Microsoft Dynamics 365 Agentic ERP Roadmap Runs Through March 2027
Microsoft is expanding the role of AI agents across its Dynamics 365 ERP portfolio, with new capabilities available now and further developments planned through March 2027.
The latest roadmap covers seven Dynamics 365 ERP applications and signals a broader shift towards agentic ERP, where AI can interact with enterprise data and support — and increasingly execute — business processes.
For ERP teams, one of the most significant developments is not tied to a particular module. Microsoft has made two connectors generally available that provide supported ways for AI agents to access Dynamics 365 ERP records and actions.
The developments also come as Microsoft moves away from its traditional twice-yearly Dynamics 365 release waves towards a continuous roadmap.
How AI Agents Connect to Dynamics 365 ERP
Microsoft’s approach keeps Dynamics 365 at the centre of the enterprise data architecture.
The ERP remains the system of record, providing business data, security and process logic, while AI agents operating through Microsoft 365 can interact with that information through Microsoft’s supported connection mechanisms.
Two connectors are now generally available:
- The Dynamics 365 ERP plugin for Copilot Cowork
- The Dynamics 365 ERP Model Context Protocol (MCP) server
Together, these provide agents with a supported route to retrieve information from Dynamics 365 and perform actions against ERP records.
AI-generated responses can also reference and link back to the underlying ERP records, giving users greater visibility into the information behind an agent’s response.
This is an important distinction from AI tools operating purely as external assistants. The agent can work with the same business information and controls that underpin the ERP environment.
Business Skills Bring Processes Into the Agent Layer
Microsoft is also introducing what it calls business skills.
These are reusable instructions that organisations can provide to ERP agents, effectively allowing companies to define how an agent should approach specific business processes.
Access to these capabilities is managed through existing Dynamics 365 security roles.
Microsoft plans to provide pre-built skills for processes such as procure-to-pay, while organisations will also be able to develop their own instructions.
Another planned capability will allow administrators to establish restrictions specifically for actions performed through Copilot Cowork or an AI agent.
This creates a distinction between what an employee can do directly within Dynamics 365 and what an agent can do on that employee’s behalf.
For sensitive activities such as financial postings or payments, this could give organisations greater control over the level of autonomy granted to AI systems.
The connectors are already generally available, but some of these more detailed agent controls remain on Microsoft’s roadmap.
Where Dynamics 365 Agents Are Appearing
Many of Microsoft’s process-specific AI capabilities remain in public preview or are still planned for future releases.
In Dynamics 365 Supply Chain Management, the procurement agent impact analysis capability can assess the potential consequences of a proposed supplier purchase order change. It can trace the change through related sales, production and transfer orders and highlight potential effects on projected inventory.
In Dynamics 365 Field Service, the Scheduling Operations Agent is designed to optimise schedules across multiple technicians. Proposed changes are presented to dispatchers for review before they are implemented.
Microsoft is also expanding its Finance Agent capabilities, including reconciliation and collections functionality across applications such as Excel, Outlook and Teams.
The Account Reconciliation Agent is designed to help resolve exceptions in batches, with controls and notifications intended to keep users informed.
Within Dynamics 365 Human Resources, Microsoft is introducing agent-assisted capabilities for areas such as candidate assessment and employee onboarding.
Across these examples, a common pattern emerges: AI analyses information or proposes an action, while a human remains responsible for approving the change.
That approach is particularly important when AI is operating in processes that can affect financial records, inventory, customer commitments or operational schedules.
Traditional ERP Capabilities Are Also Advancing
The Dynamics 365 roadmap is not limited to AI agents.
Microsoft is continuing to introduce conventional ERP functionality alongside its agentic capabilities.
Developments include support for French e-invoicing and e-reporting requirements, generative insights for demand planning and work orders that connect directly with Project Operations tasks.
This illustrates the broader direction of Dynamics 365: AI is being added to existing ERP processes rather than replacing the core transactional capabilities of the platform.
The ERP remains the underlying system of record, while AI increasingly becomes an additional layer through which users interact with that information.
The Move to a Continuous Dynamics 365 Roadmap
The latest announcements also represent a change in how ERP teams will need to plan for Dynamics 365.
Microsoft has moved away from its previous twice-yearly release-wave model towards a continuous roadmap.
Previously, release plans provided ERP teams with a structured view of capabilities expected within a particular wave. Under the new approach, individual features can move through development, preview and general availability on an ongoing basis.
This makes roadmap monitoring a more continuous responsibility for ERP teams.
With the previous Release Planner being retired, organisations will need to monitor Microsoft’s AI at Work roadmap and other product updates regularly to understand when preview capabilities become generally available.
What This Means for ERP Teams
Define agent permissions before connecting AI to ERP. The connectors are generally available, but organisations should establish clear boundaries around what agents can access and which actions they can perform. This allows teams to experiment without unintentionally expanding permissions around sensitive processes.
Treat business skills as process documentation. Reusable instructions for AI agents can also become a practical way of documenting how important processes should operate. Organisations with clear and well-maintained process definitions will be better positioned to develop reliable agent skills.
Keep humans in high-impact workflows. Many of Microsoft’s current agent capabilities follow a review-and-approve model. ERP teams should identify which activities require human approval, particularly those involving payments, financial postings, inventory commitments or customer obligations.
Monitor the roadmap continuously. With Dynamics 365 moving towards an always-on roadmap, organisations should replace periodic release planning with regular reviews. Monthly or quarterly assessments can help teams identify when useful AI capabilities move from preview to production.
Separate experimentation from production control. The availability of ERP connectors makes it easier to begin testing agents. However, pilots should operate within clearly defined security, governance and approval frameworks before AI is allowed to execute consequential transactions.
The Bigger Picture for Agentic ERP
Microsoft’s roadmap shows how the role of AI in ERP is evolving.
The next stage is not simply about adding chatbots to enterprise applications. AI agents are increasingly being given structured access to ERP data, business rules and operational workflows.
That creates the potential for agents to move from answering questions to analysing the impact of proposed changes, recommending actions and eventually carrying out selected tasks.
For ERP leaders, the challenge will be finding the right balance between automation and control.
Microsoft’s current approach, combining ERP security, agent-specific permissions and human review, points towards a gradual path to agentic ERP rather than immediate full autonomy.
As more capabilities become generally available through March 2027, organisations will have a clearer opportunity to determine where AI can safely take work off employees’ hands while keeping critical business decisions under human control.