As organizations deploy Copilots and AI agents across Microsoft 365, many discover the real risk isn’t the model—it’s their unstructured data. Email, shared drives, chat logs, and legacy repositories contain overshared, over-retained, poorly classified content. When this data is chunked, embedded, and indexed for AI, gaps in permissions, retention, and metadata become amplified.
This session introduces emerging practices at the intersection of AI, records management, and security. We’ll explore how Copilot inherits identity, ACLs, sensitivity labels, and retention policies; how to design ACL-aware RAG and hybrid vector/graph stores; and how to enrich chunks with metadata to enforce governance and compliant retrieval. Attendees will learn a practical, defensible approach to securing and governing unstructured data before—and after—it enters their AI pipelines.
- Learning Objective 1: Understand how Copilot and AI agents interact with existing identity, permissions, labels, and retention policies—and where organizations are still exposed when working with unstructured content.
- Learning Objective 2: Learn practical patterns for applying records-management and information-governance controls (classification, retention & disposition, defensible deletion) to AI training and RAG pipelines over unstructured data.
- Learning Objective 3: Discover architecture and metadata strategies (ACL-aware chunking, enriched security metadata, hybrid vector/graph filtering) that ensure AI retrieval only surfaces content users are allowed to see, while remaining auditable and compliant.
Terms: All rights reserved by the speaker, AI Agent & Copilot Summit NA, and Dynamic Communities and cannot be reproduced or distributed without express written consent by the speaker and Dynamic Communities.
Speakers
Ryan Britton, Vice President of Client Engagement, The IQ Business Group
Related Posts
