By Sekhar Ramarao
For decades, ERP has served as the enterprise's operational backbone, bringing structure, control, and a common source of truth to some of its most critical business processes. But its fundamental role has remained largely unchanged: ERP records transactions, enforces rules, and provides the system of record on which the business depends.
That role is now beginning to evolve.
The next generation of ERP will not simply record what has happened. It will increasingly understand business context, identify what needs attention, and help move work forward. With agentic capabilities entering platforms such as Microsoft Dynamics 365, the shift is from systems that primarily record and inform towards systems that can increasingly recommend, coordinate, and act.
This is what makes agentic ERP more than another addition to the enterprise technology vocabulary. It represents a meaningful change in how organizations interact with the systems at the heart of their operations.
From Assistance to Action
The distinction between a Copilot and an agent is important, but not because enterprises need another set of technology definitions.
What matters is how the work changes.
Copilots introduced a more intuitive way of working with enterprise applications. Employees could ask questions in natural language, summarize information, generate content, or receive recommendations without navigating through every screen themselves.
Agents take that model further. They can work towards a defined objective using business information, established rules and permitted actions. Instead of simply presenting information to an employee, an agent can take on parts of a process, respond to changing information and bring a person into the loop when judgment or approval is required.
A useful way to understand the progression is:
Copilot helps a person do the work. An agent can take responsibility for parts of the work. An autonomous workflow connects those actions across a broader business process.
Microsoft's current Dynamics 365 direction illustrates this progression. Its ERP portfolio includes agents for areas such as account reconciliation, supplier communication, and time and expense management, while the underlying platform is increasingly being opened to custom agent scenarios across finance and operations through Microsoft’s ERP Model Context Protocol (MCP) server. Several of the packaged agents remain in preview even as the MCP server itself has already reached general availability — a sign that the infrastructure for agentic ERP is arriving ahead of the fully packaged experiences built on top of it.
The distinction becomes much clearer when applied to an everyday enterprise process.
Take procurement.
A purchase request rarely moves directly from requirement to purchase order. There may be inventory to check, suppliers to identify, quotations to compare, approvals to secure, and exceptions to resolve. ERP manages the transaction, but people often spend considerable time coordinating what happens between those stages.
An agentic approach can begin to reduce that administrative burden. An agent might understand the requirement, work with approved vendor information, assist with quotation comparison, check relevant inventory context, prepare the next transaction and route it through the appropriate approval process.
Intertec's Supply Chain Copilot for Microsoft Dynamics 365 provides a practical example of this direction. It connects conversational requests with governed Dynamics 365 activities across purchase requirements, vendor identification, quotation comparison, and purchase-order creation, while retaining the security, business rules, workflows, approvals, and auditability established within Dynamics 365.
The interesting part is not the conversational interface. It is the connection between intent and a governed business process.
Enterprise Context Is What Makes an Agent Useful
A general-purpose AI model can understand a request to purchase 500 units of a component.
It does not inherently know which suppliers the organization has approved, whether some of those units are already available in inventory, which cost centre owns the budget, what approval threshold applies, or whether the person making the request has the authority to proceed.
The ERP does.
This is why enterprise context may prove more important to agentic ERP than raw AI capability.
Dynamics 365 sits within the flow of financial transactions, suppliers, customers, inventory, purchasing, projects, and business rules. Microsoft's ERP MCP capabilities now provide agents with standardized access to finance and operations data and business logic, while preserving the underlying permissions and auditability of the environment.
The value of an ERP agent therefore does not come simply from how intelligently it can generate a response. It comes from how reliably it can operate within the context of the business.
That also exposes a less glamorous side of the agentic ERP conversation.
Years of duplicate master data, outdated approval structures, unnecessary customizations, and fragmented integrations do not disappear when an agent is introduced. In some cases, automation simply allows those weaknesses to surface faster.
A supplier record that is wrong is still wrong when an agent retrieves it. A five-stage approval process that no longer serves a purpose does not become a good process because it has been automated.
Agentic ERP raises the value of process discipline rather than eliminating the need for it.
Autonomy Needs Boundaries
Finance, procurement, and supply chain are not environments where autonomy can simply be switched on.
Transactions carry financial consequences. Purchasing decisions have commercial implications. Changes to inventory or supplier commitments can affect operations elsewhere in the business.
The objective, therefore, should not be maximum autonomy.
It should be appropriate autonomy.
A straightforward reconciliation may be handled automatically while an exception is escalated to finance. A procurement agent may gather information and prepare a transaction, while a purchase above a defined threshold still requires human approval. A supply-chain agent may identify a potential delay and recommend an alternative without independently making a commercially significant decision.
The principle is familiar even if the technology is new: authority should follow risk.
Permissions, approval limits, segregation of duties and audit trails remain essential. If anything, they become more important when some actions are initiated by software rather than employees.
Microsoft's current agent-ready architecture reflects this requirement. Dynamics 365's ERP MCP framework is designed so that agents work with the data, business logic, and functions available within finance and operations while maintaining consistent permissions and auditability.
This is also why agentic ERP should not be treated as a standalone AI initiative. It is an enterprise applications and process transformation agenda involving data, integration, security, governance, and business-process design.
Preparing Dynamics 365 for What Comes Next
For organizations already running Dynamics 365, the next two to three years should not become a race to put an agent into every process.
A better starting point is to look for friction.
Where are employees repeatedly moving information between systems? Where are teams manually reconciling records that largely follow established rules? Which routine exceptions consume disproportionate time? Where are people re-entering information or coordinating predictable steps simply because the applications around them do not?
Those are stronger candidates for intelligent automation than processes where judgment, policy, or data quality remains unresolved.
Then comes the less exciting but more consequential work.
Are master records reliable? Are integrations stable? Do security roles still reflect the organization? Have customizations accumulated over the years? Are approval structures still necessary, or are they simply inherited from an earlier way of working?
These are not particularly futuristic questions. But they will determine whether the futuristic part actually works.
For Intertec, this evolution is a natural extension of the work its Microsoft Business Applications practice already does. Intertec's capabilities span Dynamics 365 implementation and configuration, custom development, integration, upgrades, and cloud migration, alongside Power Platform and related business application services.
That breadth becomes increasingly relevant as ERP moves beyond individual screens and transactions. An agent may start with Dynamics 365 data, interact with another enterprise application, trigger a workflow, and eventually return an action for approval. The quality of that experience depends on how well the underlying applications, data, integrations, and processes work together.
The opportunity is not to bolt intelligence onto everything that already exists.
It is to decide which work should remain human-led, which should be assisted, and which is mature enough to move towards greater autonomy.
A Different ERP Conversation for CIOs and CFOs
For CIOs, the shift changes the technology agenda. Data quality, integration, identity, security, extensibility and governance become part of the ERP conversation in a much more immediate way.
For CFOs, the impact is operational. If routine reconciliation, follow-ups, transaction preparation and predictable exception handling can increasingly be assisted or automated, finance teams can spend less time administering processes and more time on analysis, judgment and decisions that carry business value.
There is also a shared responsibility emerging between the two roles. Decisions about automation can no longer sit entirely with technology or entirely with the business. Someone must determine where autonomy creates value, where it introduces risk, and what level of control the organization is prepared to retain.
That leads to a more important question than how many agents an enterprise can deploy:
Which decisions are we prepared to let our systems make, and what must those systems know before we trust them to make those decisions?
Over the next few years, successful ERP programmes may not be judged by how many AI capabilities they activate. A better measure will be whether technology can take on more operational work without weakening the controls, accountability and business context that ERP was designed to protect.
For years, ERP has given enterprises a dependable system of record.
The next chapter is about what that foundation can help them do.
The system of record is not disappearing. It is learning how to act.
---------------------------------------------------------------------------------------------
Sekhar Ramarao leads the Microsoft Business Applications practice at Intertec Systems, with 28 years of experience across ERP, CRM, and enterprise AI in Government, Healthcare, Real Estate, and Construction, serving clients across India and the Middle East.
