How to Use Bardeen to Automate LinkedIn Outreach
LinkedIn outreach at scale has two failure modes: you copy-paste the same message to everyone and get ignored, or you personalize every message by hand and run out of hours in the day. Bardeen sits in the middle. You scrape a LinkedIn search, have AI draft a message based on each person's profile, and push the results to a spreadsheet where you review before sending. The actual sending still happens manually, which keeps you within LinkedIn's terms of service.
Bardeen is a Chrome extension, so the setup lives in your browser rather than a separate web app. That means it can interact with LinkedIn's actual UI rather than relying on an API that LinkedIn actively blocks for third parties. Here's how to build the playbook.
Install Bardeen and Create Your First Playbook
If you haven't installed it yet, get the Bardeen extension from the Chrome Web Store. Once installed, click the Bardeen icon to open the side panel. You'll create an account and get access to the playbook library.
Playbooks are Bardeen's term for automated workflows. You can build from a template or from scratch. For LinkedIn outreach, there's usually a template called something like "Scrape LinkedIn Search to Google Sheets" in the library. That's a good starting point, but you'll want to customize it significantly to add AI drafting.
For this walkthrough, we'll build from the template and then add the AI step.
Step 1: Set Up the LinkedIn Scraper
- In the Bardeen panel, search for "LinkedIn" in the playbook library.
- Open the template for scraping LinkedIn search results. It typically has three steps: scrape the current page, loop through profiles, and save to a sheet.
- Before you run anything, navigate to a LinkedIn search in your browser. Use Sales Navigator if you have it (the filter options are far more precise), or use the standard LinkedIn search filtered by "People".
- Apply your filters: job title, company size, location, industry. This is the list you're going to scrape.
When you run the scraper step, Bardeen reads the visible search results from the current page. It captures name, title, company, profile URL, and sometimes location depending on what LinkedIn displays.
The scraper reads what's visible on screen, not LinkedIn's internal data. That's the key distinction from tools that hit the LinkedIn API directly. It's slower but significantly less likely to trigger a restriction on your account.
Step 2: Add an AI Step to Draft Personalized Messages
After the scraper collects profile data, you add an AI action to draft a message for each person.
- In the playbook editor, click Add step after the scraper loop.
- Choose AI Prompt from the action list (Bardeen uses OpenAI under the hood for this).
- Write your prompt. The trick is to reference the scraped fields as variables. In Bardeen's interface, you insert variables using double curly braces, so a prompt looks like this:
Write a short, casual LinkedIn connection request message for someone with this profile:
Name: {{name}}
Title: {{title}}
Company: {{company}}
The message should:
- Be 2-3 sentences max
- Reference their role or company specifically (not generically)
- Not mention "synergies", "touching base", or any corporate clichés
- Sound like it came from a human, not a template
Do not include a subject line. Just the message body.- Map the output to a variable like
draft_message.
This step runs for each profile in the loop, generating a unique draft per person. On GPT-4o-mini, this costs fractions of a cent per message, so the economics work at volume.
Step 3: Push Everything to a Google Sheet
Add a Google Sheets: Add Row step after the AI draft step.
Map the columns:
| Sheet Column | Bardeen Variable |
|---|---|
| Full Name | {{name}} |
| Title | {{title}} |
| Company | {{company}} |
| LinkedIn URL | {{profile_url}} |
| Draft Message | {{draft_message}} |
| Status | "to review" (static text) |
| Date Added | Bardeen's {{current_date}} |
The Status column is important. It becomes your workflow tracker. Change it to "sent", "skipped", or "follow up" as you work through the list.
Reviewing and Sending
This is the step Bardeen can't automate for you, and honestly you shouldn't want it to. Every AI-drafted message needs a human eye before it goes out. The drafts are usually 80-90% good, but occasionally the AI misreads a title or writes something that sounds off for that person's specific context.
In practice, the review takes about 30 seconds per message. You read the draft, tweak the one phrase that doesn't land, copy it, and send the connection request manually on LinkedIn. For a list of 50 people, the whole review-and-send process takes maybe 45 minutes. Compare that to writing 50 personalized messages from scratch (3-4 hours) or sending the same copy to all 50 (fast, but near-zero response rate).
Rate Limits and Account Safety
This is not an optional section. LinkedIn monitors connection request volume and will temporarily restrict accounts that send too many in a short window. The specific limits aren't published, but the safe zone that practitioners report is roughly:
| Activity | Daily limit to stay safe |
|---|---|
| New connection requests | 20-25 |
| Profile views | 80-100 (with Sales Navigator) |
| Messages to connections | No hard limit, but pace them |
Bardeen doesn't enforce these limits for you. If you build a playbook that scrapes 200 results and tries to send them all in an hour, your account takes the hit. Build the pacing into your process: run the playbook once a day, work through the review sheet in batches, send manually.
One more thing: avoid scraping while on a spotty connection or VPN. Bardeen reads the live page DOM, and slow or incomplete page loads result in incomplete data.
Pushing to a CRM Instead of a Sheet
If your outreach workflow ends in a CRM rather than a spreadsheet, Bardeen has direct integrations with HubSpot, Salesforce, Pipedrive, and Notion. Replace the Google Sheets step with a CRM "Create Contact" step and map the same fields.
The draft message can go into a Note field on the contact, giving your sales reps the AI-generated context right inside the CRM alongside the contact record. Some teams find this more useful than a separate tracking sheet.
The playbook takes about 20 minutes to configure the first time. After that, it runs in under five minutes to scrape a fresh search, draft messages, and populate the sheet. The bottleneck becomes your own review time, which is exactly where the bottleneck should be.