Problem
AI consulting firms spend hours on manual research before they can qualify a single prospect. Company background, pain points, automation opportunities, competitive context, and a draft pitch all have to be assembled by hand, work that is repetitive, inconsistent, and doesn't scale beyond a handful of prospects per week.
Solution
A five-stage multi-agent pipeline. Feed it a company name and it returns a complete outreach package. A research agent gathers context, an analysis agent produces a structured brief, an n8n agent maps relevant workflows for that specific client, a presentation agent builds a deck outline, and an outreach agent drafts a personalized cold email.
Results
- 5 specialized agents running in sequence
- Complete outreach package per prospect
- Active pipeline used for client acquisition
Five specialized agents work in sequence: a research agent gathers company context, industry signals, and automation opportunities via Brave Search and Firecrawl; an analysis agent produces a structured client brief JSON with pain points and ROI estimates; an n8n agent builds a custom workflow diagram for that client based on their industry and stack; a presentation agent scaffolds a pitch deck with section headers and key talking points; and an outreach agent drafts a personalized cold email using the brief. Claude API orchestrates all reasoning; n8n handles workflow execution and data passing between stages; all outputs are structured and saved to disk for JT's review before any external action is taken. RAG layer stores previous client briefs to improve relevance of recommendations over time.
Tech Stack
- Python
- n8n
- Claude API
- RAG
- Multi-agent Orchestration
- Google Sheets API