This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
An organization plans to let its HR policy agent submit leave requests on behalf of employees. Which design decision best addresses the risk of this action?
Raise the agent's moderation level so that harmful leave submissions are blocked.
Rely on the platform's content safety filters to stop incorrect leave submissions.
Require a person to approve each leave submission before it takes effect.
An employee politely asks an HR policy agent for a coworker's leave balance, and the agent returns it. Which change addresses the cause of this behavior?
Update instructions, knowledge scope, and tool permissions so the agent declines requests for other employees' records.
Raise the moderation level so that the content filter blocks requests for another employee's data.
Report the response to Microsoft support as a failure of the platform's content filtering.
A maker's agent on the standard harness uses the default agent-level moderation. One generative answers node summarizes medical leave policies that describe medical procedures, and it sometimes returns no answer. The maker wants lower moderation for that node only. What should the maker do?
Lower the agent-level moderation on the agent's Generative AI settings page.
Set a lower moderation level in the Properties pane of that generative answers node.
Set the Moderation level on the AI & behavior tab of the agent's settings to Low.
During testing, a preview model gives an HR policy agent the best answers to multistep leave questions. The team plans to launch the agent in production next month. Which model decision fits production readiness?
Launch on a generally available model that fits the workload, and keep testing the preview model.
Launch on the preview model, and rely on the Default model as a fallback if quality drops.
Launch on the preview model after an admin turns on the preview and experimental models setting.
A team builds an agent on the GitHub Copilot harness. The team wants users to know from the start that they're working with AI, what the agent covers, and that its answers can be wrong. Which configuration meets this goal?
Turn on User feedback so that users can rate the quality of each response.
Set the Moderation level to Maximum so that inaccurate answers are filtered out.
Write a Greeting message that discloses the agent's AI use, purpose, and limitations.
You must answer all questions before checking your work.
Was this page helpful?
Need help with this topic?
Want to try using Ask Learn to clarify or guide you through this topic?