The important implementation detail is that detection sensitivity and enforcement action are separate settings. Turning up strictness does not automatically turn a warning into a blocked email—a distinction administrators should settle before deploying tighter controls.
What the strictness levels change
Microsoft Learn describes the choices as follows:
| Level | Evaluation behavior |
|---|---|
| Low | More potential violations can pass without a flag. |
| Medium | Default, balanced evaluation. |
| High | Flags more potential violations. |
| Very High | Includes minor deviations from policy definitions. |
Separately, administrators choose whether detected violations are silently recorded, shown as warnings that permit sending, or blocked until resolved. These controls concern outbound email, not a blanket policy mechanism for every customer-service channel.
That separation creates useful flexibility. As an illustrative deployment choice—not a Microsoft-prescribed configuration—a team could evaluate a sensitive-information rule aggressively while initially recording its findings for review. A tone rule could use a less sensitive setting and a visible warning. The sensible combination depends on the consequences of both missed violations and unnecessary interruptions.
The takeaway: decide what should count as a violation separately from what should happen afterward.
Where the capability fits
Governance is a skill within the Quality Management Agent in Dynamics 365 Customer Service. Microsoft’s August 5, 2026, general-availability announcement explicitly included flexible strictness levels alongside attachment checks. This is therefore an existing governance capability, not an October-only launch. Microsoft says it evaluates outbound email written by service representatives or generated by AI before sending.
Microsoft positions the feature as proactive policy enforcement. That is a useful objective, but administrators should treat it as a configurable checking mechanism rather than an assurance that every risky message will be caught. The documented Low setting can miss potential violations, while Very High can flag minor deviations.
Configure the controls deliberately
Microsoft’s documented prerequisites include the CSR Manager role, configured Copilot credits, enabled AI agents, and cross-region data-movement consent where required. To enable Governance, open Copilot Service admin center > Customer Support > Quality management, select Manage for Governance, and enable its toggle. Custom security roles also require appropriate table permissions.
For policy configuration, Microsoft Learn directs administrators to:
- Open Copilot Service workspace > Service > Governance.
- Create a policy or edit an existing one.
- Choose the email author category, enforcement action, strictness, and applicable conditions.
- Review generated instructions, then save and publish.
With no content conditions, all emails in the check are evaluated.
Test the tradeoff before tightening enforcement
Microsoft provides simulation using Excel uploads or organization records; the latter supports up to 30 emails. Results include violation status and reasons, while detection logs support ongoing review.
A practical approach is to compare clearly acceptable, clearly prohibited, and borderline messages before changing production blocking behavior. That recommendation is an operational synthesis of the documented controls, not a claim of independently tested accuracy.
Finally, Microsoft expressly says Governance should not be used for employment-impacting decisions such as compensation or seniority. Its documentation also places responsibility for applicable monitoring, notice, and consent obligations on customers. A stricter policy setting does not remove those responsibilities.
References
- Dynamics 365 Customer Service: Workforce engagement management - Manage policy strictness to enforce governance policies Microsoft 365 Roadmap · 2026-10-23T23:10:34.726181Z
- Governance Now Available for Dynamics 365 Customer Service microsoft.com
- Configure governance policies | Microsoft Learn learn.microsoft.com