Working on an interesting project right now and figured I'd share what I'm learning: helping a client automate their engineering document workflow using Claude Desktop + SharePoint/365 connectors.
The challenge: They create ~600 "data packages" per year. Each one requires pulling structured data from their ERP, plus unstructured content (text, tables, images) from certification PDFs. Currently takes 30-60 min each, mostly manual copy-paste.
The approach: Claude Desktop with MCP connectors to SharePoint + a SQL connector to their ERP. Goal is 50% time reduction.
Early findings:
- The 365 connector can pull text from PDFs but not images (yet)
- Might need Claude Code for the image extraction step
- Data normalization is the real bottleneck — some fields live in the ERP, some only in PDFs
Anyone else working on similar enterprise doc automation? Curious what approaches have worked for you.