LinkedIn engagement automation: comments, replies and lead scoring
A two-stage commenting system that joins relevant LinkedIn conversations, softly introduces a product and flags warm leads.
The problem
Commenting and replying on LinkedIn by hand does not scale. Most automation looks robotic and has no way to notice and answer the conversations it starts.
Challenges
No webhook for replies
LinkedIn offers no webhook to announce that someone has replied to a comment. The system had to poll tracked comments at regular short intervals, which adds scheduling complexity but is the only dependable way to notice a reply and respond while the conversation is still live.
Avoiding a robotic feel
Automated comments are easy to spot when they are generic or push a product straight away. The build had to produce specific, contextual reactions on relevant recent posts, and hold back any product mention until a conversation has genuinely started.
Never replying to ourselves or twice
An automation that answers replies can end up answering its own account, or the same person repeatedly. The system had to filter out its own replies to prevent loops, and cap how many automated replies any one person receives in a thread.
How the system works
Post discovery
The system looks for recent posts, under 24 hours old, that match target keywords and are worth engaging with.
Stage one: automatic comment
A genuine, contextual reaction is posted, along with a like or emoji, and it does not mention any product.
Reply detection by polling
LinkedIn offers no webhooks for this, so the system checks tracked comments every 10 to 15 minutes for new replies to its first comment.
Stage two: soft pitch reply
When someone replies, a second AI response continues the conversation and gently mentions the product, framed as a favour rather than a hard sell.
Reply cap
Database constraints limit automated replies per person per thread, so nobody is messaged repeatedly.
Lead scoring
People who engage repeatedly get a higher lead score and trigger a Slack notification so a human can follow up.
Self-reply filtering
The account's own replies are excluded, which prevents endless reply loops.
Design decisions
Two stages, with the pitch held back
The first comment is a genuine reaction with no product mention. Only when someone replies does a second response continue the conversation and mention the product gently, framed as a favour. That order keeps the contact from feeling like an advert.
Reply caps enforced by the database
Limits on automated replies per person per thread are enforced with unique constraints in the database rather than by workflow logic alone. A constraint holds even if a workflow misfires, so nobody is messaged repeatedly by mistake.
Lead scoring that alerts a human
People who engage repeatedly get a higher lead score, and a Slack notification prompts a person to follow up. We let the automation surface warm leads but leave the decision to reach out personally with a human.
Result in production
A LinkedIn engagement engine that looks and feels like genuine participation, surfaces warm leads automatically and never spams the same person twice.
Engagement on relevant posts continues without someone commenting and replying by hand.
Warm leads surface automatically and reach the team through Slack, so follow-up goes to people showing real interest.
The same person is not messaged repeatedly, so the activity reads as genuine participation.
Who this pattern fits
Built for sales and marketing teams handling high-volume outreach, the architecture also suits any team with similar volume, compliance or coordination challenges. It can be self-hosted and run by a small team.
Tools used
n8n
Workflow runtime, orchestration and version control.
Unipile
The integration layer used for LinkedIn in production.
Supabase
PostgreSQL database, authentication and storage, including the per-thread reply constraints.
Scheduled polling
Regular checks for new replies, since LinkedIn provides no webhook for them.
Related service, use case and industries
Frequently asked questions
How does it notice replies to its comments?
LinkedIn provides no webhook for this, so the system checks tracked comments on a regular short schedule for new replies. When one appears, a second AI response continues the conversation and can mention the product softly.
Who follows up with warm leads?
A person does. People who engage repeatedly receive a higher lead score, and a Slack notification tells the team so they can follow up personally. The automation surfaces the lead; the relationship stays with a human.
Does this spam people?
No. Reply caps and unique constraints per person per thread limit how much anyone is messaged, and the system tracks who has been contacted so it stops at the right point.