Improving Automated Resolution Rate with Ada

When to use this skill

Use this skill when the user wants to:

• Improve their automated resolution (AR) rate
• Reduce handoffs to human agents
• Understand why conversations aren't being resolved automatically
• Identify automation opportunities
• Analyze unresolved conversation patterns

Understanding AR

Automated Resolution (AR) measures the percentage of conversations fully resolved by the AI agent without human intervention. Key factors affecting AR:

• Knowledge coverage: Does the agent have answers to customer questions?
• Action capabilities: Can the agent perform the required tasks?
• Playbook design: Are workflows comprehensive enough?
• Handoff triggers: Are handoffs happening unnecessarily?

Workflow

Step 1: Get current AR metrics

Establish the baseline:

Use get_ada_metric to retrieve:
- Automated resolution rate (last 7 days and last 30 days for trend)
- Engaged conversation volume

Calculate week-over-week or month-over-month changes if data is available.

Step 2: Identify unresolved conversations

Find where automation is failing:

1. Use get_available_filters to understand filter options
2. Use get_conversations_by_filters with automated_resolution_status = "Unresolved"
3. Request 50-100 conversations for pattern analysis

Step 3: Analyze resolution reasons

Examine why conversations weren't resolved:

Use get_conversation on 15-25 unresolved conversations

Categorize by failure reason:

• Knowledge gap: Agent didn't have the answer
• Action limitation: Agent couldn't perform the required task
• Handoff trigger: Explicit handoff request or rule triggered
• Complexity: Multi-step issue beyond current capabilities
• Edge case: Unusual scenario not covered by playbooks

Step 4: Identify high-volume failure patterns

Look for the biggest opportunities:

Review customer_inquiry_summary and automated_resolution_reason fields

Group by:

• Topic/intent
• Failure reason
• Volume (how many conversations with this pattern?)

Step 5: Review current configuration

Understand existing capabilities:

Use get_ada_configuration to retrieve:
- Playbooks (what workflows exist?)
- Actions (what can the agent do?)
- Coaching (what guidance exists?)
- Knowledge (through search_knowledge for specific topics)

Step 6: Search for coverage gaps

For each high-volume failure pattern:

1. Use search_knowledge to check if relevant articles exist
2. Use search_coaching to check for relevant guidance

Step 7: Provide prioritized recommendations

Structure by impact and effort:

## High Impact, Low Effort
- Quick knowledge additions
- New coaching rules
- Playbook tweaks

## High Impact, High Effort  
- New action integrations
- Complex playbook creation
- API connections

## Medium Impact
- Edge case coverage
- Refinements to existing content

Example output format

## AR Analysis Summary

**Current AR**: 65% (last 7 days)
**Previous period**: 62% (prior 7 days)
**Trend**: ↑ 3% improvement

**Unresolved conversations analyzed**: 75

### Top Unresolved Patterns

| Pattern | Volume | Failure Reason | Potential AR Lift |
|---------|--------|----------------|-------------------|
| Order cancellation requests | 23 | Action limitation | +5% |
| Complex return scenarios | 18 | Knowledge gap | +4% |
| Account access issues | 12 | Handoff trigger | +2% |

### Recommendations

#### 1. Order Cancellation (Highest Impact)
**Problem**: Agent can't cancel orders; always hands off
**Solution**: 
- Create action integration with order management system
- Add playbook for cancellation flow
**Expected impact**: +5% AR

#### 2. Complex Returns
**Problem**: Return policy article doesn't cover exchanges or partial returns
**Solution**:
- Expand "Returns" knowledge article
- Add coaching for edge cases
**Expected impact**: +4% AR

#### 3. Account Access
**Problem**: Agent hands off on all password reset requests
**Solution**:
- Review handoff trigger rules
- Add self-service password reset playbook
**Expected impact**: +2% AR

Tips for better AR analysis

• Focus on high-volume patterns first (biggest AR lift potential)
• Distinguish between solvable gaps vs. intentional handoffs
• Consider whether some handoffs are appropriate (complex issues, VIP customers)
• Track AR by topic/intent if possible to identify specific weak areas
• Recommend quick wins alongside larger initiatives