This n8n workflow cleans up leads before they enter your CRM. It takes leads from Google Sheets or Telegram, even as free text, and uses AI to turn them into a standard format. Then it checks HubSpot for existing records, enriches only the missing fields with Lusha, and syncs a clean record to HubSpot.
Most CRM problems start at intake, not with enrichment or scoring. Teams have enrichment tools, automation, and dedupe rules, and the CRM still fills up with near-duplicates and records nobody trusts.
That’s because cleanup usually happens after a lead is already in the CRM. This workflow checks every lead at the door, so only clean, matched, and complete records get in.
Start from the “Sync and enrich HubSpot leads from Google Sheets and Telegram with Gemini and Lusha” template. It handles intake → normalization → matching → enrichment → CRM sync in one flow.
Input your Spreadsheet ID in both:
Configure the Telegram node to listen for messages in your designated lead channel or bot. Unstructured text messages will be parsed by Gemini.
In the Switch Logic node, set your fuzzy matching algorithm threshold:
This two-tier matching (hard + fuzzy) prevents duplicates and saves Lusha credits by avoiding unnecessary API calls.
The workflow automatically calls Lusha’s Person and Company APIs only when key fields are missing, like email, role, or firmographic data.
This targeted approach keeps costs efficient at scale. You’re not enriching everything, every time, just filling gaps.
Try sample inputs:
Confirm:
The goal: Clean CRMs aren’t the result of better cleanup jobs. They’re the result of better boundaries. Handle lead intake properly, and everything downstream becomes easier to reason about, easier to maintain, and easier to trust.
Use this n8n workflow. Before creating a record, it checks HubSpot with exact and fuzzy matching, and updates the existing contact if it finds a match.
Yes. If someone sends a Telegram message like “John Smith at Acme Corp, worth reaching out,” Gemini turns it into structured fields.
No. Lusha runs only when key fields are missing, which saves credits.
You set it. The default threshold is 80%. Raise it for stricter matching, or lower it to catch more variations.
Yes. The same logic works with Cal.com, Typeform, or forwarded emails.
One data connection. Works in Claude, ChatGPT, your CRM, or any agent you build.
