From 30 Messages a Day to 300: Building a Signal-Based LinkedIn Outreach System
March 2026
The Story
A client reached out with a familiar story. They were doing LinkedIn outreach the hard way — 30 messages a day, all manual. Copy-paste a personalized opener, send it, tag the lead by hand, move on. When someone replied “maybe later,” that lead was basically gone. No system to follow up. No way to know who was worth chasing. The frustrating part? Their reply rate was already 51% — well above industry average. They weren't bad at outreach. They just couldn't keep up with it.
They didn't need a bigger team. They needed a system that could think.
Why Most LinkedIn Outreach Fails
Here's what most people get wrong about scaling outreach: they think it's a volume problem. Send more messages, get more replies. But only a small fraction of your target market is actively looking to buy at any given time. Most of your messages are landing on people who were never going to convert — no matter how good your copy is.
So before building anything, I flipped the strategy. Instead of targeting by job title and geography alone, I designed a signal-based scoring system that identifies companies showing real buying intent — things like hiring activity, leadership changes, and growth patterns.
Each signal gets a weighted score. Layer that with behavioral indicators like profile activity and engagement patterns, and you stop guessing who to reach out to. The pipeline only feeds high-probability leads into outreach campaigns. Better targeting in, better replies out.
The System: 8 Workflows Compressed Into 3
The initial design called for 8 separate n8n workflows. That's manageable on paper, but in practice it means 8 things to monitor, 8 things that can break independently, and a tangle of sub-workflow calls passing data between them. So I consolidated everything into 3 core workflows:
Lead Acquisition Pipeline
Runs on a daily schedule. Detects buying signals, discovers matching leads, scores and deduplicates them, assigns a priority tier, and stores everything to a tracking sheet. By the time a human looks at the pipeline, every lead already has a reason attached to why they're worth reaching out to.
Lead Engagement Engine
Listens for incoming replies via webhook. An AI classifier reads each response and categorizes the intent — interested, not now, asking a question, out of office. Hot leads get an instant booking message drafted with timezone-aware scheduling. Warm leads get a contextual follow-up queued. Cold leads get tagged and archived. No lead falls through the cracks.
Pipeline Intelligence
Two jobs in one. First, it logs every outreach event into a pipeline tracker in real time. Second, it runs a weekly report — aggregating performance data, flagging stale leads for re-engagement, and generating an AI-written summary of what's working and what isn't.
The Tools Behind It
Every piece of this system runs on tools the client was either already using or could add for under $150/month:
- •n8n — The backbone. All 3 workflows live here — scheduling, webhooks, AI calls, data routing, error handling. Self-hosted on Railway.
- •Prosp.ai — Manages LinkedIn campaigns, sends connection requests and InMails, and fires webhook events back to n8n whenever a lead replies or accepts a connection.
- •Evaboot — Exports leads from Sales Navigator and detects Open Profile status, which determines whether a lead gets an InMail or a connection request.
- •Apify — Enriches leads with company-level data used for signal scoring. Lightweight, pay-per-use.
- •Claude AI (Haiku) — The brain. Classifies reply intent, drafts follow-up messages, writes booking messages with timezone logic, and generates weekly performance summaries.
- •Google Sheets — The client's CRM. Four tabs: Leads, Pipeline, Follow-Ups, Reports. Simple, familiar, no onboarding friction.
- •Sales Navigator — Prospecting and InMail sending. The client already had this — we just plugged it into the automation layer.
No enterprise software. No six-figure platform licenses. Just sharp tools wired together intelligently.
Scaling Without Getting Banned
LinkedIn doesn't like automation. Push too hard, too fast, and your account gets restricted. So the system includes a built-in volume ramp — starting at 50 messages a day and gradually scaling to 300 over 8 weeks.
The system also knows which leads have Open Profiles and routes them to InMail campaigns automatically — free, unlimited, and less monitored by LinkedIn. Connection requests stay capped at safe daily limits.
The best part: the entire system delivers value from day one, even at low volume. Auto-tagging, reply classification, follow-ups, and pipeline tracking all work whether you're sending 50 or 300 messages a day. Scaling just means turning the dial up.
How I Built It
This entire system — 3 workflows, 51 nodes, AI integrations, webhook routing, Google Sheets CRM — was built using Claude Code as my development partner.
Claude Code (CLI)
Planned the architecture, wrote the workflow logic, consolidated 8 workflows into 3, and debugged node configurations through conversation.
n8n MCP Server
Let me inspect, create, and modify workflows directly from the terminal. No switching between browser tabs and code editors.
Custom n8n Skills
Expression syntax, node configuration patterns, and workflow architecture guidance baked into the development environment.
GitHub MCP
Repo management and version control without leaving the conversation.
From first design to deployed workflows: built in a single session.
The Takeaway
This project reinforced something I keep coming back to: the biggest wins in automation aren't about doing things faster — they're about doing the right things in the first place.
Before this system, the client had a 51% reply rate but couldn't convert consistently. Leads went cold. Follow-ups didn't happen. The pipeline had no memory. After — every reply gets classified instantly, every warm lead gets a follow-up, every hot lead gets a timezone-aware booking message, and a weekly AI-generated report tells you exactly what's working.
You don't need a sales team of 10 to run outreach at scale. You need the right signals, the right automation, and a system that never forgets a lead.
Built with Claude Code, n8n, Claude AI, Prosp.ai, Evaboot, Apify, and Google Sheets.