# Agent Copilot (Agent Assist)

## Study Notes
| Topic | Description |
|---|---|
| Agent Copilot | AI-powered real-time guidance system for agents |
| Also Known As | Agent Assist, Copilot Assistant |
| Purpose | Provides real-time recommendations and knowledge during customer interactions |
| Activation | Requires Premium edition and Customer Insights module |
| Benefit | Reduces handle time, improves first-contact resolution, enhances agent confidence |

---

## Navigation
Admin → Architect → Agent Copilot
OR
Admin → Contact Center → Agent Assistance → Copilot Configuration

---

## Agent Copilot Overview

Agent Copilot is an AI-powered assistant that provides real-time guidance and recommendations to agents during customer interactions. It analyzes the conversation in real-time and suggests relevant knowledge articles, scripts, and next steps to improve interaction quality and resolution.

### Key Capabilities
- **Real-time recommendations** - Suggests actions based on conversation context
- **Knowledge article suggestions** - Recommends relevant articles automatically
- **Script guidance** - Provides talking points and recommended language
- **Sentiment analysis** - Monitors customer emotion and suggests de-escalation
- **Next action recommendations** - Predicts optimal next steps
- **Agent learning** - Improves over time with agent feedback

### How It Works
1. Agent answers contact
2. Copilot monitors conversation in real-time
3. AI analyzes conversation intent and context
4. System searches knowledge base for relevant information
5. Recommendations displayed in agent interface
6. Agent reviews and applies suggestions
7. Feedback loop improves future recommendations

---

## Edition & Module Requirements

| Requirement | Details |
|---|---|
| Minimum Edition | Premium Edition required |
| Module | Customer Insights add-on module |
| License Type | Agent licenses with Copilot enabled |
| Setup | Admin configuration in Architect |
| Integration | Knowledge management system required |

---

## Study Notes - Copilot Features
| Feature | Description | Use Case |
|---|---|---|
| Knowledge Recommendations | AI-suggested articles from knowledge base | Technical support, FAQs |
| Script Guidance | Real-time conversation scripts and talking points | Sales, compliance-heavy calls |
| Sentiment Monitoring | Real-time emotion analysis of customer | De-escalation, empathy guidance |
| Next Action Suggestions | Recommended next steps for agent | Call routing, transfer decisions |
| Agent Performance Tips | Real-time coaching during interaction | Training reinforcement |
| Historical Context | Customer interaction history suggestions | Personalization, context |
| Product Recommendations | Sales-specific recommendations | Upsell, cross-sell opportunities |
| Compliance Reminders | Real-time compliance guidance | Regulatory requirements |

---

## Implementation Guide

### Step 1: Prerequisites & Planning
1. Ensure organization has Premium edition
2. Purchase Customer Insights module
3. Audit existing knowledge base
4. Document Copilot use cases by queue
5. Plan knowledge article optimization
6. Review agent readiness and training needs

### Step 2: Knowledge Base Configuration
1. Navigate to Admin → Knowledge Management
2. Create/organize knowledge articles
3. Tag articles with metadata (category, queue, intent)
4. Add keywords and synonyms for better matching
5. Ensure article quality and accuracy
6. Set article access permissions

### Step 3: Enable Agent Copilot
1. Go to Admin → Architect → Agent Copilot
2. Enable "Agent Copilot" toggle
3. Select knowledge base source
4. Configure recommendation parameters
5. Set recommendation types to display
6. Choose recommendation frequency

### Step 4: Customize by Queue
1. Create queue-specific Copilot settings
2. Configure knowledge sources per queue
3. Set relevance thresholds
4. Define script templates
5. Establish sentiment trigger rules
6. Test queue-specific configurations

### Step 5: Agent Training & Rollout
1. Train agents on Copilot interface
2. Explain recommendation types
3. Practice with sample interactions
4. Gather initial feedback
5. Monitor early adoption
6. Provide ongoing support

### Step 6: Monitoring & Optimization
1. Review Copilot engagement metrics
2. Monitor agent utilization of recommendations
3. Track recommendation accuracy
4. Gather agent feedback
5. Optimize knowledge articles
6. Adjust recommendation parameters

---

## How to Implement

| Phase | Description | Timeline |
|---|---|---|
| Planning | Audit knowledge base and define use cases | Week 1-2 |
| Setup | Configure Copilot and knowledge sources | Week 2-3 |
| Content | Create/optimize knowledge articles | Week 3-5 |
| Training | Educate agents on features and usage | Week 5-6 |
| Pilot | Deploy to single queue with monitoring | Week 6-7 |
| Rollout | Enable across all queues | Week 7-8 |
| Optimization | Monitor and tune performance | Ongoing |

---

## Agent Copilot Architecture
```
Incoming Contact
    ↓
Agent Accepts Contact
    ↓
Copilot Monitoring Begins
├── Real-time Conversation Analysis
├── Intent Detection
└── Context Extraction
    ↓
AI-Powered Recommendation Engine
├── Knowledge Base Search
├── Relevance Scoring
├── Sentiment Analysis
└── Prediction Models
    ↓
Recommendation Generation
├── Knowledge Articles
├── Scripts & Talking Points
├── Next Action Suggestions
├── Sentiment De-escalation Tips
└── Product/Service Recommendations
    ↓
Display to Agent Interface
    ↓
Agent Reviews Recommendations
    ↓
Agent Applies (or Dismisses) Suggestions
    ↓
Feedback Loop Updates AI Model
```

---

## Real-Time Recommendation Flow
```
Customer Says: "I've been trying to reset my password for hours"

Copilot Analyzes:
├── Intent: Password Reset Help
├── Sentiment: Frustrated/Angry
├── Context: Technical Issue
└── Duration: Extended problem
    ↓
Copilot Recommendations Display:

1. KNOWLEDGE ARTICLE (High Confidence)
   ├── "Password Reset Troubleshooting"
   ├── Relevance: 94%
   └── Steps: 5-7 minute resolution

2. SENTIMENT GUIDANCE (Urgent)
   ├── Suggest: Apologize for inconvenience
   ├── Tone: Empathetic
   └── De-escalation: Acknowledge frustration

3. NEXT ACTION (Suggested)
   ├── Offer: Manual password reset
   ├── Escalation: If still unresolved
   └── Followup: Offer premium support

4. SCRIPT SUGGESTION (Optional)
   ├── "I completely understand how frustrating that is..."
   ├── "Let me walk you through the fastest solution..."
   └── "If this doesn't work, I'll reset it for you personally"
    ↓
Agent Applies Recommendations
    ↓
Customer Issue Resolved
    ↓
System Captures Feedback
```

---

## Copilot Interface Components
```
Agent Desktop View:

┌─────────────────────────────────────────┐
│  Current Interaction                     │
│  Customer: John Smith                    │
│  Queue: Technical Support                │
│  Duration: 3:45                          │
├─────────────────────────────────────────┤
│  COPILOT RECOMMENDATIONS                 │
├─────────────────────────────────────────┤
│                                          │
│  📚 KNOWLEDGE ARTICLES                   │
│  ├─ Password Reset Guide (94% match)    │
│  ├─ Two-Factor Auth Setup (87% match)   │
│  └─ Account Recovery (76% match)        │
│                                          │
│  😟 SENTIMENT ALERT                      │
│  ├─ Customer: FRUSTRATED                │
│  └─ Suggestion: De-escalate & empathize│
│                                          │
│  ➡️  NEXT ACTIONS                        │
│  ├─ Offer manual password reset         │
│  ├─ Provide security questions          │
│  └─ Escalate if unsuccessful            │
│                                          │
│  💬 SUGGESTED SCRIPT                     │
│  "I understand how frustrating this is. │
│   Let me walk you through the quickest  │
│   way to get this resolved..."          │
│                                          │
│  [👍 Helpful] [👎 Not Helpful]           │
└─────────────────────────────────────────┘
```

---

## Real Flow Scenario: Sales Queue with Copilot
```
Agent: "Hi, thanks for calling. How can I help?"

Customer: "I'm interested in upgrading my plan"

Copilot Recommendations Appear:

1. KNOWLEDGE - Sales Playbook
   ├─ Current Plan Analysis
   ├─ Available Upgrades
   └─ Pricing Information

2. PRODUCT RECOMMENDATIONS
   ├─ Upsell: Premium Plan (+40% revenue potential)
   ├─ Cross-sell: Support Package
   └─ Offer: 2-month discount if upgrading today

3. NEXT ACTION SUGGESTIONS
   ├─ Qualify: Ask about usage patterns
   ├─ Present: Show cost-benefit analysis
   └─ Close: Offer contract details

4. SCRIPT SUGGESTION
   "Based on your usage, the Premium Plan
    would save you money and give you these
    benefits: [list]. Can I set that up?"
    ↓
Agent Applies Script & Recommendations
    ↓
Customer Upgrades (upsell successful)
    ↓
System Logs: Agent applied recommendation
    ↓
Next Sale: System learns and improves recommendations
```

---

## Real Flow Scenario: Support Queue with Sentiment
```
Customer Calls (Angry Tone)

Agent Answers

Copilot Immediately Detects:
├─ Sentiment: NEGATIVE (85% confidence)
├─ Emotion: Frustrated/Angry
├─ Risk: Potential churn
└─ Recommended Action: De-escalate NOW
    ↓
Sentiment Alert in Copilot:
"Customer is frustrated. 
 Suggested response:
 'I'm sorry you're experiencing this issue.
  I'm going to personally make sure we get
  this resolved for you right now.'"
    ↓
Copilot Provides Knowledge:
├─ Issue Resolution Articles
├─ Escalation Path (if needed)
└─ Retention Options
    ↓
Agent Applies De-escalation Approach
    ↓
Customer Sentiment Improves
    ↓
System Logs Improvement
    ↓
Issue Resolved Successfully
```

---

## Omnichannel Copilot Support

### Voice Interactions
```
Real-time Copilot assistance during calls
├── Conversation transcription
├── Intent analysis
├── Real-time knowledge suggestions
└── Sentiment monitoring
```

### Chat Interactions
```
Suggested responses during chat
├── Pre-written messages
├── Quick knowledge links
├── Canned responses with personalization
└── Sentiment-based guidance
```

### Email Interactions
```
Copilot assistance with draft responses
├── Knowledge recommendations
├── Tone suggestions
├── Template recommendations
└── Compliance checking
```

### Social Media Interactions
```
Real-time assistance for public responses
├── Tone and brand consistency
├── Knowledge suggestions
├── De-escalation for negative sentiment
└── Escalation recommendations
```

---

## Usage Scenarios

| Scenario | Solution | Outcome |
|---|---|---|
| High call volume with complex issues | Copilot suggests knowledge articles | Reduced AHT, faster resolution |
| New agents lacking experience | Real-time script and guidance | Improved FCR, faster ramp-up |
| Compliance-heavy calls | Compliance reminders and scripts | Reduced risk, better compliance |
| Frustrated customers | Sentiment analysis with de-escalation tips | Improved satisfaction, retention |
| Sales team underperforming | Upsell/cross-sell recommendations | Increased revenue per interaction |
| Quality issues with call handling | Real-time coaching suggestions | Improved quality scores |
| Agent knowledge gaps | Targeted knowledge recommendations | Improved FCR, fewer escalations |
| Multilingual support | Language-specific scripts and guidance | Consistent quality across languages |

---

## Knowledge Base Organization for Copilot
```
Knowledge Base Structure:

├─ TECHNICAL SUPPORT
│  ├─ Password Reset
│  │  ├─ Steps 1-5
│  │  ├─ Common Issues
│  │  └─ When to Escalate
│  ├─ Two-Factor Auth
│  ├─ Account Recovery
│  └─ Billing Issues
│
├─ SALES
│  ├─ Plan Comparison
│  ├─ Pricing Information
│  ├─ Promotional Offers
│  ├─ Upsell Scenarios
│  └─ Contract Terms
│
├─ CUSTOMER SUCCESS
│  ├─ Onboarding Steps
│  ├─ Best Practices
│  ├─ Feature Usage
│  └─ Integration Guides
│
└─ COMPLIANCE
   ├─ Required Disclosures
   ├─ Privacy Policies
   ├─ Legal Requirements
   └─ Prohibited Actions
```

---

## Copilot Performance Metrics

### Recommendation Quality
| Metric | Target | Purpose |
|---|---|---|
| Recommendation Accuracy | >85% | Agent finds recommendations helpful |
| Agent Acceptance Rate | >60% | Agents actively use suggestions |
| Relevance Score | >4/5 | Recommendations match context |
| Time to Resolution | -15% | Faster with Copilot assistance |
| Knowledge Article Match Rate | >90% | Correct article suggested |

### Agent Performance
| Metric | Target | Purpose |
|---|---|---|
| First Contact Resolution | +10-25% | Copilot guidance improves outcomes |
| Average Handle Time | -10-20% | Faster resolution with suggestions |
| Customer Satisfaction | +5-15% | Better agent performance |
| Quality Score | +5-10% | Improved call quality |
| Agent Confidence | +20-30% | Subjective improvement |

---

## Best Practices

### Knowledge Base Optimization
- **Keep articles current** - Update quarterly or when processes change
- **Write for clarity** - Use simple language agents can understand quickly
- **Include visuals** - Screenshots and diagrams help comprehension
- **Provide examples** - Real scenarios help agents apply knowledge
- **Tag thoroughly** - Use keywords and metadata for better matching
- **Test accuracy** - Verify all knowledge is correct before publishing

### Copilot Configuration
- **Start simple** - Begin with high-confidence recommendations only
- **Tune relevance** - Adjust thresholds based on agent feedback
- **Monitor adoption** - Track which recommendations agents use
- **Gather feedback** - Ask agents what would help them more
- **Iterate quickly** - Update knowledge and rules frequently
- **A/B test** - Try different recommendation approaches

### Agent Enablement
- **Provide training** - Agents need to understand Copilot features
- **Share success stories** - Show how other agents use it effectively
- **Encourage experimentation** - Let agents find what works for them
- **Make feedback easy** - Simple thumbs up/down for recommendations
- **Celebrate improvements** - Recognize agents who adopt well
- **Continuous learning** - Regular coaching on Copilot usage

### Monitoring & Optimization
- **Track recommendations** - See which articles agents use most
- **Monitor accuracy** - Ensure recommendations are helpful
- **Gather sentiment** - Ask agents about Copilot effectiveness
- **Review metrics** - Check impact on FCR, AHT, CSAT
- **Optimize content** - Update articles agents find unhelpful
- **Plan improvements** - Use data to guide enhancements

---

## Common Implementation Scenarios

### Scenario 1: Technical Support with Knowledge-Heavy Topics
```
Configuration:
├── Knowledge base with 200+ articles
├── Intent-based recommendation
├── Sentiment monitoring enabled
├── De-escalation scripts
└── Escalation pathways

Expected Results:
├── FCR improvement: 15-25%
├── AHT reduction: 12-18%
└── Agent confidence: +25%
```

### Scenario 2: Sales Team with Upsell Focus
```
Configuration:
├── Product recommendation engine
├── Upsell/cross-sell playbooks
├── Customer history integration
├── Pricing recommendations
└── Contract template suggestions

Expected Results:
├── Revenue per call: +15-30%
├── Sales conversion: +10-20%
└── Agent productivity: +20%
```

### Scenario 3: Multilingual Support
```
Configuration:
├── Knowledge base in multiple languages
├── Language-specific scripts
├── Tone guidance per language
├── Cultural sensitivity prompts
└── Translation recommendations

Expected Results:
├── FCR consistent across languages
├── Quality standardization
└── Customer satisfaction: +10-15%
```

---

## Troubleshooting Guide

| Issue | Cause | Resolution |
|---|---|---|
| No recommendations appearing | Knowledge base empty or not indexed | Populate knowledge base and index |
| Irrelevant recommendations | Poor knowledge article tagging | Review and improve metadata/keywords |
| Agents ignoring recommendations | Not helpful or slow to appear | Adjust relevance thresholds and review content |
| Slow recommendation loading | Too many articles to search | Add more specific keywords and metadata |
| Sentiment detection inaccurate | Model needs more training data | Collect more interactions and retrain |
| High false positive sentiment | Threshold too sensitive | Adjust sensitivity settings lower |
| Copilot not working for all queues | Not enabled for specific queues | Enable in queue configuration |
| Knowledge articles outdated | No review process | Establish content review cycle |
| Low agent adoption | Agents don't understand value | Provide additional training |
| Module not appearing | License not purchased or enabled | Verify Premium edition and module purchase |

---

## Agent Copilot vs. Traditional Knowledge Base

| Feature | Agent Copilot | Traditional Knowledge |
|---|---|---|
| Real-time suggestions | Yes, automatic | Manual search required |
| Context awareness | AI-powered, contextual | Search-based only |
| Sentiment analysis | Yes | No |
| Next action prediction | Yes | No |
| Script guidance | Automatic | Manual lookup |
| Learning capability | Improves over time | Static |
| Setup complexity | Medium-High | Low |
| Ongoing maintenance | High (ML tuning) | Medium |
| Agent productivity | +20-30% potential | Baseline |
| Customer satisfaction | +5-15% improvement | Baseline |

---

## Sentiment Analysis in Copilot

### Sentiment Detection Levels
```
Extremely Negative (-2)
├─ Angry, frustrated, hostile
├─ Risk: High churn likelihood
└─ Action: Immediate de-escalation

Negative (-1)
├─ Disappointed, concerned
├─ Risk: Medium churn likelihood
└─ Action: Empathy + quick resolution

Neutral (0)
├─ Standard interaction tone
├─ Risk: Low
└─ Action: Normal service

Positive (+1)
├─ Satisfied, pleased
├─ Risk: None
└─ Action: Reinforce positive experience

Extremely Positive (+2)
├─ Happy, delighted
├─ Opportunity: Upsell/cross-sell
└─ Action: Leverage positive sentiment
```

---

## Integration Scenarios

### With Workforce Optimization
```
Copilot + WFO
├── Agent assisted by Copilot
├── Interaction recorded
├── Quality scored with AI
├── Coaching recommendations generated
└── Coaching delivered back to agent
```

### With Predictive Routing
```
Copilot + Predictive Routing
├── Best agent routed to contact
├── Copilot assists during interaction
├── Recommendations improve outcome
└── System learns for future routing
```

### With Analytics
```
Copilot + Analytics
├── Copilot recommendation usage tracked
├── Impact on metrics measured
├── Dashboards show Copilot effectiveness
└── Data guides optimization
```

---

## Interview Cheat Sheet

| Question | Answer |
|---|---|
| What is Agent Copilot? | AI-powered real-time guidance system for agents |
| Also known as? | Agent Assist or Copilot Assistant |
| What are the requirements? | Premium edition + Customer Insights module |
| What does it recommend? | Knowledge articles, scripts, next actions, sentiment guidance |
| How does sentiment analysis help? | Detects customer frustration and suggests de-escalation |
| Where is it configured? | Admin → Architect → Agent Copilot |
| What channels does it support? | Voice, chat, email, social media |
| How does it improve performance? | FCR +10-25%, AHT -10-20%, CSAT +5-15% |
| What's required for success? | Quality knowledge base and agent training |
| How does machine learning help? | System learns from recommendations agents use vs. ignore |
| Can agents reject recommendations? | Yes, agents decide what to apply |
| How long until ROI? | 4-8 weeks to see significant improvement |
| What if knowledge base is empty? | Copilot won't have content to recommend |
| How does it work with omnichannel? | Provides channel-specific guidance (voice, chat, email, etc.) |
| What's the biggest success factor? | Quality, current, well-organized knowledge base |

---

## Key Takeaways

- **Real-Time AI Assistance** - Copilot provides in-the-moment guidance during interactions
- **Knowledge-Driven** - Quality depends on knowledge base content and organization
- **Sentiment Awareness** - Monitors emotion and suggests appropriate responses
- **Omnichannel Support** - Works across voice, chat, email, and social channels
- **Premium Feature** - Requires Premium edition and Customer Insights module
- **Significant Impact** - FCR improvements of 10-25% typical
- **Machine Learning** - System improves recommendations based on agent actions
- **Agent Adoption Critical** - Success depends on agents trusting and using recommendations
- **Ongoing Content Management** - Knowledge base requires regular updates
- **Quick ROI** - 4-8 weeks to measurable improvements

---

## Migration Path from Manual Knowledge Search

### Phase 1: Assessment (Weeks 1-2)
```
├── Audit current knowledge base
├── Identify content gaps
├── Plan reorganization
└── Set quality standards
```

### Phase 2: Content Preparation (Weeks 2-4)
```
├── Create/update knowledge articles
├── Add metadata and keywords
├── Organize by intent and queue
└── Quality review all content
```

### Phase 3: Copilot Setup (Weeks 4-5)
```
├── Enable Agent Copilot
├── Configure recommendations
├── Set relevance thresholds
└── Establish monitoring
```

### Phase 4: Agent Training (Week 5-6)
```
├── Educate on Copilot features
├── Practice with sample interactions
├── Explain recommendation types
└── Share best practices
```

### Phase 5: Pilot & Optimization (Weeks 6-8)
```
├── Deploy to single queue
├── Monitor and gather feedback
├── Optimize recommendations
└── Plan full rollout
```

### Phase 6: Full Deployment (Week 8+)
```
├── Enable across all queues
├── Provide ongoing support
├── Monitor metrics
└── Continuous improvement
```

---

## Additional Resources

### Official Documentation Links
- Genesys Cloud Agent Copilot Guide: https://help.genesys.com/genesyscloud/current/en-us/AgentCopilot.html
- Knowledge Management Setup: https://help.genesys.com/genesyscloud/current/en-us/KnowledgeManagement.html
- Customer Insights Module: https://help.genesys.com/genesyscloud/current/en-us/CustomerInsights.html

### Support Contacts
- Genesys Sales: sales@genesys.com
- Genesys Support: https://support.genesys.com
- Community Forums: https://community.genesys.com

---

## Document Version Info
**Last Updated:** March 2026  
**Source:** Genesys PureCloud Official Documentation  
**Version:** 1.0