Every morning I wake up to a fresh batch of qualified business prospects. Doctors, dentists, law firms, CPAs — all in South Florida, all with verified phone numbers, all ready for outreach.
I didn't hire a VA. I didn't buy a lead list. I built an AI agent that runs 5 times a day across 4 different campaigns, and it's been doing this autonomously for over two weeks.
Here's how it works.
The Architecture
The system is simple in concept: search for local businesses, check if we already have them, and add qualified ones to our outreach queue. But the devil is in the details.
Each campaign runs on a cron schedule. The agent searches Google Maps for specific business types in specific geographies — "dentists Boca Raton," "law firms Fort Lauderdale," "dermatologists Palm Beach Gardens." It extracts business names, phone numbers, categories, and locations.
Then comes the part that actually matters: deduplication.
Why Dedup Is the Real Feature
About 40% of search results return businesses we already have in the queue. Without dedup, you're wasting outreach on people you already contacted, or worse, contacting them twice and looking unprofessional.
The agent checks each prospect against the existing queue by phone number and business name. It's not fancy — it's just thorough. Every run logs exactly how many were added vs. skipped, so I can see the hit rate declining as the queue saturates a geography.
When I started seeing 10+ duplicates per run in Miami and Fort Lauderdale, I knew it was time to expand to new areas: Coral Springs, Jupiter, Deerfield Beach, Plantation. The saturation signal is built into the system.
Four Campaigns, One Queue
I run four campaigns with different angles:
- Receptionist Campaign — targeting practices that would benefit from AI phone answering
- Reviews Campaign — businesses where online reputation management matters
- AI Campaign — firms that might want general AI automation
- Boring Campaign — the catch-all for financial advisors, wealth management, CPAs
Each campaign searches for slightly different business types and pitches, but they all feed into the same prospect queue with the same dedup logic. This means a dentist found by the Reviews campaign won't get re-added by the Receptionist campaign.
The Numbers
In two weeks, the system has built a queue of 250+ qualified prospects. That's roughly 18 new businesses added per day, with another 8-10 filtered out as duplicates.
The campaigns run at 10 AM, 1 PM, 3 PM, 5 PM, and 8 PM UTC. Each run takes about 30 seconds. Total compute time per day: under 3 minutes.
Compare that to manual prospecting. A human researcher might find 20-30 businesses per hour, but they'd burn out checking for duplicates and they wouldn't maintain consistency across weeks.
What I'd Do Differently
If I were starting over, I'd build the saturation detection earlier. I wasted a few days running searches in Miami that returned mostly duplicates before expanding to new geographies.
I'd also add category diversification sooner. Starting with just doctors, lawyers, and dentists was fine, but adding orthodontists, chiropractors, dermatologists, and immigration attorneys opened up much richer territory.
The Takeaway
The unsexy truth about AI automation is that it's not about the AI doing something magical. It's about the AI doing something boring with perfect consistency.
Searching Google Maps isn't hard. Checking for duplicates isn't hard. Logging results isn't hard. But doing all of that 5 times a day, every day, for weeks, without getting tired or sloppy? That's where the value is.
The best AI automations aren't replacing creative work. They're replacing the repetitive work that humans do badly because we get bored.
My agent doesn't get bored. It just keeps finding prospects.
Top comments (0)