What Happens When an AI Agent Goes Rogue in your Microsoft Tenant with Chance Weaver

The speed of AI innovation means that proofs of concepts that once took years now reach production in weeks. Nathan Taylor and Chance Weaver analyze how workflows went from chat interfaces to sophisticated agentic workflows within the Microsoft ecosystem. They even discuss Sourcepass MCOE’s experience helping partners move from theoretical interest to functional deployment using tools like Copilot Studio and the newly announced Agent365.

As you know, identifying business problems is the first step toward effective automation. Chance shares his experience building a "Chief of Staff" agent to manage global roadblocks and explains why teaching internal teams to build their own agents is the most effective training method. They also address the "token economy," where efficiency in model selection becomes a financial necessity rather than just a technical preference.

What You’ll Learn:

The structural advantages of Agent 365 for managing agent identity and auditing
Differences between responsive Copilot, automated Cowork, and autonomous Scout
Strategies for achieving token efficiency in a consumption-based pricing model
Why local AI processing on Copilot PCs might change your cost structure
The importance of data governance to prevent IP leakage into public models

About the Guest:

Chance Weaver is the Global VP of AI Adoption for Pax8, where he helps partners implement and monetize AI technology through the Pax8 marketplace. He brings nearly three decades of experience in technology and business development to the role. Before joining Pax8, he founded and scaled a highly successful, Microsoft focused MSP in Southern California that he later sold to a large platform company, where he went on to lead its Microsoft practice and AI strategy across dozens of acquisitions. Today he leads global initiatives that help the partner community have deeper business conversations and pinpoint where the real opportunity to implement AI lies.

About the Host:

Nathan Taylor is the Senior Vice President and Global Microsoft Practice Leader at Sourcepass, where he leads the Sourcepass Center of Excellence for Microsoft. His work is grounded in a simple idea: Microsoft should not be complicated. By removing complexity, confusion, and frustration from the Microsoft ecosystem, Nathan helps organizations focus on outcomes while getting the most from their Microsoft investment.

Episode Highlights:

[00:12:12] The Chief of Staff Agent Concept

Chance explains his creation of an autonomous agent that monitors global projects and surfaces roadblocks daily. This tool acts as a proactive assistant, identifying potential issues before they disrupt operations for the leadership team.

[00:15:15] Security and IP Protection in AI

Uploading sensitive business data to public models like Anthropic without enterprise licensing can result in a loss of intellectual property. Nathan and Chance stress why working with a partner like Sourcepass MCOE is vital for maintaining data sovereignty.

[00:20:32] Solving the Identity Crisis with Agent365

Managing agents under a user's personal identity creates massive security and auditing risks. Agent 365 introduces a dedicated agent identity, allowing for full audit trails and professional lifecycle management of automated tasks.

[00:22:51] Microsoft’s AI Licensing Realities

Current pricing models for tools like Agent 365 often feel out of reach for the average SMB. Microsoft is currently iterating on these structures, and businesses should expect significant changes as the licensing settles over the next six months.

[00:25:27] Preparing for the Token Economy

Consumption-based pricing means that token efficiency is now an art that IT leaders must master. Understanding which models to use will determine the long-term profitability and sustainability of AI projects.

[00:26:40] Local Processing and Copilot+ PCs

While most current Microsoft AI tools rely on cloud processing, the arrival of NPU-equipped hardware offers a path to lower costs. Running smaller models locally could eventually offset the token consumption costs associated with cloud-only workflows.

Episode Resources:

Reach out to Sourcepass MCOE to see how your organization can build its own AI agents: https://sourcepassmcoe.com/demystifying-microsoft-contact

Chance Weaver on LinkedIn

Nathan Taylor on LinkedIn

Quotes:

"Agent365 being able to bring agent identity—having an identity for an agent, being able to track that identity, have full audit capability—is incredibly powerful. It solves a lot of those what do we do from a security standpoint from an agent? to manage agents for clients in a way that they would have difficulty doing before."
"The token economy is coming. And you know, you gotta make sure as an SMB that you're working with a partner that understands which models you should use to basically—I call it token efficiency—to make sure that you have token efficiency when you're building out agents or when you're building out automations."
"If you go to an SMB and you say, 'Hey, what are the problems in your business?' a lot of times they're like, 'I don't know.' They are just heads down trying to get stuff done on a day-to-day basis. So being able to have the conversations to help identify that or teach them a little bit about AI, I think this is incredibly powerful."