the same name were the other trap, so the AI checks that the name and phone match before it writes anything.
Results
About 30 seconds per agent, with several running at once
About 1 cent of AI cost per agent, so roughly $20 for 2,000 agents instead of $4,000
Icebreakers, recent listings and contacts in every brief
No paid proxies needed
Tech: Python, Gemini AI, nodriver, LiteLLM, Streamlit
Full case study: https://mranas.dev/work/real-estate-lead-enrichment
Like this project
Posted Jun 2, 2026
A business that cold calls real estate agents around San Antonio, Texas had 2,000+ agents to research. Now the team uploads the Excel list and clicks Start. AI searches Google, reads the top 5 sites and writes a 30-second call brief per agent, for about 1 cent each.
