D365 Customer Service: Governance Tips

Key insights

• Governance overview: Microsoft added a preview policy-enforcement layer in Dynamics 365 Customer Service that checks outbound emails before they reach customers.
  It evaluates both representative-authored and AI-generated messages and can flag or block content that breaks rules.
• Email channel and attachments: The feature focuses on outgoing email and also reviews text-based attachments like PDF and Word files.
  Administrators can include attachments in the same pre-send checks that apply to message text.
• Enforcement and strictness levels: Administrators choose actions such as flagging or blocking and set strictness to Low, Medium, High, or Very High.
  These settings let teams tune enforcement to match compliance, security, and brand needs.
• Copilot Service admin center setup: Turn on governance from the Copilot Service admin center and manage policies in the Copilot Service workspace.
  Enabling the governance toggle activates the preview layer for your environment.
• Prerequisites: Admins need the CSR Manager role and agents need the Customer Service Representative role, plus privileges to governance-related Dataverse tables.
  You also need Copilot credits, consent for possible data movement, and AI agents enabled in the Power Platform admin center.
• Policy creation and simulation: Create or edit policies to set name, content type (Email), author type (representative or AI), enforcement action, and conditions, then save and publish.
  Use the simulation option to test policies on sample messages before enforcing them in production.

Video overview

Dian Taylor - [MVP] (Dynamics 365 Talk) published a practical YouTube walkthrough that explains how to configure governance in Dynamics 365 Customer Service. In the video, she demonstrates a new preview policy-enforcement layer that evaluates outbound email content before it reaches customers, and she explains why organizations may want those guardrails in place. Taylor shows both the administrative steps and the operational behavior, so viewers can see how policies act on messages written by agents as well as those that are AI-generated. Consequently, the video serves as a hands-on primer for teams planning to protect brand tone, compliance, and security across email channels.

Key features demonstrated

First, the video highlights the ability to validate each outgoing email against business-defined rules and to apply enforcement actions that either flag or block content. Taylor notes that the system checks messages regardless of authorship, and therefore governance becomes an operational content-safety control rather than only a role-and-configuration model. In addition, she points out default system policies that target issues such as Email Template Adherence, Groundedness, and Profanity so organizations start with sensible coverage out of the box. Moreover, she explains how the preview shows results and behavior so teams can understand impacts before enforcement reaches customers.

Second, the walkthrough covers expanded scope beyond plain text: administrators can evaluate text-based attachments like PDFs and Word documents as part of the pre-send check. Taylor demonstrates configurable strictness levels — Low, Medium, High, and Very High — which let organizations tune sensitivity based on risk appetite. She also shows how attachment scanning and strictness combine to form a more complete pre-send control that fits different regulatory and brand needs. As a result, the feature reads as a layered safety net that tries to balance protection with productivity.

How to configure the governance feature

Taylor walks viewers through the setup using the Copilot Service admin center and the Copilot Service workspace, starting with preconditions that must be in place. She explains that admins typically need the CSR Manager role and that service representatives require their own role; additionally, extra privileges or global access to specific Dataverse tables are necessary for full administration. Taylor also reminds watchers that the environment must have AI agents enabled and that organizations will need Copilot credits and consent for potential data movement across regions when required. Thus, the prerequisites combine security, licensing, and access considerations that teams must plan for.

Next, Taylor shows the specific steps to enable and manage governance: in the admin center go to Customer Support then Quality management, select Manage for Governance and toggle the feature on. Policy creation happens in the Copilot Service workspace under Service > Governance, where administrators can pick a default system policy or create a new one. She demonstrates filling out policy name and description, choosing content type (usually Email), deciding whether the rule applies to representatives or autonomous AI, and selecting enforcement behavior such as flagging or blocking. Finally, Taylor emphasizes using the simulation option to test policies before publishing them to production so teams can validate outcomes safely.

Tradeoffs and operational challenges

Taylor discusses several tradeoffs that organizations must weigh when they enable pre-send governance, starting with enforcement style: blocking prevents risky content from reaching customers but can interrupt agent workflows, whereas flagging preserves speed but may let some issues slip through. Therefore, teams must balance the desire for strict control against the need for quick, empathetic responses to customers. She also highlights the risk of false positives, especially at strict settings, where overly aggressive rules can frustrate agents and require extra review steps. Consequently, tuning and monitoring become ongoing tasks rather than one-time setup activities.

Another challenge in her demo is the performance and privacy tradeoff when scanning attachments and evaluating more complex content. While broader scanning improves safety, it can add latency and raise questions about where data is processed, which matters for regional compliance. Taylor advises that simulation does not always capture every live scenario, so organizations should plan staged rollouts and monitor results closely. In addition, governance depends on accurate content understanding from AI, and issues like hallucination or inconsistent grounding mean administrators cannot treat policy checks as a silver bullet.

Recommendations and next steps

Based on the video, Taylor recommends starting with simulation and using conservative strictness settings while the team measures impacts on response time and customer satisfaction. She suggests involving legal, compliance, and frontline agents when defining policy conditions so the rules align with real-world needs and brand voice. Furthermore, regular review cycles and clear escalation paths help teams refine rules and reduce friction for agents who interact with blocked or flagged content. In practice, this approach reduces surprises and helps teams build confidence in the governance layer over time.

In conclusion, the YouTube walkthrough by Dian Taylor provides a clear, actionable tour of the new governance preview in Dynamics 365 Customer Service and explains both the technical steps and operational decisions involved. For organizations adopting AI features in customer service, the demonstration makes a strong case for adding content-safety controls while also highlighting the careful tradeoffs required. Consequently, teams should treat governance as an evolving capability that needs testing, monitoring, and cross-team collaboration to deliver safe and efficient customer communications.

Keywords

configure governance D365 Customer Service, Dynamics 365 Customer Service governance best practices, D365 Customer Service governance setup guide, Dynamics 365 permissions and governance, governance policies D365 Customer Service, D365 Customer Service security and compliance, configure service management governance Dynamics 365, D365 governance configuration tutorial