AI Lead Gen Agent: 50 Personalized Leads to Inbox Weekly
Built a 37-node n8n agent that searches Apollo, enriches via Clay, scores leads with Claude across 6 dimensions, routes into 3 tiers, then loads tiered personalized email sequences into Instantly. Zero manual prospecting.
Watch the walkthrough
3-6 minute screen-share showing Problem → Solution → Result
The Problem
Manual Prospecting That Does Not Scale
Finding qualified leads meant bouncing between LinkedIn, Apollo, and company websites. Each prospect required 15-20 minutes of research before writing the first email. At that rate, 10 prospects per day was the ceiling. For a growing consultancy, that volume was not enough to fill a pipeline.
Generic Outreach Killing Response Rates
Without enrichment data, every prospect got the same templated email. No reference to tech stack, recent funding, or hiring patterns. Response rates sat below 2%. The emails read like spam because they were written without context.
No Lead Scoring or Prioritization
Every lead was treated equally. A CEO at a recently funded 30-person SaaS got the same outreach as a director at a 500-person enterprise with no buying signals. Time was wasted on leads that were never going to convert, while spreadsheet-based CRM gave zero visibility into which leads had been contacted or went cold.
The Solution
Architecture diagram — click to zoom
Perceive: ICP Config + Apollo + Clay Enrichment
A single config node holds all ICP parameters: target titles, location filters, company size range (11-50 employees), sender identity, value proposition, API credentials. Apollo People Search finds matching contacts. Results deduplicate by email. Each lead pushes to Clay for enrichment: tech stack, funding stage, recent funding rounds, hiring signals, LinkedIn activity, company bio.
Decide: AI Scoring + Tier Classification
A code node scores each lead 0-100 across 6 dimensions: title match (25 pts), company size fit (20), funding signals (20), hiring signals (15), tech stack depth (10), LinkedIn activity (10). Leads classify into Hot (70+), Warm (40-69), Cold (under 40). This is where AI credit and research effort route to the highest-value leads automatically.
Act: Tiered Personalized Outreach
Hot leads trigger the full treatment: website scrape, HTML-to-text clean, Claude generates a deep research brief plus a 3-email sequence (connection observation → relevant insight → value prop with booking link). Each email under 100 words. Warm leads get a lighter touch: 2-sentence research note plus 2-email sequence. Cold leads mark as Skipped in Airtable. Every sequence pushes to Instantly for automated sending.
Guardrails and Audit
Every lead lands in Airtable with score, tier, enrichment data, research brief, and email sequence. Continue-on-fail error handling wraps every external API call (Clay, website scrape, Instantly). Cold-lead filter stops wasted API credits. Slack notifications post after each run with lead counts and ICP summary. Modular design: swap Apollo or Instantly without touching scoring logic.
The Impact
Quantitative Results
- 50 leads per week enriched, scored, and loaded into personalized sequences automatically
- Zero manual prospecting time after ICP configuration
- Hot leads receive 20+ minutes of AI-generated research per lead in seconds
- 37 nodes across 5 stages, modular swap-in for Apollo or Instantly
Strategic Value
- AI credits and human-equivalent research effort route to the highest-value leads automatically. Cold leads filter out before consuming resources
- Originally shipped as a client deliverable, now the production lead generation engine for DTS Consulting. Proven twice
Have a similar problem?
Tell me what is going on and I will tell you what I would do about it. No obligation.
Get in touch