I came across an interesting acquisition sourcing project this week.
The requirement was simple on paper:
Monitor broker emails, broker websites, and marketplaces for acquisition opportunities. Use AI to qualify deals. Push everything into a CRM automatically.
The catch?
The client wanted it built in n8n.
I normally reach for n8n. That's where I'm most comfortable. But automation is automation. The tool matters less than the system design.
The real problem wasn't moving data between tools.
The problem was signal vs noise.
A broker email is easy.
But once you start monitoring marketplaces and websites, you quickly run into:
Marketplace homepages
Seller landing pages
Pricing pages
Account activation emails
CRM notifications
Duplicate opportunities
Listings with missing financials
Without filtering, your CRM becomes a junk drawer.
So I built a pipeline that:
- Monitors Gmail inboxes
- Monitors marketplaces using Apify
- Uses Claude to determine whether something is actually an acquisition opportunity
- Extracts company name, industry, geography, revenue, EBITDA, asking price, risks, and next actions
- Rejects non-deal content automatically
- Deduplicates opportunities
- Pushes opportunities into Airtable for review
- Syncs approved opportunities into HubSpot
- Creates companies and deals automatically
Tech stack:
n8n
Claude
Airtable
HubSpot
Apify
One thing I found interesting was teaching Claude the difference between:
“Here's a business for sale”
and
“Here's a marketplace homepage talking about businesses for sale”
The first few tests looked successful until I realized the system was confidently analyzing landing pages instead of actual opportunities.
After refining the workflow, it now correctly ignores generic pages and only processes real listings.
The result is a sourcing pipeline that can continuously monitor deal sources without flooding the CRM with garbage data.
Still a few pieces left to build (enrichment, reporting, outreach), but the core engine is now working end-to-end.