Correct AI answers

Correcting AI answers helps train your AI Agent using:

• validated responses

• human-reviewed replies

• support workflows

• operational feedback

This improves:

• answer accuracy

• consistency

• long-term AI performance

Open a conversation

To review and correct AI answers:

1. Open your AI Agent workspace

2. Go to Inbox

3. Open a conversation

Inside the conversation view, you can review how the AI responded to customer questions.

Review AI responses

When reviewing conversations, check if the AI:

• answered correctly

• used the right tone

• followed instructions

• referenced accurate information

Look for:

• missing details

• incorrect answers

• outdated information

• weak support guidance

Correct incorrect answers

If the AI response is inaccurate:

1. Edit or replace the response

2. Provide the correct answer

3. Save the updated response if training options are available

This helps reinforce better answers for future conversations.

Use AI Analysis

Mando AI may provide AI-generated analysis for conversations.

This can help identify:

• customer intent

• response quality

• missing information

• escalation opportunities

AI Analysis can make conversation reviews faster and more consistent.

Improve recurring support workflows

Many teams focus corrections on:

• high-frequency support questions

• refund requests

• shipping policies

• onboarding questions

• troubleshooting flows

Improving common workflows usually creates the biggest impact on AI quality.

Combine corrections with better knowledge

Correcting answers works best when combined with:

• updated website content

• improved internal articles

• better snippets

• synced support documentation

The AI performs best when both:

• training data

• knowledge sources

are kept updated.

Review conversations regularly

Ongoing review helps teams:

• identify weak responses

• improve support quality

• maintain consistency

• adapt to business changes

Many businesses review conversations weekly to improve AI performance continuously.

Archive outdated workflows

As policies or products change:

• remove outdated corrections

• update old workflows

• refresh support guidance

Keeping training data current helps maintain accurate responses.

Best practices for correcting answers

For better long-term results:

• focus on recurring questions first

• keep corrections clear and concise

• maintain consistent tone

• avoid conflicting guidance

Consistent corrections usually improve AI reliability significantly.

Next steps

• Use AI Analysis

• Understand conversation details

• Understand trained answers