Sunday Coffee & Code: Procurement Automation: RFP Agent Factory Progress Update

Last week, I shared the first working version of my offline PoC RFP Agent Factory - using Microsoft Agent Framework, Ollama, ChromaDB and Docling MCP (all the real focus of the work) to autonomously and automatically generate draft RFP Responses and Response Assessments. This weekend, a few significant improvements came together: ● Structured JSON extraction - instead of generating answers directly from the RFP, the Agent code now converts the RFP into structured JSON while preserving the exact hierarchy and wording. ● Iteratively processes each requirement by an Agent to generate targeted responses. ● Word output now split between Response for further processing and a Reasoning Audit record. ● Model upgrades and quality improvements: Moving to Qwen3:14b has resulted in better compliance with instructions and responses along with clearer hints when things went wrong (and overall, it seems clearer when reasoning) - while still running fully offline, automatically and autonomously. This approach delivers more consistent and complete content while maintaining a clean mapping between each requirement and its response. Importantly, this same structured dataset can power future automated RFP assessment workflows in addition to responses. It is encouraging to see how quickly this is progressing into something practical for proposal teams and procurement professionals alike. If you would like to compare approaches I would be happy to connect. ► Next steps: May be some RAG improvements and validation testing Agent to check for hallucinations.