microsoft / microsoft/dynamics365patternspractices

[PATTERN]: GMS Mailbox Agent

Open
#336 0 comments 0 reactions 1 assignee View on GitHub

@rachel-profitt is already working on this.

Since May 5, 2026.

needs review pattern
Dominant language
Python
Stars
227
Forks
106
Avg merge
2d 13h
Merged PRs (30d)
1

Description

Contact details

saphalyamohanty@kpmg.com

Organization type

Partner / ISV / Independant consultant

End-to-end business process

Case to resolution

Which business process area is this article related to?

Customer Service

Which business process is this article related to?

Work and Dynamically Prioritize cases

Which pattern or practice is this request related to?

Manage Cases

Enter any additional comments or information you want us to know.

Business Goals and Drivers
The GMS Mailbox Agent is an agent first, D365 native capability designed to help Global Mobility Services teams manage high volume engagement mailboxes by improving email classification, SLA aware prioritization, routing recommendations, and response drafting—while ensuring mandatory human review before any client communication is sent. The solution uses Dynamics 365 Customer Service as the system of record for case creation and lifecycle management, with Microsoft Dataverse providing a secure and auditable data layer for engagement scoped email content, AI outputs, and operational logs. Microsoft Copilot Studio acts as the reasoning layer (classification/prioritization/drafting), and Power Automate provides reliable event driven orchestration and guaranteed processing—ensuring the agent augments operations without replacing governance or control
Operationally, the end to end process begins when an email is received and ingested into D365, creating or updating a case. A Power Automate cloud flow triggers on case creation to persist email metadata/content to Dataverse, invoke Copilot Studio agent actions for classification and prioritization, then run SLA evaluation logic (commonly via child flows) to compute urgency and SLA risk. When a response is needed, a second agent action generates a draft reply using engagement scoped historical context (prior mailbox threads, similar cases, approved templates/rules), and the draft plus evidence references are stored back in Dataverse. Users review and approve the draft inside the D365 model driven experience; only after explicit approval is the reply sent—ensuring the agent never sends client emails autonomously.
Security and compliance are enforced through defense in depth: users authenticate to D365 and are authorized via role based access control (RBAC), while Dataverse row level security enforces engagement boundaries (e.g., EngagementID scoping) for all data access. Copilot Studio does not bypass these controls; instead, the agent accesses data via scoped actions that retrieve only engagement permitted records and write back results under the same security constraints. This architecture provides strong auditability (classification decisions, confidence scores, overrides, SLA updates, draft approvals) and supports regulated operating requirements by ensuring the AI layer is governed, bounded, and human validated.

Solution Architecture GMS_MailboxAgent v0.1.docx

Specify the date you expect the article to be completed and ready for review.

05/05/2026

Code of Conduct
  • I agree to follow this project's Code of Conduct

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.