> ## Documentation Index
> Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent AI

Intelligent AI assistant for contact center agents to boost productivity.

***

## Overview

Agent AI provides real-time assistance to human agents:

* Suggested responses based on conversation context.
* Automated call and chat summaries.
* Knowledge base search integration.

***

## How It Works

<div class="ascii-art">
  ┌─────────────────────────────────────────────────────────────────┐
  │                    Live Conversation                            │
  │                                                                 │
  │       Customer: "I received the wrong item in my order          │
  └─────────────────────────────────┬───────────────────────────────┘
  │                                 ▼
  ┌─────────────────────────────────────────────────────────────────┐
  │                       Agent AI Engine                           │
  │                                                                 │
  │       ┌─────────────┐  ┌─────────────┐  ┌─────────────┐         │
  │       │  Context    │  │  Knowledge  │  │  Response   │         │
  │       │  Analysis   │  │  Search     │  │  Generation │         │
  │       └─────────────┘  └─────────────┘  └─────────────┘         │
  └─────────────────────────────────┬───────────────────────────────┘
  │                                 ▼
  ┌─────────────────────────────────────────────────────────────────┐
  │                    Agent Desktop Widget                         │
  │                                                                 │
  │  Suggested Response:                                            │
  │  "I apologize for the inconvenience. I can help you with a      │
  │   replacement or refund. Let me pull up your order details."    │
  │                                                                 │
  │  Relevant Knowledge:                                            │
  │  • Wrong Item Policy (confidence: 95%)                          │
  │  • Return Process Guide (confidence: 87%)                       │
  │                                                                 │
  │  Recommended Actions:                                           │
  │  \[Create Return] \[Issue Refund] \[Escalate to Supervisor]        │
  └─────────────────────────────────────────────────────────────────┘
</div>

***

## Features

### Real-Time Suggestions

Contextual response suggestions during conversations:

| Feature              | Description                              |
| -------------------- | ---------------------------------------- |
| **Auto-suggest**     | Suggestions appear as customer speaks    |
| **One-click use**    | Insert suggestion with single click      |
| **Edit before send** | Modify suggestions as needed             |
| **Learn from edits** | System improves from agent modifications |

Configuration example:

```yaml theme={null}
Suggestions:
  trigger: continuous
  display_count: 3
  confidence_threshold: 0.7
  sources:
    - knowledge_base
    - canned_responses
    - conversation_history
  tone: professional
```

### Automated Summaries

Generate conversation summaries automatically:

| Summary Type    | When Generated                      |
| --------------- | ----------------------------------- |
| **Real-time**   | Updated as conversation progresses  |
| **After-call**  | Complete summary at interaction end |
| **Disposition** | Structured outcome summary          |

Summary configuration example:

```yaml theme={null}
Auto-Summary:
  format: structured
  sections:
    - customer_issue
    - resolution
    - follow_up_required
  length: concise
  auto_save_to_crm: true
```

### Knowledge Assistance

Integrated knowledge search:

* Search triggered automatically by conversation context.
* Manual search with natural language queries.
* Results ranked by relevance.
* Source citations included.

***

## Configuration

### Enable Agent AI

1. Navigate to **Agent AI** → **Configuration**.
2. Enable Agent AI for desired queues.
3. Configure feature settings.
4. Test with pilot agents.

### Feature Settings

| Setting              | Options                         |
| -------------------- | ------------------------------- |
| **Suggestion mode**  | Continuous, On-demand, Disabled |
| **Auto-summary**     | Enabled, Disabled               |
| **Knowledge search** | Auto, Manual, Both              |
| **NBA**              | Enabled, Disabled               |

### Integration

Connect Agent AI to:

* **Search AI** — For knowledge retrieval.
* **CRM** — For customer context.
* **Case management** — For action execution.
* **Quality AI** — For coaching feedback.

***

## Agent Experience

### Desktop Widget

Agent AI appears as a widget in the agent console:

<div class="ascii-art">
  ┌─────────────────────────────────────┐
  │  Agent AI                    \[─] \[×]│
  ├─────────────────────────────────────┤
  │                                     │
  │  Suggested Response                 │
  │ ┌─────────────────────────────────┐ │
  │ │ I understand your concern about │ │
  │ │ the billing charge. Let me      │ │
  │ │ address them.                   │ │
  │ └─────────────────────────────────┘ │
  │  \[Use] \[Copy] \[👍] \[👎]              │
  │                                     │
  │  Knowledge                          │
  │  • Billing FAQ (95%)                │
  │  • Refund Policy (88%)              │
  │  \[Search manually]                  │
  │                                     │
  │  Quick Actions                      │
  │  \[Refund] \[Credit] \[Escalate]       │
  │                                     │
  └─────────────────────────────────────┘
</div>

### Feedback Loop

Agents can rate suggestions:

* **Thumbs up** — Good suggestion.
* **Thumbs down** — Unhelpful suggestion.
* **Edits** — System learns from modifications.

Feedback improves suggestion quality over time.

***

## Analytics

### Agent AI Metrics

| Metric                    | Description               |
| ------------------------- | ------------------------- |
| **Suggestion acceptance** | % of suggestions used     |
| **Time saved**            | Handle time reduction     |
| **Knowledge utilization** | Searches and clicks       |
| **NBA conversion**        | Recommended actions taken |

### Quality Impact

Track quality improvements:

* First contact resolution rate.
* Customer satisfaction scores.
* Quality evaluation scores.
* Handle time trends.

***

## Best Practices

### Deployment

1. Start with a pilot group of agents.
2. Gather feedback and iterate.
3. Roll out gradually by queue/team.
4. Monitor adoption and adjust.

### Knowledge Quality

* Keep knowledge base current.
* Remove outdated content.
* Add content for common queries.
* Monitor search failures.

### Agent Training

* Introduce Agent AI in agent training.
* Explain feedback mechanism importance.
* Show how to use suggestions effectively.
* Address concerns about AI assistance.

***