LinkedIn GTM Lead Engine
The LinkedIn GTM Lead Engine is a fully autonomous prospecting system that mines three different LinkedIn surfaces for buyer signals — active job posts, user posts, and comment threads under high-engagement posts — then enriches, scores, and sequences the leads into a live drip campaign.
Manual LinkedIn prospecting caps at ~30 quality leads per rep per week. Comment threads under viral posts are gold — but nobody has time to scrape and enrich them at scale.
Three parallel Apify actors (job posts, user posts, post comments) feed one enrichment pipeline. Each raw lead gets email-extracted, LLM-scored for fit, deduped against the CRM, then dropped into a 21-day Brevo drip. Replies and unsubscribes auto-suppress mid-sequence. Runs weekly, unattended.
Workflow breakdown
Phase 1 — Multi-surface Apify scrape
Three Apify actors run in parallel for maximum surface coverage.
- LinkedIn Jobs actor filtered by target ICP keywords + geography
- User Posts actor scraping recent activity from a maintained watchlist of accounts
- Post Comments actor targeting high-engagement threads (posts above N reactions)
- All three actors write raw results into a shared n8n execution context
Phase 2 — Enrichment & scoring
Every raw record gets a full enrichment pass before it's allowed into a sequence.
- Email extraction via multi-provider fallback (Apollo → Hunter → regex on public profile)
- GPT-4 fit-scoring against the client's ideal customer profile (JSON output, cached per contact)
- Deduplication against Supabase (prior sequenced contacts + suppression list)
- Enrichment failures land in a manual-review queue, not the sequence
Phase 3 — Brevo 21-day drip
Scored leads flow into a live sequence with automatic suppression.
- 21-day sequence: 5 touches, spaced with intentional gaps (not daily hammering)
- Reply detection via Brevo webhooks auto-pauses the sequence for that contact
- Unsubscribes propagate back to the suppression list in Supabase
- Weekly report auto-generated: leads in, leads scored, leads sequenced, replies received
Technical highlights
Parallel Apify orchestration
- Three actors invoked concurrently, results merged with source-attribution retained
- Actor failures don't block the other two — partial pipeline runs still ship value
LLM scoring with structured output
- GPT-4 called with strict JSON schema + validation retry loop
- Score, tier, and 1-line reasoning stored per contact for audit
- Cache layer prevents rescoring the same LinkedIn profile within 30 days
Suppression as a first-class concept
- Every outbound touch checks the suppression list before firing
- Reply, unsubscribe, and manual-flag all feed the same suppression store
- No lead ever gets two competing sequences from adjacent runs
Why this matters
Comment mining alone unlocks a lead surface most tools ignore entirely
The full pipeline runs unattended — the operator's job becomes closing replies, not prospecting
Every enrichment and suppression decision is logged, so campaigns are auditable and defensible