Making Excel AI Friendly
MDN (Markdown Notation) project defining a simple, AI-readable spec for representing tabular and contextual data for AI systems
MDN + AI Experiments
Bridging Excel and Markdown for structured AI context
I was frustrated with getting excel data into AI in a token efficient manner that preserved the business logic of the document. This is my proposal to address that issue. MDN (Markdown Numbers) project defines a simple, AI-readable spec for representing tabular and contextual data. Alongside the spec, we're building tools to convert Excel ↔︎ MDN so teams can move between spreadsheets and AI-friendly formats without friction. This is all part of work I have been doing on the Bri AI document system.
The GitHub repo includes:
- Draft MDN specification
- Example Excel + MDN files
- Early-stage converter code
- Ongoing discussion and iteration
What is MDN?
MDN (Markdown Numbers) is an AI-optimized spreadsheet format that combines YAML metadata, CSV sheet data, and JSON blocks for formulas and formatting. It's designed to solve the challenge of making spreadsheets AI-friendly while preserving business logic.
Key Benefits
- AI-Optimized: Linearly structured for efficient token processing
- Business Logic Preserved: Keeps all formulas and calculations
- Human Readable: Uses plain text with YAML, CSV, and JSON
- Version Control Friendly: Text-based, works with Git and similar tools
- Token Efficient: Minimal overhead for AI context windows
- Bidirectional: Converts between Excel and MDN in both directions
Format Structure
The MDN format follows a consistent linear structure with clearly delimited sections:
- YAML Header: Document metadata and context
- CSV Sheets: Tabular data for each worksheet
- Formulas: JSON block with cell formulas and relationships
- Formatting: JSON block with styling and presentation rules