Configure the Quality Assurance Agent
The Quality Assurance Agent brings AI-powered support to contact center supervisors and quality managers across their day-to-day activities. Rather than limiting oversight to manual spot-checks, the agent autonomously monitors conversations in real time, proactively alerts supervisors when quality issues arise, and enables timely intervention, keeping service standards high without adding operational overhead.
Quality evaluation is the agent's core skill. It assesses conversations both during and after completion, delivering quality scores, compliance checks, and coaching insights across every interaction. When supervisors need to act, the agent surfaces the right conversation at the right moment—through a real-time alert, a flagged coaching opportunity, or a post-conversation evaluation for follow-up review.
Core capabilities
The Quality Assurance Agent delivers three types of value:
- Real-time quality evaluation: The agent monitors active conversations as they happen, scoring quality criteria and flagging issues so supervisors can act immediately.
- Post-conversation evaluation: After conversations close, the agent scores interactions against your defined evaluation criteria and surfaces coaching opportunities.
- Supervisor monitoring and intervention: Supervisors get holistic visibility across all active conversations, with the ability to monitor, consult, join, or transfer conversations at any moment.
Prerequisites
Before you enable the Quality Assurance Agent, confirm:
- The Omnichannel administrator role is assigned to the user configuring the agent.
- A pay-as-you-go plan is configured for your environment.
- Enable AI agents is turned on in the Power Platform admin center.
- Consent for potential cross-region data movement is provided.
Configure the Quality Assurance Agent
- In the Copilot Service admin center, go to Customer Support > Quality management.
- Select Manage for Quality Evaluation Agent. The Quality Evaluation Agent page appears.
- In the Prerequisites section, check the status of three setup steps:
- Step 1: Connection References
- Step 2: Power Automate Flows
- Step 3: Copilot Studio Agent
- If Step 1 shows In progress or Incomplete, select Manage connections, then select Update connection references to use your connector.
- If any Power Automate flows in Step 2 are turned off, open and enable each flow in Power Automate.
- Once all three steps show Ready, select Publish to complete the setup.
Important
After you complete connection reference configuration, do a hard refresh of the page to see the updated status. If prerequisites still appear incomplete after a hard refresh, follow the manual setup guide linked in the admin center.
Enable evaluation for conversations
With prerequisites confirmed, enable the Quality Assurance Agent for the conversation record type:
- On the Quality Evaluation Agent page, in the Enablement by record type section, select Conversation, and save.
- In Specify data, review the default data fields included for conversation evaluation. Select Manage data to customize which fields the agent uses as input. You can include up to 10 one-to-one data types and six one-to-many data types beyond the defaults.
Including richer data, such as conversation transcripts, channel type, and queue name, gives the agent more context to produce accurate evaluations.
Set scoring and thresholds
The scoring configuration defines what "good" and "poor" mean in your contact center.
- In the Evaluation criteria score section, select Enable scoring for criteria. Confirm by selecting Turn on in the dialog that appears.
- From the Set threshold value out of 100 dropdown, select the score that distinguishes an acceptable interaction from one that needs follow-up.
Conversations that score below the threshold appear as candidates for coaching review in supervisor reports. Set this threshold at a level that reflects your organization's actual quality standards.
Important
Evaluation scoring can't be turned off after you enable it. Plan your scoring model carefully before activating this setting, including which behaviors to score, how to weight them, and what threshold to set.
Create evaluation criteria
Evaluation criteria are the forms the Quality Assurance Agent uses to score conversations. Each form contains questions with defined answer choices and scoring logic.
Users with the Quality Manager or Quality Administrator role create criteria forms in the admin center. Design questions around the behaviors and outcomes your organization cares about most, such as greeting adherence, resolution accuracy, empathy, and required disclosures.
For example, a quality evaluation might score a conversation across these dimensions:
| Criterion | Score |
|---|---|
| Communication | 100 |
| Accountability | 100 |
| Resourcefulness | 100 |
| Empathy | 30 |
| Overall evaluation score | 82.5 |
Well-designed criteria forms are specific enough to give the agent clear scoring guidance but broad enough to apply consistently across a wide range of conversation types.
Create evaluation plans
An evaluation plan defines which conversations to evaluate, how often, and under what conditions. You can target conversations based on attributes such as queue, workstream, channel, or representative.
The Quality Assurance Agent supports two types of evaluation plans:
- Scheduled plans: The agent runs automatically on the schedule you define, evaluating conversations that match your criteria.
- Real-time plans: The agent evaluates conversations as they happen, providing scores and alerts during active interactions.
For conversations that require immediate review, supervisors can also trigger on-demand evaluation at any time without waiting for the next scheduled run.
Monitor conversations in real time
The Quality Assurance Agent gives supervisors holistic visibility across all active conversations. From the supervisor dashboard, supervisors can:
- Monitor an ongoing engagement: View all active conversations with real-time sentiment analysis and quality scores.
- Intervene by initiating a consult or transfer: Join a conversation, provide guidance to the representative through internal messaging, or transfer the conversation to a different representative or queue.
- View quality metric dashboards: See aggregated quality scores across all open and recently completed conversations.
- Join the conversation directly: When supervisor intervention is needed, participate in the conversation alongside the representative.
This real-time monitoring capability means supervisors shift from reactively hunting for quality problems to proactively acting on the insights the agent surfaces during every conversation.