Improving CSAT with Ada

When to use this skill

Use this skill when the user wants to:

• •Improve their CSAT score
• •Understand why customers are giving low satisfaction ratings
• •Get actionable recommendations for improving customer experience
• •Analyze patterns in negative feedback
• •Identify gaps in their AI agent's responses

Workflow

Step 1: Get current CSAT metrics

Start by understanding the baseline:

Use get_ada_metric to retrieve CSAT rate for the last 7 days (or user-specified timeframe)

Report the current CSAT percentage and compare to any available historical data.

Step 2: Identify low-CSAT conversations

Use filtering to find problem areas:

1. Use get_available_filters to see available filter options
2. Use get_conversations_by_filters with CSAT scores of 1-2 to find low-satisfaction conversations
3. Request 50-100 conversations for meaningful pattern analysis

Step 3: Analyze conversation patterns

Pull transcripts and summaries to understand root causes:

Use get_conversation on a sample of 10-20 low-CSAT conversations

Look for:

• •Common customer inquiries that led to dissatisfaction
• •Points where the agent's response was unhelpful
• •Missing information or incorrect answers
• •Tone or empathy issues
• •Handoff failures or delays

Step 4: Review current configuration

Understand what the agent is working with:

Use get_ada_configuration to retrieve:
- Playbooks
- Guidance/custom instructions
- Actions
- Coaching rules
- Company description

Step 5: Search for gaps

Check if relevant content exists:

1. Use search_knowledge with topics from low-CSAT conversations
2. Use search_coaching to find relevant coaching rules

Identify:

• •Topics with no knowledge coverage
• •Outdated or incomplete articles
• •Missing coaching for common scenarios

Step 6: Provide actionable recommendations

Structure recommendations as:

1. •

   Quick wins - Changes that can improve CSAT immediately

   • •New coaching rules for common failure patterns
   • •Updates to existing knowledge articles
   • •Tone/empathy guidance additions
2. •

   Medium-term improvements - Changes requiring more effort

   • •New playbooks for uncovered scenarios
   • •Knowledge base expansions
   • •Action integrations
3. •

   Strategic changes - Longer-term considerations

   • •Process changes
   • •Integration improvements
   • •Training data updates

Example output format

## CSAT Analysis Summary

**Current CSAT**: 72% (last 7 days)
**Conversations analyzed**: 50 low-CSAT conversations

### Key Findings

1. **Refund policy confusion** (15 conversations)
   - Customers frustrated by unclear refund timeline
   - Agent responses lack specific day counts

2. **Shipping status inquiries** (12 conversations)
   - Agent can't access real-time tracking
   - Customers handed off unnecessarily

### Recommendations

#### Quick Wins
- Add coaching: "When discussing refunds, always mention the 5-7 business day processing time"
- Update knowledge article "Refund Policy" with specific timelines

#### Medium-Term
- Create playbook for shipping status that integrates with tracking API
- Add action to pull order status from backend system

Tips for better analysis

• •Be specific about timeframes when querying metrics
• •Analyze at least 50 conversations for meaningful patterns
• •Cross-reference findings with current configuration
• •Prioritize recommendations by potential CSAT impact
• •Consider both content gaps AND tone/empathy issues