Personalized cold email drafts from a lead search

Personalized cold email drafts from a lead search

An n8n workflow. A form takes a description of your product, a job title, a place and some keywords. An Apify Actor finds
matching leads with verified emails, the workflow reads each lead's website, and GPT-4.1 drafts a cold email for each
one. The drafts land in a Google Sheet and nothing is sent.

What is inside

• workflow.json: the workflow as n8n exports it, twenty nodes: nineteen that run and one sticky note with the title.
  It holds no credential.
• README.md: these notes.

What we changed from the source

• The author's sheet, the sample product and the sample lead are gone. The sample lead was a named person, so the
  example in Generate Multiline Icebreaker is now an invented one.
• The old example reply answered with an icebreaker key, while the instructions and Add Row use email. The new
  example uses email.
• The source prompt had nn and n- where line breaks were meant. We wrote \n there.
• join("/n") in the last message, which put the text /n between page summaries, is now join("\n").
• The form's placeholders no longer name the author's product, name or site.

Before you start

• An n8n instance, on n8n Cloud or your own server.
• An Apify account and its API token, with credit for the lead search.
• An OpenAI API key that can use gpt-4.1.
• A Google account, and a Google Sheet for the drafts.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. In Leads Scraper, replace YOUR_TOKEN in the Authorization header with your Apify API token. Keep Bearer
   before it.
3. Add an OpenAI credential and select it in Summarize Website Page and Generate Multiline Icebreaker.
4. In Google Sheets, make a sheet with a tab and these headers in the first row: first_name, last_name,
   designation, email, website_url, location, phone_number, Outreach Content.
5. Add a Google Sheets OAuth2 credential and select it in Add Row. Replace YOUR_SHEET_ID and YOUR_SHEET_TAB_ID
   with your sheet's ID and the tab's gid, or switch both fields to From list and pick them. The sheet ID sits between
   /d/ and /edit in the sheet's address, and the gid comes after gid=.
6. Read the email template in Generate Multiline Icebreaker and change what it claims about the product (below).
7. For a first run, lower max_result in the body of Leads Scraper from 100 to 3 or 5. Save, open the form's test
   address, fill it in and watch the sheet.
8. Press Publish at the top of the canvas (or Shift+P) when you are happy.

How a run goes

1. The form asks for About your Product, Designation, Location, Keywords (optional), Your Name and
   Product URL.
2. Code in JavaScript percent-encodes every text field, so Product Managers reaches the search as
   Product%20Managers. Its comment says it also lowercases, but the code only encodes. If a search with spaces returns
   nothing, skip this node for the search fields.
3. Leads Scraper posts the search to an Apify Actor and waits for its dataset. Apify's listing for the Actor ID in the
   address, read on 2 November 2026, names it Leads Finder by microworlds. The body asks for verified emails, leaves
   out catch-all domains and returns at most max_result leads.
4. Only Websites & Emails drops leads without a website or an email. Scrape Home fetches each home page, HTML
   collects its links and Edit Fields keeps the lead's details beside them.
5. Loop Over Items takes one lead at a time. Split Out, Filter, Code and Remove Duplicate URLs turn the links
   into a list of different internal paths, and Limit keeps the first three.
6. For each path, Request web page for URL, Markdown and Summarize Website Page fetch the page and have GPT-4.1
   write a two-paragraph abstract. Aggregate gathers them.
7. Generate Multiline Icebreaker writes the email from the abstracts, the lead's name and headline, and your product
   description. Add Row appends the lead's details and the draft to the sheet, and the loop moves to the next lead.

Filter keeps only links that start with /, so absolute links to the same site never reach Code. The home page
itself is not summarized, only up to three pages linked from it.

Change the look

• Number of leads: max_result in Leads Scraper.
• Who counts as a lead: the other fields of the same body, such as email_status and
  contact_email_exclude_catch_all_domains.
• Pages read per lead: the number in Limit.
• The email: the second message in Generate Multiline Icebreaker, which holds the structure and the rules. Keep the
  email key in its JSON format, because Add Row reads message.content.email.
• Columns: the mapping in Add Row. headline and links are not saved.

Good to know

• OpenAI bills up to four GPT-4.1 calls per lead, three summaries and one email. Apify bills the lead search. The
  Actor's listing, read on 2 November 2026, shows a small start fee and a charge per lead that depends on your Apify
  plan.
• Nothing is sent. The template tells the model to say the tool costs a few cents per run and consistently converts.
  Keep only claims your product can back.
• The leads are named people with email addresses and phone numbers. Check the cold-email and data protection rules where
  you and they are before you contact anyone.
• Add Row appends, so the same search run twice adds the same people twice.
• A run through many leads can reach the time limit of your n8n plan. Lower max_result first.
• Your product description and each lead's website text go to OpenAI. Your search goes to Apify.