The problem.
A real estate agency with five brokers managed leads from four channels: each broker's personal WhatsApp, a shared mailbox, real estate portal forms (Idealista, Fotocasa, Pisos.com), and calls. Typical volume: +300 new contacts a month.
The chaos was structural. Each broker had their own way to track leads — notebook, personal Excel, memory + WhatsApp history. They'd tried two CRMs and both failed: brokers saw them as extra work and stopped using them within weeks. Result: nobody knew the real pipeline, how many active leads existed, who was working them, or when last contact had been.
What we built.
Multichannel ingestion
n8n monitors the four channels: WhatsApp Business API for the shared number, IMAP for the mailbox, portal webhooks where available + scraping where not, and call recordings transcribed via Whisper. Each lead arrives normalized to a common format.
AI deduplication and enrichment
Same lead writing across channels? The system detects it via fuzzy match on phone, email, and message context. Enriches with the public info available (LinkedIn, company, etc.) and assigns to the right broker by zone or workload.
Auto-fill CRM
The CRM (Pipedrive in this case) gets the complete deal automatically: source channel, message thread, suggested score, contact data. The broker only has to act, not record.
Reminders and reports
Each broker receives a daily summary: leads to call back today, deals about to go cold, performance vs week. The owner sees the master pipeline updated in real time.
The brokers stopped resisting the CRM because they didn't have to fill it out. The system fills it for them. Result: 12 hours a week back, +35% visits scheduled.
Stack used.
n8nPipedriveWhatsApp Business APIWhisperGPT-4oPostgreSQL