Skip to content

ODSP2160 Pre-recorded On demand

Make agent-generated documents consistent and auditable

Ivan Tadic

Summary

From Microsoft's AI summary

Ivan Tadic of Doctavian described separating AI reasoning from document rendering. A model maps requests to structured parameters, and a deterministic engine produces the document, so identical data and templates give identical output. The architecture uses Claude for reasoning and an MCP server as the contract layer. In a demo, Claude retrieved templates, asked for missing information, generated an Acme quote, and created a draft signing envelope. The signed document was sealed with a qualified certificate and tamper-evident audit trail, with eIDAS-compliant signatures. Viewers were invited to register for a free developer account.

Session description

Microsoft's catalogue description

LLMs are great at drafting but not designed for repeatable output. When AI generates a contract or invoice, consistency matters. See how Doctavian uses MCP to interpret intent, render deterministic output, and add eIDAS-compliant digital signatures and audit trails. Learn a practical approach to building consistent, compliant agent-driven document workflows in Microsoft 365 and beyond.

← All sessions