DialNexa Annotations And Call Feedback

Use DialNexa call annotations to mark transcript turns, provide corrected behavior, and generate prompt improvement suggestions.

DialNexa annotations let users turn a bad or important transcript turn into prompt improvement feedback. From the call detail transcript, users can select a turn, explain expected behavior, and submit an annotation that creates a prompt suggestion for review.

A good annotation does not say the agent was bad. It says what the agent should have done instead.

Annotation Fields

Annotations should be specific enough to improve a prompt.

Field	Purpose
Selected transcript turn	The exact call moment being reviewed.
Feedback	What was wrong, risky, confusing, or missing.
Expected responses	One or more examples of better behavior.
Agent id context	Connects the suggestion to the right agent.

Create Useful Feedback

1. Pick one turn

   Annotate the exact moment instead of the entire call.

2. Describe the mismatch

   Explain what the caller expected and what the agent did.

3. Add expected response examples

   Give the model prompt reviewer concrete target behavior.

4. Submit and review suggestions

   Use prompt suggestions as candidates, not automatic edits.

Annotation Quality Checks

Too vague

Rewrite feedback to include the caller situation and desired behavior.

Too broad

Create multiple annotations for separate issues.

No expected response

Add at least one better answer so the suggestion has direction.

Wrong agent context

Confirm the call belongs to the agent you want to improve.

Related Reading

Call Detail Page

Open transcript turns.

Prompts

Apply improvements carefully.

Testing Agents

Retest the fixed behavior.