Session

Ollama on the Finance Laptop: Hunting the Shadow AI That Lives on the Endpoint

Most shadow AI advice stops at the browser: block the chatbot, add DLP, done. Meanwhile the AI has moved onto the device. Local models run on NPU laptops through Ollama and LM Studio and never touch a SaaS catalogue. MCP servers sit in config files on admin and developer machines, holding tokens and running commands. AI features arrive switched on through ordinary app updates. In Microsoft's own UK research, 71% of employees said they had used unapproved AI tools at work.

Microsoft's new local-AI discovery in Defender, Intune and the Microsoft 365 admin centre is a start. This session tests how far it gets. Ryan Mangan, a double Microsoft MVP for Windows 365 and AVD, walks through a layered detection model (inventory, process and network telemetry, config files, identity consent and the browser). He shows what each layer catches and misses on real devices, and fills the gaps with free scripts, KQL and Intune remediations. You'll finish with a paved road rather than a ban, including Windows 365 and AVD as a governed place to run local AI.

You'll leave with:
- A map of where AI hides on Windows devices, and which tool sees each layer
- KQL queries and Intune remediation scripts, shared on GitHub
- A paved-road policy that offers sanctioned local AI instead of blocking everything


Level 300. For endpoint, Intune and security admins. Live demos. Vendor-neutral: Microsoft-native tooling and free scripts only.

Ryan Mangan

Founder & CEO, EfficientEther | Double Microsoft MVP, FBCS CITP | Author, Mastering Azure Virtual Desktop

York, United Kingdom

Actions

Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.

Jump to top