Autonomous AI agents are moving from demo to daily reality. But who governs them, and who pays for them?

In this episode of Guardians of M365 Governance, Christian Buckley, Joy Apple and I (Ragnar Heil) compare what we run every day with what Microsoft is shipping. We cover shadow AI agents, Microsoft Agent 365, the new OneDrive Copilot experience and the question every customer asks first: what will it cost?

In this episode
🤖 Building in public: how we run 20–30 personal bots for daily briefings, research and video ideas
📊 The new OneDrive Copilot experience: interactive dashboards, document-to-presentation and charts
🛡️ Agent 365 live: registering and governing non-Microsoft agents (Anthropic, Amazon Bedrock, Databricks, Google Vertex AI, Salesforce, ServiceNow, Snowflake, UiPath)
👀 Shadow AI: why agents like OpenClaw, Hermes and Grok bots stay invisible on unmanaged devices
💰 Copilot cost governance: my vibe-coded dashboard on the new Microsoft 365 Copilot cost API, and why one prompt burned about 9,000 credits
⚖️ The big question: tokens are visible, but wasted hours are not

Key takeaways

Every agent and every prompt needs a clear objective
Visibility comes first: use Agent 365 to see which agents exist
Plan for the blind spot, because agents on unmanaged devices won’t appear
Give end users a self-service view of their credit consumption
Measure cost against value, not just against tokens

Hosts
- Christian Buckley – 16-time Microsoft MVP and Microsoft Regional Director, Dallas
- Joy Apple – Microsoft MVP, Director of Success at Orchestry, Tulsa
- Ragnar Heil – 10-time Microsoft MVP, Senior Business Development Manager at HanseVision (Bechtle Group), Germany

Who Governs Your autonomous AI Agents? Agent 365 vs Shadow AI

Ragnar365 Nuggets: Microsoft 365 AI-powered Workplace

Who Governs Your autonomous AI Agents? Agent 365 vs Shadow AI