LinkedIn comment lead qualifier

LinkedIn comment lead qualifier

An n8n workflow for LinkedIn posts that ask people to comment a keyword. Give it a post's address through an Airtable row and it
reads the comments and keeps the people who used the keyword. It then looks up each one's profile and company, marks the company
YES or NO against your fit rules and drafts a short message. Everything is saved to Airtable.

What is inside

• workflow.json: the workflow as n8n exports it, 26 nodes: 21 that work and five sticky notes. It holds no credential.
  Placeholders stand in for the Airtable base and tables, the webhook's path and the Apify token in four headers. In
  writeScript they also replace the author's own credibility line, offer and call to action.
• README.md: these notes.

Before you start

• An n8n instance, on n8n Cloud or your own server. A server of your own needs an address Airtable can reach.
• An Apify account and its API token. The workflow calls four Apify scrapers.
• A Gemini API key from Google AI Studio.
• An Airtable account with a personal access token that can read and write records, and a base with two tables, Posts
  and Results (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Google Gemini (PaLM) API credential with your key and select it in Google Gemini Chat Model. That one node
   serves every AI step.
3. In getPost, getComments, getProfile and getCompany, replace YOUR_TOKEN in the Authorization header with your
   Apify token, keeping Bearer in front of it.
4. Add an Airtable Personal Access Token credential and select it in getPostUrl, isDuplicate and addLead.
5. In those three nodes, choose your base and tables where they hold YOUR_BASE_ID, YOUR_POSTS_TABLE_ID and
   YOUR_RESULTS_TABLE_ID.
6. In writeScript, replace YOUR_CREDIBILITY_LINE, YOUR_BONUS_OFFER and YOUR_CALL_TO_ACTION with your own words.
7. In Webhook, replace YOUR_WEBHOOK_PATH with a long random string, then copy the Production URL.
8. In Airtable, add an automation on the Posts table with a Run script action, started by a trigger you choose, such as a
   button field or a new row. Give the script an input variable named recordId, set to the row's id, and paste the
   script below with the Production URL in it.
9. Save, then press Publish at the top of the canvas (or Shift+P).

The Airtable base

Posts holds one row per LinkedIn post, with a field named Post that holds the post's address. Results holds what the
workflow saves, with these columns:

| Column | Type | What the workflow puts in it |
|---|---|---|
| User | text | The commenter's name |
| Role | text | The title of the first job on their profile |
| Industry | text | The business type Gemini gives the company |
| Company | text | The company's name |
| Employees | number | The company's employee count |
| Country | text | The country of its headquarters |
| Qualified? | single select, YES and NO | Gemini's answer on the fit rules |
| Tagline | text | A one-line description from Gemini |
| Website | text | The company's website |
| CompanyLinkedin | text | The company's LinkedIn page |
| UserLinkedin | text | The commenter's profile, which addLead matches on |
| Comment | text | What they wrote |
| Post | text | The post's address |
| Script | long text | The drafted message |

The file's mapping also lists an Email column that nothing fills. The script the automation runs:

const { recordId } = input.config();
await fetch('YOUR_PRODUCTION_URL', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ id: recordId }),
});

What the webhook takes

Webhook takes a POST with the row's id in id, and getPostUrl reads that row from Posts:

{
  "id": "recXXXXXXXXXXXXXX"
}

How a run goes

1. getPostUrl reads the Posts row, and getPost scrapes the post at the address in its Post field.
2. getComments scrapes up to 100 of the latest comments. filterAuthor drops the post author's own and any with no name.
3. Loop Over Items takes one commenter at a time. hasIntent asks Gemini whether the post names a keyword and the
   comment uses it. A NO goes back to the loop.
4. A YES goes to getProfile. isDuplicate searches Results for that name, and If lets the commenter on only when
   the name is new and the first job on the profile links to a LinkedIn company page.
5. getCompany reads that page. qualifyLead has Gemini classify the company and answer YES or NO, and companyInfo
   gathers the fields.
6. writeScript drafts the message, and addLead upserts the row into Results, matched on UserLinkedin. Every
   commenter who gets this far is saved, with YES or NO in Qualified?.

Change it

• Who counts as a fit: the criteria in the message of qualifyLead. As exported: any industry, a headquarters in the
  United States, Canada or Europe, 10 to 500 employees and any role.
• The keyword rule: the message of hasIntent.
• The message: the system message of writeScript (first name only, up to 80 words, no emoji) and the three lines you
  replaced in its text.
• How many comments: limit and page_number in the body of getComments, 100 and page 1 as exported.
• Saving only the YES rows: add an If node after companyInfo on qualified, with its false branch going back to
  Loop Over Items.
• The columns: the mapping in addLead.

Good to know

• A run costs two Apify scrapes for the post and its comments, then up to two more for each commenter who uses the
  keyword. Gemini is called once per comment for the keyword check, and twice more for each commenter who goes on. The
  first sticky note links the four actors by id. Open each to see its price, since some Apify actors charge per result
  or by rental.
• The workflow drafts messages and sends nothing. Read each draft before you use it.
• isDuplicate compares names, so two people with one name count as one. A name with an apostrophe may break its
  Airtable formula.
• getProfile and getCompany are set to continue when a scrape fails, so one failure does not stop the run.
• Anyone who learns the webhook's address can start a run on your keys. That is why its path should be long and random.
• The Apify token sits in the nodes' headers. Take it out before you share or export the workflow.
• LinkedIn's terms restrict automated collection, and the people listed are named individuals. Check the rules that
  apply to you, data protection law included, before you run it on a real post.