Built production-grade AI briefing system with multi-agent architecture

I originally built this system for a real operating company as a production-grade intelligence and distribution tool integrated into their daily workflow. This version is a sanitized demonstration build with proprietary branding, infrastructure, and internal business data Removed for portfolio and submission purposes!!

This project is a production-grade AI workflow that autonomously generates structured market briefings using a multi-tool AI agent architecture built in n8n. The system orchestrates real-time weather data, live news retrieval, structured LLM outputs, editorial review agents, and automated email formatting into a fully automated daily intelligence pipeline.

The architecture enforces schema validation, tool-calling discipline, structured output parsing, and multi-stage AI review to ensure reliability and consistency. The result is a self-operating briefing engine capable of transforming raw external data into structured, professionally formatted analytical reports without manual intervention.

Designed a multi-agent system (Primary Agent + Review Agent) for content generation and validation

• Implemented structured output enforcement using JSON schema parsing

• Built real-time tool-calling architecture (weather APIs + news APIs)

• Automated sentiment classification based on live external data

• Engineered HTML templating system for automated report generation

• Integrated automated email distribution pipeline

• Implemented data persistence layer for source tracking

• Created deterministic formatting structure for consistent output quality

• Developed modular architecture allowing easy expansion or tool replacement.

Tools

• n8n – Workflow orchestration & automation

• OpenAI (GPT-4.1-mini) – Language model & agent reasoning

• LangChain Nodes (n8n integration) – Agent + tool architecture

• News API – Real-time news retrieval

• OpenWeatherMap API – Real-time weather data ingestion

• Structured Output Parser (JSON Schema) – Deterministic output enforcement

• Gmail API – Automated email distribution

• Microsoft Excel Integration – Source tracking and persistence

• Custom Code Nodes (JavaScript) – HTML rendering & transformation logic

Methodology

The system follows a deterministic AI orchestration model:

1. Trigger Layer

Scheduled execution initiates the workflow.

2. Data Acquisition Layer

• Agent calls multiple weather tools for temperature retrieval.

• Agent calls News API restricted to recent articles.

3. Reasoning & Analysis Layer

• Primary AI Agent synthesizes live data.

• Sentiment classification determined from supply/demand dynamics and weather impact.

• Structured output enforced through schema validation.

4. Editorial Control Layer

• Secondary Review Agent validates sentiment alignment, formatting structure, and analytical coherence.

• Ensures no direct quotes, formatting errors, or structural issues.

5. Rendering Layer

• Custom JavaScript replaces dynamic placeholders in HTML template.

• Structured content injected into styled email layout.

6. Distribution & Persistence Layer

• Email delivered automatically.

• Source links logged to spreadsheet for archival reference.

Learnings

• Multi-agent systems significantly improve output reliability compared to single-pass LLM workflows.

• Structured output parsing is critical for production AI systems.

• Tool discipline improves reasoning consistency when strict ordering is enforced.

• Separating reasoning and rendering layers prevents formatting corruption.

• Schema enforcement reduces hallucinations and ensures deterministic automation.

• Modular workflow design simplifies debugging and scaling.

• Production AI requires guardrails, validation, and output review — not just prompt engineering.

Relevant URLs

(Replace with your demo-safe versions)

• n8n: https://n8n.io

• OpenAI API: https://platform.openai.com

• News API: https://newsapi.org

• OpenWeatherMap API: https://openweathermap.org/api

• Microsoft Graph API: https://learn.microsoft.com/en-us/graph/overview

• Gmail API: https://developers.google.com/gmail/api