Home/Case studies/Blitz Ploiești
How I automate my own work as a real-estate agent at Blitz Ploiești.
Not a project for a client. Not a fictional case study for marketing. I am the client. I use these workflows every day, every week — and I'll update this study as I optimize and measure.
Who I am and what problem I'm solving.
I work as a real-estate agent at Blitz Ploiești with a 2.5% + VAT commission. The Prahova market has roughly X active agents competing for a relatively limited property pool. Differentiation comes down to reaction speed and lead-qualification quality.
Manually, an agent can make 30-50 real prospecting calls per day. That leaves limited time for meetings and viewings — which are the activities that actually generate commissions.
The business question: how do you call 3-5x more property owners without wasting time on the unqualified ones?
My answer: N8N + AI workflows that take over the heavy lifting (prospecting, initial qualification, follow-up) and let me focus on what matters (viewings, negotiation, closing).
Automated owner prospecting
Listings posted directly by owners (not by agencies) are pure gold in real estate. The first agent to call has a 70% chance of getting an exclusive. Manually? It's impossible to monitor 4-5 portals 24/7.
How it works
- Firecrawl scans OLX, Storia, Imobiliare.ro every 15 minutes for new Prahova listings.
- OpenAI classifies: real owner vs. agent in disguise (many agents post "from owner" to get more views).
- For every valid listing: extracts the data (location, surface, price), compares with recent transactions in the zone, calculates a potential score.
- Listings scoring >7/10 land on my priority call list in Notion, sorted descending.
- For each, a personalized WhatsApp message draft (with details from the listing) is ready to send with one click.
Results (real data, updated monthly)
- ~171 outreach attempts/month vs. ~50 manually possible
- Under 60 minutes from posting to first contact (vs. 8-24h manually)
- Daily time invested: ~30 min for list review + targeted calls (vs. 4-5h manually)
- Workflow cost: ~€8/month APIs + VPS
WhatsApp chatbot for buyer qualification
Buyer inquiries come in large volumes from property portals. 70% are "just looking", 20% are serious but without realistic budget, 10% are qualified buyers. The question: how do you filter fast?
How it works
- All WhatsApp inquiries to my agent number are handled by an AI agent.
- The agent replies in seconds with a warm message, asks 6 BANT questions distributed conversationally (not interrogation-style): approximate budget, timing, mortgage pre-approval, whether they've sold their current property, exact target zone, whether they've already viewed others in Prahova.
- OpenAI calculates a BANT score (0-10) and writes a summary into a Notion CRM.
- Leads scoring >7 automatically get a Cal.com link for booking a viewing in the next free slots.
- Leads scoring 4-6 enter a nurture sequence (weekly email + WhatsApp with relevant listings).
- Leads scoring <4 get a polite reply and exit the pipeline.
Results
- Reply in under 30 seconds 24/7 (vs. 2-4h manually during business hours)
- ~10h/week saved from conversations that were going nowhere
- Higher viewing-conversion rate — pre-qualified leads have a higher materialized-viewing rate
Automated expired-exclusive follow-up
When an exclusive expires, the owner often re-posts the listing on their own or with another agent. Matching listings with history is gold — but manually it's impossible.
How it works
- Own database with all expired Prahova exclusives (historic scraping).
- Daily workflow comparing new listings against history: similarity score on text + perceptual hash on images.
- Match with >0.85 confidence = listing from the same owner, expired from someone else.
- Reach-out message at 3-5 days after re-posting (not too early — owner still has hope; not too late — another agent catches them).
- Auto-personalization: "I noticed your listing is back online. I work with active buyers in zone X. Can I help?"
What's not working perfectly yet.
This study isn't marketing. Here's what I still need to improve:
- The chatbot still gives generic answers when buyers ask unexpected technical questions. I'm working on fine-tuning with more context from my closed-transaction history.
- The scraper loses listings when portals change their HTML structure. Average: 1-2 days of "partial downtime" per month until I adjust the selectors.
- BANT qualification doesn't catch Romanian cultural nuances (Romanian buyers often say "no rush" when they actually want to view tomorrow). I iterate the prompt monthly.
- The study doesn't yet have hard numbers on closed transactions directly attributable to the workflows (vs. traditional channels) — attribution in real estate is complicated. This point is under construction and I'll publish when the data is sufficient.
Want these 3 workflows for yourself? Bundle €299, or €1,499 with setup.
"Real Estate Complete" pack — all three workflows described above, ready to import into your N8N. Or buy them with turnkey setup and I'll build everything on your server.