Skip to content

BRK216 Breakout Wed 18 Nov 13:45 PST

Multi-model AI in Copilot: Performance and value with built-in trust

Peter Bergen, Nagu Rangan, Natan Szymczyk

Summary

From the recording

Copilot offers Auto, which chooses the model, reasoning effort and harness per task, or explicit model selection. Microsoft says models ship on release day or very close to it. Everyday tasks sit within the user subscription licence, while advanced models use usage-based credits; tasks complete without charging above the admin-set monthly limit. Cost management does not require an Agent 365 licence. Use-case insights start this week in Frontier with Cowork, covering over 6,000 use cases. Cloud models are due in the EU data boundary by the end of this year. Most statuses lacked dates, and captions were auto-generated.

Session description

Microsoft's catalogue description

No single AI model is best for every task. As AI advances, organizations need access to the latest models without adding complexity for users, cost uncertainty for leaders, or administrative risk for administrators. Join us to see how Microsoft Copilot provides one governed front door to multi-model AI, with Microsoft evaluation and integration, intelligent routing with Auto, and cost management tools to you optimize AI value while maintaining security, compliance, and control.

Demo timestamps

Microsoft's player does not support links to a specific time, so open the recording and skip to the time shown.

  1. Skip to 0:13:07

    Auto mode adapting reasoning effort and generating artifacts

    Live demo showing Auto adjusting reasoning effort to task complexity, then a work-grounded example with the fictional company Caldova, where Copilot analyzes supply chain data and creates charts and a PowerPoint.

Watch on-demand (opens in a new tab)

← All sessions