How we automated "Social Listening" to double our LinkedIn engagement
What we set out to do
Manual "social listening" is a grind. We wanted to find every relevant industry conversation happening on LinkedIn in real-time, but we didn't have the hours to manually search keywords, vet profiles, and brainstorm thoughtful comments. The goal? Identify qualified leads through their activity and engage them authentically without the manual labor.
How we set it up
We built a three-stage "Growth Loop" using PhantomBuster, Zapier, and OpenAI:
- The Scraper: We used PhantomBuster to monitor specific industry keywords and scrape every post containing them.
- The Brain (AI Classification): Those posts were sent to Zapier, where OpenAI analyzed the poster’s profile. It asked: "Is this organization a qualified lead for our product?"
- The Filter: If "No," the lead is discarded to save credits. If "Yes," OpenAI drafts a comment that we can use to engage, based on our specific brand voice and the context of their post.
- The Action Station: Everything (the post link + the AI draft) lands in a Google Sheet for our team to review and post daily.
What happened
The results exceeded our expectations:
- 100% Increase in engagement on our company page.
- Significant follower growth by showing up in the "right" comment sections.
- Content Fuel: We now have a database of what our target leads are actually talking about, which dictates our own content strategy.
Tips for others
- Don’t fully automate the "Post" button: We kept the final step manual. AI writes the draft, but a human hits "send." This ensures we maintain a genuine human connection and catch any AI hallucinations.
- Refine your prompt: Be very specific with OpenAI about your "Brand Voice." Tell it to avoid corporate fluff so the comments feel like they’re coming from a peer, not a bot.
Comments
1 comment
This is a really solid setup - thanks for sharing it so clearly.
I like how you’ve kept the human in the loop at the very end. That balance between scale and authenticity is where most of these workflows fall apart, and you handled it well.
Also the “content fuel” point is 🔥 - feels like an underrated byproduct of doing this right.
Curious - have you noticed the quality of conversations (not just engagement) improving over time as you refine the prompts?
Please sign in to leave a comment.