1 Job Ad, 7 Boards, 3 Minutes: The n8n Poster Agent

By Lazar Milicevic · Published July 22, 2026 · 7 min read

Every recruiting AI demo starts at resume screening. That's the wrong end of the funnel. If you hire under 20 roles a year, you're not drowning in resumes — you're drowning in copy-pasting the same job description into seven boards and having zero clue which one delivered your last hire.

Here's the n8n workflow that fixes the sourcing side: one JD, seven boards, three minutes, every applicant tagged by source automatically.

The real bleed: you're renewing the wrong board

Before a single resume exists, someone on your team just spent 47 minutes logging into LinkedIn, then Indeed, Glassdoor, Wellfound, and three niche boards. Same JD pasted seven times. Salary field reformatted for each. Logo re-uploaded. Three weeks later a hire lands and nobody can tell you which board it came from.

So next quarter you renew the $84/month LinkedIn slot because it feels safe, and you cancel the $6/month niche board that actually delivered the hire. That's the real bleed. Screening AI stacked on top of a broken sourcing funnel is polishing the wrong thing.

The fix is two things:

• One-click multi-board posting from a single source of truth.
• A tracking token on every apply URL so source attribution is automatic, not manual.

Neither needs a SaaS subscription. Both fit in n8n plus Airtable.

Intake: one Airtable row per role

The whole system starts from a single Airtable table called roles. One row equals one open req. Columns:

• role_title (single line)
• jd_markdown (long text, markdown)
• salary_band (e.g. $95k–$120k)
• location (single line)
• employment_type (single select: full-time, contract, part-time)
• target_boards (multi-select: LinkedIn, Indeed, Wellfound, Hacker News Who's Hiring, Dribbble Jobs, RemoteOK, WeWorkRemotely)
• status (single select: draft, posting, live, closed)

The multi-select is the switch. Tick the boards you want, flip status to posting, save. An Airtable automation fires a webhook into n8n with the row ID as payload.

Markdown for the JD is deliberate. It's the one format that survives conversion into HTML, plain text, JSON payloads, and email bodies without turning into a mess of escaped quotes.

The normalizer: one JD, seven shapes

Each board wants the JD in a different shape. If you skip this step you'll spend the rest of your life fighting formatting bugs. In n8n this is a single Function node that takes the markdown JD and outputs a keyed object.

// n8n Function node - JD normalizer
const md = $input.item.json.jd_markdown;
const salary = $input.item.json.salary_band;
const role = $input.item.json.role_title;

// LinkedIn: HTML with specific allowed tags
const linkedinHtml = md
  .replace(/^### (.*)$/gm, '<h3>$1</h3>')
  .replace(/^## (.*)$/gm, '<h2>$1</h2>')
  .replace(/^\* (.*)$/gm, '<li>$1</li>')
  .replace(/(<li>.*<\/li>\n?)+/g, m => `<ul>${m}</ul>`)
  .replace(/\*\*(.*?)\*\*/g, '<strong>$1</strong>');

// Indeed: plain text, line breaks preserved
const indeedText = md.replace(/[#*_`]/g, '').trim();

// Wellfound: structured JSON
const wellfoundPayload = {
  title: role,
  description_html: linkedinHtml,
  salary_min: parseInt(salary.match(/\$(\d+)k/)[1]) * 1000,
  salary_max: parseInt(salary.match(/–\$(\d+)k/)[1]) * 1000,
  remote_policy: 'remote_ok',
  equity_min: 0.0,
  equity_max: 0.5
};

// Niche boards: email body
const emailBody = `New role: ${role}\nSalary: ${salary}\n\n${md}`;

return {
  linkedin: linkedinHtml,
  indeed: indeedText,
  wellfound: wellfoundPayload,
  email_body: emailBody
};

That node outputs one payload with seven keys. Downstream nodes pull the shape they need. Add a new board later, add a new key here. That's the only place format logic lives.

Fan-out: Switch node, one branch per board

A Switch node reads target_boards and routes to enabled branches. Each branch is either an HTTP Request node (for boards with real APIs) or a Send Email node (for boards that accept email submissions).

Rough breakdown of what actually works in practice:

Rate-limit note: LinkedIn's Talent Solutions API needs partner approval and approval takes weeks. If you're posting fewer than five roles a month, the email-to-post pattern via LinkedIn Recruiter or scheduled browser automation is what most solo operators actually use. Skip the API pain until volume justifies it.

Every branch ends the same way: a Create Record node writes to an Airtable postings table. Schema:

• role_id (link to roles)
• board (single select)
• posted_at (datetime)
• external_id (whatever ID the API returned, or empty for email)
• tracking_token (unique string, generated with crypto.randomUUID())
• apply_url (the tokenized URL — see below)

The tracking token: the piece 90% of builds skip

This is the whole trick. Every board lets you set an apply URL. Instead of pointing that URL at your careers page, point it at a webhook on your n8n instance with the tracking token as a query parameter.

https://n8n.yourdomain.com/webhook/apply?t=8f2a91b3-4c7d-4e12-9f8a-1b2c3d4e5f6a

When an applicant clicks apply on Indeed:

1. Browser hits your n8n webhook with the token.
2. n8n looks up the token in the postings table.
3. Reads which board that token belongs to.
4. 302-redirects the applicant to your hosted application form (Tally, Fillout, Typeform, whatever) with ?source=indeed pre-filled in a hidden field.
5. When they submit, the source rides along into your applicants table.

// n8n webhook handler for apply clicks
const token = $input.item.json.query.t;

// Airtable lookup happens in the next node; assume it returns:
const posting = $node["Airtable Lookup"].json;

const formUrl = `https://tally.so/r/YOUR_FORM?source=${posting.board}&role_id=${posting.role_id}`;

return {
  statusCode: 302,
  headers: { Location: formUrl }
};

Now every applicant has a board of origin, automatically. No human tagging. No self-reported "how did you hear about us" dropdown that everyone lies on.

The view that changes how you spend money

Once source tags are landing on every applicant, build one Airtable view that will change your entire hiring budget.

Group applicants by source_board. Add a formula field qualified — whatever your bar is. For a mid-level engineering role that might be:

IF(
  AND(
    {years_experience} >= 3,
    {portfolio_url} != "",
    {location_match} = TRUE()
  ),
  1, 0
)

Then a rollup on the boards table: monthly spend divided by count of qualified applicants from that board in the same window. That's your real cost per qualified applicant (CPQA).

In a test batch I ran for a small agency hiring three roles over six weeks, the numbers landed like this:

Same JD. Same three minutes of setup work. Wildly different economics. They killed the LinkedIn slot the next cycle and doubled the niche board budget. First hire came from Wellfound. Second from the $6/qualified niche board. LinkedIn produced volume but almost no qualified conversions for that specific role.

You can't make that call without source attribution. And you can't get source attribution without the tokenized apply URL. Everything else is theater.

What this workflow does not do

• Rank resumes.
• Draft outreach messages.
• Schedule interviews.
• Rewrite the JD for tone or SEO.

Those are separate workflows, triggered by new rows in the applicants table. Same n8n instance, different graph. The point of this piece is the sourcing layer. Get that right, then bolt screening on top.

Why bizflowai.io helps with this

The poster agent is one of the smaller pieces we ship for clients running lean hiring ops. Most builds we deploy through bizflowai.io combine this sourcing layer with the downstream screening workflow — resume parsing, JD-match scoring, first-touch outreach — so the source tag survives all the way through to hire. That's how you get a closed-loop CPQA number that actually informs next quarter's board budget instead of guessing based on which invoice arrived most recently.

Want more like this?

I publish practical AI automation, GenAI engineering, and faceless content workflows on YouTube every week.

Subscribe to bizflowai.io on YouTube — never miss a new tutorial.

Planning an AI automation project or need a second opinion on your architecture?

Connect with me on LinkedIn — Lazar Milicevic, GenAI Engineer & bizflowai.io Founder.

Visit bizflowai.io for our services, case studies, and AI consulting.