Every founder knows the grind: find leads, qualify them, add them to a list, repeat. I got tired of it, so I built a system that does it automatically — every few hours, around the clock.
The Problem
Manual prospecting is soul-crushing. You open Google Maps, search for dentists in Boca Raton, click through websites, copy phone numbers into a spreadsheet, and repeat until your eyes glaze over. Even with a VA, you are paying someone to do robotic work.
I wanted something different: a pipeline that finds, qualifies, and queues prospects while I sleep.
The Architecture
The system runs on cron jobs — four campaigns firing at staggered intervals throughout the day:
- Receptionist campaign (10 AM, 8 PM) — finds medical offices, dental practices, and law firms that would benefit from AI receptionists
- Reviews campaign (1 PM) — targets businesses with poor or few online reviews
- AI campaign (3 PM) — broader search for businesses ripe for AI automation
- Boring campaign (5 PM) — wealth managers, CPAs, dermatologists — the unsexy but profitable verticals
Each job runs independently. It searches Google Maps for businesses in South Florida, scrapes their websites with Firecrawl, evaluates fit based on criteria like business type and online presence, then pushes qualified leads into a Redis queue.
How It Actually Works
The AI agent (yes, it is an AI doing the prospecting) follows a simple loop:
- Search — Pick a category and location. Query Google Maps API for businesses matching the campaign criteria.
- Scrape — Hit each website with Firecrawl to pull contact info, services offered, and technology signals.
- Qualify — Does this business answer phones? Do they have an existing chatbot? Are they in our target revenue range? The agent makes judgment calls here.
- Queue — Qualified prospects go into Redis with their name, phone, website, and campaign tag. Duplicates are filtered automatically.
- Log — Everything gets written to daily memory files so I can review what was found.
The results speak for themselves. In the last three days alone, the system queued 67 prospects across all four campaigns. That is 67 businesses I did not have to manually research.
The Dialer Connection
Queued prospects do not just sit there. A separate outreach dialer pulls from the Redis queue and makes AI-powered phone calls using Retell. The calls are natural-sounding conversations that qualify interest and book follow-ups.
Last week, the dialer burned through 37 calls from the queue in a single day. Some connected, some went to voicemail, some were not interested — but the pipeline kept refilling faster than the dialer could empty it.
What I Learned
Stagger your jobs. Running all four campaigns at once hammers the APIs and produces worse results. Spreading them across the day gives better data and stays under rate limits.
Redis queues are perfect for this. Simple, fast, and you get deduplication almost for free with sets. No need for a fancy database when you are just staging leads for outreach.
Campaign segmentation matters. The receptionist campaign finds different prospects than the reviews campaign, even in the same zip code. Different angles surface different opportunities.
Let the AI make judgment calls. Early versions had rigid filters that missed good prospects. Letting the agent evaluate fit contextually — like noticing a dental practice still uses an answering service — catches opportunities that rules-based systems miss.
The Numbers
- 4 campaigns running autonomously
- ~20 prospects per day added to queue
- 67 prospects in the last 3 days
- Zero manual research required
- Cost: Pennies per prospect (API calls + compute)
Try It Yourself
You do not need a complex setup. A cron job, a search API, a scraping tool, and a queue. The AI agent is the glue — it handles the fuzzy logic that makes prospecting actually work. Start with one campaign, one vertical, one city. Scale from there.
The best sales pipeline is the one that runs while you are not looking.
Top comments (0)