Session Abstract

For years, governance in Microsoft 365 focused on content: sites, documents, permissions, and sprawl. With Copilot, agents, and skills, we're increasingly governing behaviour at scale. As organisations move from AI-assisted work to AI-driven action, new questions emerge: Who owns the decisions agents embody? How do we manage competing instructions, rules, and knowledge sources? And what happens when AI systems work exactly as designed, but produce outcomes nobody expected? In this session, we'll explore: • Why agent sprawl may become the next governance challenge in Microsoft 365 • How skills and instructions encode organisational judgement and decision-making • What organisations must do to build trust, accountability, and oversight into AI-driven work As AI starts acting, governance must evolve from managing information to managing outcomes. The organisations that succeed won't be those with the most agents, but those with the clearest understanding of who owns the decisions those agents make.

Rebecka Isaksson

Rebecka Isaksson is the Founder of KnowFlow Value and a Content AI & M365 Knowledge Strategist, helping organizations turn Content Chaos into Copilot Ready. A Microsoft MVP and former Director of Knowledge Management at Microsoft HQ, she brings 15+ years of expertise in collaboration, content governance, and AI-powered knowledge strategies. She’s the creator and host of the Knowledge Fika (now 2.0) podcast, where Swedish fika meets smart conversations on AI and knowledge in M365 - featuring Microsoft Execs, Experts and MVPs. Beyond the stage, Rebecka mentors and coaches emerging KM professionals and AI transformation leaders, fuelling the next wave of talent in this space. A sought-after keynote speaker and panellist at global events, Rebecka delivers practical insights on how AI accelerates knowledge strategies - and vice versa - and why culture change is the real game-changer. Recognized as a Microsoft MVP in 2024 (SP) and since 2026 also for Copilot, she continues to champion ethical, human-centric approaches to AI adoption.