D365 in trading: AI-assisted process discovery from a project leader’s perspective

11 October 2026 · 7 min read · Manfred Schönberger

During an assignment in the trading sector, one of my responsibilities was to lead the process analysis for a D365 implementation. The brief covered capturing all current processes with AI support and conducting the subsequent fit-gap analysis. The central question was which workflows the ERP standard can support and where the business needs a different solution. The following sections explain the requirements for delivering that assignment.

Turn the brief into a manageable scope

“Capture all current processes” was the starting point of the brief. Project control required a structure that made the scope and ownership of each process visible. A process map was intended to provide the framework for discovery and the subsequent comparison with D365.

In trading, connections cross departmental boundaries. A customer order can involve sales, warehousing and finance. Purchasing and goods receipt affect stock and supplier invoices. Discovery therefore needed to follow business transactions from trigger to outcome and capture handovers between teams.

Defining the overall scope required consideration of supporting workflows. Examples include master data maintenance, returns, stocktaking and financial close. Agreement with the owners was intended to establish complete boundaries, covering frequently discussed processes alongside less visible activities.

Use a common language for current processes

Discovery was intended to explain how a transaction was handled at the time of analysis. Each process required consistent information: its trigger and outcome, the roles and systems involved, and the rules governing decisions. Manual intermediate steps and exceptions also needed to be considered.

Work instructions were intended to provide a starting point. Discussions with employees were meant to establish whether actual practice matched the documented procedure. A handover through email, an additional spreadsheet or a personal query needs to remain visible. Responsibilities and data flows can become unclear at these points.

For me as project leader, separating discovery from evaluation was essential. The existing workflow needed to be documented first. This would provide the basis for assessing future working practices. An existing step earns its place in the target process through its purpose and the business requirements it serves.

Use AI as a discovery tool

The brief called for AI support in current-process discovery. Interview notes and approved work instructions were intended to be brought together in a consistent process description. AI should identify activities and roles, compare conflicting statements and flag missing information as questions.

The requirements for an AI instruction were clear: it needed to use the supplied material and leave gaps visible. A suitable formulation is: “Describe the current process using only the supplied material. Reference the originating passage for each activity. Separate confirmed information from assumptions and unanswered questions.” The resulting draft was intended to prepare professional review.

Professional confirmation needed to remain with the people who understood and owned the workflow. Completeness and the sequence of activities required validation. My responsibility as project leader was to lead discovery and the subsequent analysis across the full process scope. Progress and ownership of unresolved questions therefore needed to be traceable.

Move from current processes to the D365 standard

Fit-gap analysis required clarification of the business requirement behind the existing workflow. An additional approval may provide an essential control. It may also have arisen from a limitation in the previous system. The distinction affects the future solution.

The comparison was intended to start with D365 standard processes and configuration options. AI can identify candidate documentation and prepare review questions. A proposed match still needs to be verified in the intended application and configuration.

A sound fit assessment required concrete evidence covering the data, roles and interfaces involved. The proposed process needed to be demonstrable using a suitable transaction. The D365 application and features used must be recorded in the project scope; a general process name alone does not establish the product fit.

Record gaps so that decisions can be made

Fit-gap analysis was intended to provide a basis for decisions. Each finding required a documented requirement, evidence and a next review step. The following overview summarises the information needed.

A pending assessment is useful when the missing information is clear. A premature classification creates certainty that the project may later have to correct. A transparent status was therefore as important to project control as the functional assessment.

Connect discovery with acceptance

The analysis was also intended to support later system validation. Approved requirements can support test cases, with AI structuring starting conditions, actions and expected outcomes. Professional review needed to consider relevant exceptions alongside the normal workflow.

For invoice approval, absence, rejection and insufficient permissions would be possible cases to investigate. This example explains a potential testing approach; it is not a confirmed finding from the assignment. The actual cases follow from discovery and the agreed rules.

The intended outcome was a traceable connection between current processes, future workflows and later system validation. Changes to a requirement needed to be traceable to the relevant decision and test. This documentation requirement belonged in the initial structure of the working material.

The assignment therefore required process knowledge and system expertise to be connected. The intended outcome was a shared basis for D365 implementation decisions: which workflows to adopt, where changes are needed and which questions must be resolved before implementation.