← sources

I built a lead-research pipeline and deliberately did not use an AI Agent.

source: reddit (reddit:AI_Agents)published 2026-09-23 06:01ingested 2026-09-24 02:36

retained text

I built a production-style prototype for researching and qualifying marketing agencies in Bengaluru using n8n.

The workflow:

* Accepts a niche, location, and company limit
* Searches with Tavily
* Removes directories and duplicate domains
* Fetches and cleans website content
* Uses ScrapeGraphAI when normal extraction fails
* Uses OpenRouter for structured lead analysis
* Scores leads from 0–100
* Stores results in Google Sheets
* Generates a Markdown report
* Logs failures through a separate Error Handler workflow

One design decision I found interesting: I used a deterministic pipeline with a Basic LLM Chain instead of an AI Agent.

The AI analyzes the company, but JavaScript enforces the score boundaries and qualification rules:

* 80–100: High
* 60–79: Medium
* 0–59: Low

I chose this because lead scoring should be predictable and auditable. An agent could introduce unnecessary tool-selection and execution variability.

What would you improve?

Would you use an AI Agent for this workflow, or keep it as a controlled pipeline?

GitHub link in the comments:  

document fetched 2026-09-24 00:20 · canonical https://reddit.com/r/AI_Agents/comments/1wnxog1/i_built_a_leadresearch_pipeline_and_deliberately