Webcast Recap: Securing AI agents in Microsoft Dynamics 365 ERP

As AI agents touch more features of Microsoft Dynamics 365 ERP, security teams must face a new reality. AI is no longer just looking at data or making recommendations. It's doing real work on core business processes such as posting journals, updating records, and triggering activity inside the ERP itself.

That shift was one of the main themes of a recent 2025 year-end security webinar hosted by MSDW and featuring Alex Meyer and Frank Vukovits of Fastpath, now part of Delinea. The presenters discussed what organizations learned about security in 2025 and what they need to do differently as AI agents take on more responsibility inside ERP systems. The message was clear: organizations must secure AI agents in much the same way they secure human users within their systems.

Why AI agents are different from traditional automation

Meyer and Vukovits emphasized that AI agents differ from traditional automation. Earlier workflows and scripts typically stayed within a single application and performed a limited set of tasks. AI agents, by contrast, can access large volumes of data, interact with multiple systems, and take direct action much faster than human users.

Meyer said security teams can't rely on old assumptions any longer. That's because AI agents aren't there just to help or answer questions. Now, they can act for users and move between systems, which becomes risky if their access isn't kept in check.

Understanding that is especially important in an ERP environment, such as Microsoft Dynamics 365 Finance & Operations, where actions often have financial, regulatory, and audit consequences.

AI agents acting inside D365 Finance & Operations

AI agents differ from traditional automation in their ability to take action within and across business systems. That difference is critical when thinking about security and auditability in D365 F&O. Unlike a sidecar-style copilot, agents can perform many of the same tasks as human users.