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