Building visibility on LinkedIn through genuine engagement (commenting on relevant posts, replying to responses) doesn't scale manually, but most engagement automation is either obviously robotic or has no way to detect and respond to replies.
- 01Post discovery
The system searches for recent LinkedIn posts (under 24 hours old) matching target keywords.
- 02Stage 1: auto comment
The automation posts a genuine, specific reaction to the post (never mentioning any product at this stage), plus a reaction/like.
- 03Reply detection via polling
Since LinkedIn offers no webhook for comment replies, the system polls each tracked comment on a schedule (every 10 to 15 minutes), checking for new replies.
- 04Stage 2: soft pitch reply
If someone replies to the original comment, a second stage AI reply engages further and, only at this stage, introduces a soft, favor framed mention of the relevant product, never a hard pitch.
- 05Reply cap
Each person is capped at a maximum number of automated replies per thread (tracked via a unique constraint on person + comment), so the automation never spams the same person.
- 06Lead scoring
People who engage repeatedly are scored as warmer leads and flagged (including Slack alerts) for human follow up.
- 07Self reply filtering
The system always filters out the account's own replies to prevent infinite reply loops.
A LinkedIn engagement engine that looks and feels like genuine participation in a conversation, surfaces warm leads automatically, and never spams the same person twice.
Originally built for the sales & marketing, the same architecture applies to any team facing a similar volume, compliance, or coordination problem. The stack (n8n, Unipile, Supabase, and others) is stable, self hosted where it matters, and designed to be operated by a small team without a platform crew.
- LinkedIn comment automation
- LinkedIn engagement automation
- LinkedIn reply automation
- n8n LinkedIn automation
- n8nWorkflow runtime and orchestration. Every scheduled job, webhook, and branch lives here so the logic is version controlled and observable.
- UnipilePart of the production stack for this build.
- SupabasePostgres, auth, and storage. Holds the source of truth for records the workflows read and write.
- Scheduled pollingPart of the production stack for this build.
I would separate the enrichment pipeline from the sending pipeline more strictly. Coupling them made it too easy for a bad enrichment run to poison outreach quality for a day.
How to automate LinkedIn engagement without it feeling automated. A strong authority piece on the two stage reply design and reply caps.
Does this spam people?+
No. Reply caps and unique constraints per person per thread prevent over messaging. The system tracks who has been contacted and stops when it should.