Plans and permissions:
- Add-on: AI agents, billed separately
- See scores on cases in Quality review: Admin, Manager, or Team Lead role, or a custom role with Quality review set to Review quality for all agents or Review quality for certain agents
- See score trends in the AI agent performance report: Admin, Manager, or Team Lead role, or a custom role with AI agent performance set to Can view
Use AI experience scores and outcomes to understand how your AI agents perform without reading every transcript. Assembled uses AI to evaluate every conversation an AI agent handles, and gives it two complementary scores:
- AI outcome says what happened to the case, such as whether the AI agent resolved it or handed it off.
- AI experience says how well the interaction went for the customer.
Together, they show how often AI agents resolve or hand off cases, where conversations loop or stall, how customers likely felt, and where to improve your workflows, prompts, or knowledge.
Scores are given to conversations that AI agents handle directly, not to cases where support team members use Copilot.
To see them, go to AI quality > QA review, Report > AI agent performance, or AI support > Workflows.
Where to find the scores
| Where | What you see |
|---|---|
| AI quality > QA review | Each case’s AI outcome and AI experience in the case list, on the Email agent, Chat agent, and Voice agent tabs. Open a case to see its sub-scores and the explanation. See Quality review. |
| Report > AI agent performance | Outcomes, rates, and experience scores over time, for all your AI agents. |
| AI support > Workflows | Outcomes and performance for each workflow. See Workflows page. |
AI outcome
The AI outcome describes how the case ended. When you open a case in Quality review, it’s shown as Case status.
| Outcome | What it means |
|---|---|
| Resolved | The AI agent fully answered the customer’s question without help from your support team. |
| Handed off | The AI agent escalated the conversation to your support team. |
| Abandoned | The customer left the conversation before it was resolved. |
| Active | The conversation is still in progress. |
AI experience score
While the AI outcome says what happened, the AI experience score says how good the interaction was for the customer. Each scored conversation gets an overall rating of Excellent, Good, or Poor, based on three sub-scores: resolution progress, efficiency, and sentiment. When you open a case, you also see a short explanation of the score.
Positive AI experience in Quality review is the percentage of scored cases rated Good or Excellent.
Resolution progress
How far the AI agent moved the customer toward a solution. Use it for cases that weren’t resolved, to see how much the AI agent did before the case was handed off or abandoned. When you open a case, this sub-score is labeled Resolution.
| Rating | What it means |
|---|---|
| Resolved | The AI agent solved the customer’s issue completely. |
| Partially resolved | The AI agent made progress, but didn’t finish. |
| Triaged | The AI agent helped diagnose the issue or gather information. |
| Uncertain | The AI agent didn’t clearly understand what the customer needed. |
| Not resolved | The AI agent couldn’t provide useful help. |
The AI agent performance report counts Resolved and Partially resolved as positive.
Efficiency
How smooth and direct the conversation was. Use it to find flow problems, unclear prompts, or loops.
| Rating | What it means |
|---|---|
| Smooth | Solved in a single, efficient exchange. |
| Quick | Mostly efficient, with little back-and-forth. |
| Moderate | Noticeable extra steps or repetition. |
| Repetitive | The AI agent and the customer got stuck going in circles. |
| Stuck | The conversation stalled without progress. |
The AI agent performance report counts Smooth and Quick as positive.
Sentiment
How the customer likely felt by the end of the conversation. It combines how the customer felt with how that changed over the conversation. Use it to monitor the customer experience and the risk of dissatisfaction.
| Rating | What it means |
|---|---|
| Satisfied | A strongly positive experience. |
| Improving | The customer’s sentiment started low but improved. |
| Neutral | No strong emotional signals. |
| Dissatisfied | Negative throughout: the customer started and ended frustrated. |
| Declining | The customer’s sentiment got worse over the conversation. |
The AI agent performance report counts Satisfied, Improving, and Neutral as positive.
How resolution and containment rates are calculated
The AI agent performance report turns outcomes into two rates:
- Resolution rate = Resolved cases ÷ Total cases. It counts only cases the AI agent resolved.
- Containment rate = (Resolved cases + Abandoned cases) ÷ Total cases. It counts every case the AI agent handled without handing off.
For example: your chat agent handles 100 cases. It resolves 60, 10 customers leave, and it hands off 30. The resolution rate is 60 ÷ 100 = 60%, and the containment rate is (60 + 10) ÷ 100 = 70%.
Use the scores together
The AI outcome shows effectiveness: what ultimately happened to the case. The AI experience score shows quality: how the interaction went. Looking at both tells you not just whether AI agents handled the work, but how customers experienced it.
| What you see | What it suggests | What to do |
|---|---|---|
| Many resolved cases, but poor experience scores | The AI agent resolves cases, but with friction. | Make your workflows more efficient. |
| Strong experience scores, but many handoffs | Customers have good conversations, but the AI agent can’t finish the job. | Add knowledge, or automate more of the work. |
| More Repetitive efficiency and Dissatisfied sentiment | Customers are getting stuck. | Review your conversation design, prompts, and routing. |