Command Palette

Search for a command to run...

Sign in
AI Assistant

AI Analytics Dashboard

Full transparency into every AI decision. Track approval rates, confidence scores, and accuracy metrics. Watch your AI get smarter over time—and prove it with data.

Key Metrics

Approval Rate

Percentage of AI actions approved by you or auto-approved based on confidence thresholds

94% means 94 out of 100 AI actions were approved

Confidence Score

AI's self-assessed certainty for each decision, ranging from 0 to 1

0.89 average means the AI is typically 89% confident

Accuracy

How often approved AI actions led to positive outcomes

96% means 96% of actions achieved their intended result

Actions Processed

Total number of AI actions taken in the selected time period

127 today means the AI handled 127 customer interactions

Knowledge Quality Metrics

Track the health and effectiveness of your knowledge base:

Evidence Coverage

Percentage of AI responses that include citations from your knowledge base

95% means 95 out of 100 answers cite a source

Stale Source Alerts

Number of knowledge sources flagged as outdated and needing review

3 stale sources means 3 items are over 90 days old

Gap Acceptance Rate

Percentage of AI-suggested knowledge gaps that you added to the knowledge base

78% means you added 78 out of 100 suggested topics

Citation Accuracy

How often the cited source correctly answers the customer's question

97% means 97 out of 100 citations were relevant

Understanding Confidence Scores

Every AI action includes a confidence score that reflects how certain the AI is about its decision:

High Confidence0.90 - 1.00

AI is very certain. These actions can typically be auto-approved.

Good Confidence0.80 - 0.89

AI is reasonably sure. May warrant a quick review.

Moderate Confidence0.70 - 0.79

AI is somewhat uncertain. Review recommended before approval.

Low ConfidenceBelow 0.70

AI flagged this for human review. Always requires manual approval.

Decision History

The decision history shows every AI action with full context:

Action Description - What the AI did (e.g., "Replied to Sarah about availability")

Confidence Score - How certain the AI was about this action

Status - Whether it was approved, auto-approved, pending, or rejected

Timestamp - When the action occurred

Linked Context - Customer, booking, or conversation associated with the action

Decision Statuses

Auto-Approved

AI action met confidence threshold and was executed automatically

Approved

You or a team member reviewed and approved the AI action

Pending

AI action is waiting for human review before execution

Rejected

You rejected the AI's proposed action and took over manually

Improving AI Performance

Your AI learns from every interaction. Here's how to help it improve:

1

Review pending actions promptly

The faster you approve or reject, the faster the AI learns your preferences.

2

Add knowledge when AI hesitates

Low confidence often means missing information. Add to your knowledge base.

3

Adjust autonomy levels gradually

As approval rates improve, consider increasing AI autonomy for efficiency.

4

Check weekly trends

Monitor the learning rate metric to ensure consistent improvement.

Configuring Auto-Approval

Set thresholds for when AI can act without waiting for approval:

Confidence Threshold - Actions above this score are auto-approved (default: 0.90)

Action Type Rules - Different thresholds for messages, bookings, reminders

Risk Level Override - High-risk actions always require approval regardless of confidence

Troubleshooting Low Approval Rates

If your approval rate drops below 90%, consider these fixes:

Problem: AI suggests wrong services

Solution: Update your knowledge base with service details and common customer requests

Problem: AI responds with incorrect information

Solution: Add FAQs and correct information to the knowledge base

Problem: AI doesn't understand customer intent

Solution: Review rejected actions and add examples to help the AI learn patterns

Problem: Too many actions requiring approval

Solution: Consider raising autonomy level if approval rate is consistently high

Learning Rate Explained

The learning rate shows how quickly your AI is improving:

Example: 2.3% improvement/week

This means your AI's performance metrics (approval rate, accuracy, confidence) are improving by an average of 2.3% each week. A positive learning rate indicates the AI is successfully adapting to your business patterns.

Pro Tips

  • • Check your analytics dashboard at least weekly to spot trends
  • • Export reports for team meetings or business reviews
  • • Compare performance across different action types
  • • Use insights to optimize your knowledge base content