AI Business Idea Hunter: Automated Research with Notion Memory & Anti-Duplicate Logic

[kevindo]

Hello n8n Community!

I’m excited to share my latest workflow: an AI Agent that hunts for high-quality online business ideas daily.

Finding viable business ideas often involves sifting through endless spam and duplicates. I built this workflow to solve that problem using n8n’s powerful automation and AI.

What does it do?

This agent runs daily to search, analyze, and validate business concepts from English sources. It doesn’t just dump data—it “remembers” what it has found to ensure you only get fresh, unique ideas.

Key Features:

• Multi-Source Search: Searches via Serper API with an automatic fallback to Tavily API if the primary source fails (High Availability).

• Stateful Memory (Anti-Duplicate): Connects to a Notion database to retrieve a 7-day history. It uses this context to filter out duplicates effectively.

• AI Quality Control: Uses Claude 3 Haiku (or GPT) to analyze ideas based on:

  • Revenue Potential

  • Setup Cost & Timeline

  • Automation Level

  • Risk Assessment

• Smart Filtering: Automatically rejects scams, MLM, and purely manual service jobs.

• Dual Output:

  • Telegram: Sends a concise summary of top ideas (Score 7/10+).

  • Notion: Saves detailed analysis to a database for tracking.

How it works (Architecture):

1. Trigger: Runs daily at 6 AM.

2. Search: Queries multiple keywords (SaaS, AI automation, etc.).

3. Branching: Checks if Serper returns results; if empty, switches to Tavily.

4. Memory: Fetches “Idea History” from Notion (last 7 days).

5. AI Analysis: Sends raw data + History context to LLM. LLM is prompted to return structured JSON.

6. Validation: Code node parses JSON and checks for errors.

7. Storage: Saves new ideas to Notion and updates the History log.

Setup Requirements:

To run this workflow, you will need:

1. API Keys: Serper API, Tavily API (optional backup), and OpenAI/Anthropic API.

2. Notion Template: Duplicate this template to your workspace:[ NOTION TEMPLATE LINK HERE]

3. Configuration:

   • Update the Notion Database IDs in the workflow nodes (Look for the red sticky notes in the canvas).

   • Connect your Telegram Bot.

---

Get the Workflow

Download JSON / View Code here: [GITHUB GIST LINK HERE]

(Click “Raw” on GitHub to copy the code easily)

Note: Please ensure you enable the LLM Notion EN node if it appears disabled after importing.

---

Demo Video:

See the workflow in action: [ YOUTUBE LINK HERE]

I built this to save time for solo founders and automation enthusiasts. Feedback and improvements are welcome!

[Andrew_Bell]

This is a well-thought-out workflow and a great example of combining AI with practical automation. I particularly like the use of Notion as a memory layer to prevent duplicate ideas and the fallback mechanism between Serper and Tavily for improved reliability. The structured AI evaluation, scam filtering, and dual output to Telegram and Notion make the solution both useful and production-friendly. Thanks for sharing such a detailed breakdown of the architecture and setup requirements—it’s valuable for anyone looking to build intelligent research pipelines with n8n. Looking forward to seeing how this evolves and what additional features the community contributes.

[kevindo]

Thanks Andrew_Bell, I’m glad you liked it.

I frequently use Notion as a memory storage location for ideas, and AI directly retrieves ideas from there. This ensures the AI doesn’t repeat duplicate ideas.

I look forward to receiving more feedback so I can have more ideas for further improvements.

Incidentally, you can also check out other workflows I’ve shared - this one was approved by the n8n admin. You can view it here: Vinh Automation Tools