Understanding AI Experience Scores & Outcomes – Assembled

Your AI agents are automatically evaluated (via AI) through two main complementary scores:

AI Experience Score tells you how well the interaction went.

AI Outcome tells you what happened to the case.

Together, they give you a scalable way to understand AI performance and customer experience — without manually reviewing transcripts.

These scores help you:

Understand how often AI resolves vs. escalates cases
Spot friction, loops, and poor experiences
See how customers likely felt
Identify where to improve flows, prompts, or knowledge
Track performance trends over time

Scores are available for conversations handled directly by the AI agent (not human-handled Copilot conversations). You can find them in QA, Workflow Monitoring, and AI Agent Performance Reporting.

AI Outcome

AI outcome describes how the case ended. There are four statuses:


Resolved	The AI agent fully answered the customer's question without assistance from your support team.
Handed Off	The conversation was escalated to your support team.
Abandoned	The customer disengaged before they got to resolution.
Active	The conversation is still in progress.

This will help you understand:

Containment rate: how often the AI handled the conversation without escalating to support (Containment rate = Resolved + Abandoned / total cases)
Escalation patterns
Where customers drop off

AI Experience Scores

While AI outcome tells you what happened, the AI Experience Score tells you how good the interaction felt to the end user.

Each conversation receives an overall rating: Excellent, Good, or Poor.

This overall score is based on three subcomponents.

Resolution Progress Sub-score

How far the AI moved the customer toward a solution. Use this for cases where the AI outcome wasn't fully Resolved to understand how much progress the AI agent made before the case was handed off or abandoned.


● Fully Resolved	The stated issue was completely solved by the AI agent.
● Partially Resolved	Some progress, but not fully complete.
● Triaged	Helped diagnose or gather information.
● Unclear	Customer need wasn’t clearly understood.
● Not Resolved	The AI agent could not provide useful help.

Efficiency Sub-score

How smooth and direct the interaction was. Use this to identify flow problems, poor prompts, or looping behavior.


● One-Shot	Solved in a single, highly efficient exchange
● Direct	Mostly efficient with minimal back-and-forth
● Frictional	Noticeable extra steps or repetition
● Looping	AI and customer got stuck in inefficient cycles
● Dead End	Interaction stalled without progress

Sentiment Sub-score

How the customer likely felt by the end of the conversation. This is a combo measure of both the trajectory and the sentiment over the course of the conversation. Use this to monitor experience and dissatisfaction risk.


● Delighted	Strongly positive experience
● Recovered	Sentiment started low but improved during the interaction
● Neutral	No strong emotional signals
● Degraded	Sentiment worsened over the course of the interaction
● Frustrated	Negative throughout. Customer started frustrated and ended frustrated.

Using the Scores Together

AI outcome shows effectiveness — what ultimately happened to the case.

AI Experience Score shows experience quality — how well the interaction went.

Looking at both together helps you understand not just whether AI handled the work, but how customers experienced it.

For example, if AI resolves many cases but experience scores are poor, the AI may be technically successful but creating friction — focus on improving flows and efficiency. If experience scores are strong but many conversations are handed off, customers are having good interactions but the AI lacks capability — expand knowledge or automation coverage. If looping behavior and frustrated sentiment increase, review conversation design, prompts, or routing to prevent customers from getting stuck.