Send us Fan Mail

Episode Summary

Microsoft’s new Copilot experience brings Chat and Co-Work together under Home while adding Code and Autopilot as distinct modes. This episode explains how each mode is intended to be used, how the managed runtime could simplify application governance, and why organizations need to prepare for consumption-based costs. It also covers Microsoft IQ, plugins, model access, advanced SharePoint capabilities, OneDrive, and Copilot Search.

Topics Covered

• Copilot Home, Code, and Autopilot
• Chat versus Co-Work
• Office editing inside Copilot
• The Today screen
• Mentioning Copilot in Teams
• Personal Autopilot agents
• End-user coding
• Copilot managed runtime
• Microsoft IQ and Dataverse data
• Plugin and MCP governance
• FinOps for AI and Copilot credits
• Model-family access controls
• Hybrid billing for SharePoint
• Copilot Search in SharePoint and OneDrive

Key Takeaways

• Communicate the redesigned Copilot interface before users encounter it.
• Match quick work to Chat and complex multi-step work to Co-Work.
• Use Autopilot for autonomous personal work without abandoning enterprise-agent investments.
• Establish spending policies and approval paths before expanding usage.
• Decide which business processes are suitable for Code, Autopilot, Co-Work, or enterprise agents.
• Treat advanced SharePoint capabilities as a distinct cost and governance category.

Chapters

00:00:00 Welcome to the Copilot Briefing
00:00:32 Microsoft’s Copilot announcements
00:00:58 The new Copilot app
00:02:31 Office inside Copilot
00:04:13 The Today screen
00:04:59 Mentioning Copilot in Teams
00:05:55 Introducing Autopilot
00:06:46 Co-Work versus Autopilot
00:08:03 Autopilot triggers and actions
00:09:47 Home, Autopilot, and Code
00:10:06 End-user application development
00:11:38 Copilot managed runtime
00:13:18 Microsoft IQ and business data
00:14:46 The plugin registry
00:15:47 FinOps for AI
00:19:06 Agent costs and spending policies
00:21:01 Controlling model access
00:22:06 Cost visibility and usage limits
00:24:38 Copilot in SharePoint
00:28:06 OneDrive and Copilot Search
00:29:00 Business priorities and takeaways

Links

Copilot in SharePoint updates - https://techcommunity.microsoft.com/blog/spblog/sharepoint-ai-innovations-hit-ga-powering-new-copilot-app-and-agents/4555724?WT.mc_id=MVP_320501

The new Copilot app - https://blogs.microsoft.com/blog/2026/09/25/introducing-the-new-copilot-with-home-code-and-autopilot/

Send us Fan Mail

Episode Description

Running AI at scale is expensive, and consumption-based services can make budgets harder to predict. Steve Corey explains Microsoft’s MAI family of in-house models, the role of multi-model orchestration, and why routine work should not automatically be sent to the most expensive model. The episode reviews MAI models for reasoning, coding, image generation, voice, and transcription, then considers how efficient routing could affect Copilot credits, pay-as-you-go usage, custom agents, and AI return on investment.

✅ Connect With Steve

☑️ Socials - https://stevecorey.social

🌍 Web - https://stevecorey.com

📝 Show Notes and Resources 📝

Topics Covered

• The cost of operating AI at scale

• Copilot licensing and consumption-based pricing

• Copilot credits and variable AI costs

• Microsoft’s MAI model family

• Multi-model orchestration and automatic routing

• Reasoning, coding, image, voice, and transcription models

• Pay-as-you-go cost management

• Cost-conscious custom agent design

• Matching model capability to task complexity

• AI solution feasibility and return on investment

Key Takeaways

• Not every request requires a frontier model.

• Efficient routing can reserve expensive models for genuinely complex work.

• Model choice directly affects the consumption profile of AI solutions.

• Custom agents can use different models for routine and critical steps.

• Organizations should continually reassess which models their workloads use.

• Lower-cost models may make usage more manageable, but consumption remains variable.

Chapters

00:00 Welcome to the Copilot Briefing

00:32 Why AI Costs Matter

01:37 The Copilot Pricing Challenge

03:14 Introducing Microsoft MAI

03:49 Why Microsoft Built Its Own Models

05:19 Multi-Model Orchestration

06:12 How Model Routing Works

06:56 MAI Model Lineup

07:01 Reasoning Model

07:53 Coding Model

08:23 Image Model

09:25 Voice Model

10:11 Transcription Model

10:49 Implications for Copilot Costs

12:18 Pay-As-You-Go Economics

13:20 Custom AI Cost Control

13:52 The Bottom Line

15:20 Podcast Takeaways

🔗 Links

MAI Models - (Microsoft Techcommunity)

Send us Fan Mail

Episode Summary

Microsoft Copilot Cowork tasks can produce very different costs even when they initially appear similar. This episode explains the major factors behind that variation, including model choice, context retrieval, tool usage, reasoning depth, runtime, and generated outputs. It also presents a practical framework for distinguishing light, medium, and heavy tasks and describes how the /cost command can be used to review the credits consumed by a completed task. The central recommendation is to compare Cowork’s cost with the time and labor required to perform the same process manually.

✅ Connect With Steve

☑️ Socials - https://stevecorey.social

🌍 Web - https://stevecorey.com

Topics Covered

• Copilot credits and usage-based billing
• Model selection and AI workload complexity
• Context retrieval from files, email, chats, and transcripts
• Connectors, MCP servers, agents, and plugins
• Runtime and long-running Cowork processes
• Light, medium, and heavy task categories
• Additional files and presentation outputs
• Reviewing task usage with /cost
• Comparing AI costs with manual labor
• Choosing between Copilot, agents, Researcher, and Cowork

Key Takeaways

• Cowork costs reflect the full workload, including context, tools, reasoning, runtime, and outputs.
• Broad organizational searches generally require more resources than single-document tasks.
• The /cost command provides a concrete way to inspect the cost of a completed run.
• Not every task requires Cowork; some may be better suited to regular Copilot or an agent.
• Business value should be evaluated by comparing AI usage with the cost of the existing manual process.

Chapters

01:00:00 Introduction
01:00:32 Why Cowork task costs differ
01:00:52 Usage-based billing and Copilot credits
01:02:07 Different AI workloads require different resources
01:02:44 Document summaries versus deep analysis
01:03:15 Models, context retrieval, and tool calls
01:05:10 How runtime affects cost
01:06:08 Cowork as a long-running agent harness
01:07:16 How additional outputs increase cost
01:07:37 Light, medium, and heavy tasks
01:07:45 Characteristics of a light task
01:08:43 Characteristics of a medium task
01:09:12 Characteristics of a heavy task
01:10:20 Checking task cost with /cost
01:11:20 Comparing Cowork with manual labor
01:12:00 Choosing between Cowork, agents, and Researcher
01:13:07 The podcast takeaway
01:13:56 Measuring process time and business cost
01:14:16 Business outcomes and closing