Skip to content

README.md

Latest commit

 

History

History
44 lines (31 loc) · 2.86 KB

README.md

File metadata and controls

44 lines (31 loc) · 2.86 KB
Raw

Autonomous AI Lead Ingestion & Omnichannel Conversion Pipeline

An event-driven automation system designed for North Peak Heating & Cooling to capture, qualify, and convert inbound business leads autonomously. By integrating an LLM qualification layer directly into the data ingest pipeline, this system minimizes manual triage, shortens response times from hours to under a minute, and ensures consistent follow-up sequencing.

🚀 System Architecture & Workflow

The system is split into two primary operational scenarios that execute autonomously based on state changes in the underlying database.

1. Main Lead Processing Pipeline

  • Trigger: Instantly watches for new inbound lead rows populated via custom customer forms into the core tracking database.
  • AI Qualification Layer: Passes raw lead metrics (Service requested, Urgency level, Homeownership status, Location, and Issue description) to Groq AI (llama-3.3-70b) using an optimized system prompt payload.
  • JSON Parsing & Database Enrichment: Extracts structural JSON elements (score, urgency, summary, recommended_action) from the LLM response, injecting them back into the data pipeline to classify the lead's hot/cold state.
  • Omnichannel Routing: Simultaneously triggers a personalized, dynamic auto-reply via Gmail (embedding self-booking scheduler links) and shoots a structured emergency alert straight to the internal sales team via a Telegram Bot API.

2. Multi-Day Follow-Up Sequence

  • Evaluates non-converted records daily using conditional split-routing.
  • Day 1 & Day 2: Fires contextual, personalized email sequences reminding the lead to claim their consultation slot.
  • Day 3: If still unbooked, automatically flags the lead state as Lost in the pipeline database and closes out the active instance cleanly.

🛠️ Technical Stack & Frameworks

  • Orchestration & Infrastructure: Make.com (Event-driven visual canvas routing)
  • Intelligence Engine: Groq AI Engine (llama-3.3-70b Neural Network via custom HTTP endpoints)
  • Data Layer: Google Suite (Forms for front-end ingestion, Sheets as a light, relational CRM schema)
  • Messaging & Interface Fabrics: Telegram Bot API (Team notification arrays), Calendly API (Self-service appointment booking engine), Gmail SMTP (Dynamic outreach)

🧠 AI Agent Prompting Strategy

The Groq AI interface relies heavily on strict JSON constraint prompting to eliminate string hallucination and ensure deterministic data injections back into the database columns.

System Target Payload

{
  "score": (Value 1-10 evaluated dynamically on homeownership status, complete contact details, and core emergency context),
  "urgency": "HIGH" | "MEDIUM" | "LOW",
  "summary": "Max 2-sentence lead context profile extraction.",
  "recommended_action": "Actionable strategy rule generated for internal agents."
}