How do I use AI-suggested scores?
Required Feature Flags
The following feature flags and permissions are required to use this feature:
Feature Flag | Description |
SmartScore V2 | Enables AI-powered scoring on line items that have AI guidelines configured |
Required Permissions:
Evaluate (
quality.evaluations.evaluate) — to perform evaluations and review AI suggestions
When AI suggestions are enabled, evaluagent reads the conversation, looks at the scorecard's line item guidelines, and proposes a score for you to review. You stay in control — accept the suggestion, change it, or stick with your own answer.
Step 1: Open a contact to evaluate
Go to Evaluate > click Contacts to evaluate and open the contact you want to score.
The evaluation form opens alongside the conversation transcript or recording.
Step 2: Spot the AI suggestions
Line items with an AI suggestion show an indicator next to the score options. Click the indicator (or the line item) to see:
Element | What it shows |
AI suggested score | The score the AI recommends for this line item |
AI reasoning | A short explanation of why the AI suggested that score, with references to the conversation |
If a line item doesn't have AI guidelines, you won't see a suggestion. Score it manually.
Step 3: Review the AI's reasoning
Don't accept a suggestion blindly. Read the reasoning and check it against the conversation.
Ask yourself:
Did the AI pick up on the right moments in the conversation?
Does the reasoning match the scorecard guideline for this line item?
Is there context the AI might have missed (tone, customer history, your own knowledge of the situation)?
The AI is a second opinion, not the final answer.
Step 4: Accept, change, or override
Once you've reviewed the suggestion, choose how to record your decision:
Accept — the AI's score becomes your score for the line item
Change — pick a different score from the available options. Use this when the AI got close but isn't quite right
Override — keep your own score and ignore the suggestion. Use this when you disagree with the AI
You can add a comment explaining why you changed or rejected a suggestion. This is useful for later coaching and for refining the line item guidelines over time.
Step 5: Finish and publish the evaluation
Score any remaining line items, add overall feedback, and publish the evaluation as you normally would.
The system records each AI suggestion alongside your final decision, so reports show:
Whether the AI suggested a score for the line item
What the AI suggested
What you decided
Any comment you added
When AI suggestions work best
AI suggestions are most useful when:
Line item guidelines are clear and specific
The transcript or recording is complete and accurate
The criteria are objective (e.g. "Did the agent verify the customer's identity?")
They're less reliable when:
Guidelines are vague or open to interpretation
Context outside the conversation matters
Tone or cultural nuance is part of the score
In those cases, treat the suggestion as a prompt to look more closely, not as the answer.
Tips for getting the most from AI scoring
Read the reasoning every time, even when you agree with the score
Be consistent in how you accept or override across evaluations
Add a comment when you override — it helps your team spot patterns
If the AI keeps getting a particular line item wrong, raise it with your administrator. The line item's AI guidelines may need updating