Welcome to this AI Agent & Copilot Podcast, where we analyze the opportunities, impact, and outcomes that are possible with AI.
In this episode, I speak with Kenny Mullican about several use cases for Model Context Protocol (MCP) specific to his company’s Dynamics 365 ERP installation.
Highlights
MCP Overview and ERP Use Cases (1:50)
Mullican explains that MCP allows AI agents to communicate with other systems, acting as a bridge to ERP systems. MCP provides tools that AI agents understand, reducing the need for application programming interfaces, or APIs, to connect to those systems.
MCP can provide accurate and up-to-date information from ERP systems, avoiding hallucinations that are common in AI while avoiding the pitfalls of outdated or incorrect information. He notes the use of MCP in approval workflows and describes a project where over 1,000 customizations were documented and indexed for AI to access. The use of MCP in determining invoice discrepancies and pricing issues is highlighted, showing its effectiveness in complex scenarios.
Connecting AI to Company Data (11:55)
Kenny discusses the integration of AI with company data, using a data framework and Power BI. He explains the process of building an MCP to connect AI to semantic models and data warehouses, emphasizing the benefits of AI accessing real-time data for immediate responses. He also notes the availability of out-of-the-box MCPs for connecting to systems like Fabric.
MCP for Finance & Operations (14:10)
Paragon Films’ MCP for F&O allows AI to perform actions including creating sales orders, moving inventory, and triggering workflows. The potential risks and governance issues of AI performing actions in the ERP system are discussed. He highlights the importance of careful control and management of AI actions to prevent unintended consequences.
MCP Learnings and Best Practices (17:20)
Mullican shares insights on the differences in handling MCPs between Anthropic Claude and Microsoft Copilot. He discusses the importance of understanding the capabilities and limitations of different AI tools. He advises starting with easy MCP use cases, such as connecting to Microsoft Learn and company data, before moving to more complex integrations.
“Immortal Employee” MCP (21:40)
Beyond ERP, Paragon Films has what it calls an “immortal employee” MCP, which captures and documents the knowledge of experienced employees. He describes a project where an important employee is using AI to document his knowledge before retirement. The process involves creating documents, summarizing transcripts, and storing them in a SharePoint library. Then an agent can access them using MCP and build an interactive web page detailing the process(es) so the knowledge is available to others.
Mullican concludes by encouraging viewers and listeners to start experimenting with MCP and its various use cases.
More Kenny Mullican and MCP Insights:
- Microsoft MCP Server Fully Exposes Dynamics 365 to AI Without Middleware
- How AI Frees Leaders to Think Differently About System Building
- Governed Diversity Means Supporting a Portfolio of Specialized AI Tools
- How AI ‘Collapses the Feedback Loop’ in Software Development
Ask Cloud Wars AI Agent about this analysis
