Automated Market Intelligence Pipeline

A public automation system for collecting, ranking and briefing market signals. Deterministic orchestration comes first; language-model reasoning is used where it adds a structured brief.

Content monitoring becomes brittle when every stage depends on a model. Collection, normalization, deduplication and ranking should continue to work when a model is unavailable or returns invalid output.

A working automation project, not a marketed SaaS. It has no published customer metrics.

I designed and implemented the workflow around deterministic collection and ranking, with language-model reasoning used for structured briefs and explicit fallbacks.

The workflow gathers signals from RSS and search sources, normalizes and deduplicates them, ranks relevance, optionally creates structured briefs with Claude and a Gemini fallback, and records runs before email delivery.

• n8n orchestration
• RSS and search-source ingestion
• Normalization, deduplication, scoring and ranking
• Claude reasoning with Gemini fallback
• Deterministic non-model fallback
• Structured JSON output
• Scheduled execution, webhooks and run history
• Email delivery

• n8n
• RSS and search sources
• Claude
• Gemini
• n8n Data Tables
• Resend

• Language-model reasoning for structured briefs
• Gemini fallback model
• Deterministic fallback when no model result is available

• Keep collection and ranking deterministic so the pipeline can operate without a model.
• Use models for briefing rather than workflow control.
• Make fallback behavior inspectable and explicit.

• Signal collection, normalization and deduplication
• Scoring and ranking
• Structured model-generated briefs with fallback behavior
• Scheduled execution, webhooks and run history
• Email delivery

None listed.

Working automation

A working automation that produces structured briefs from ranked market signals with inspectable fallback behavior.