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

# Manage Interruptions

<Badge icon="arrow-left" color="gray">[Back to NLP Topics](/ai-for-service/automation/natural-language/nlp-topics)</Badge>

Users rarely follow a linear conversation path. They may ask an unrelated question mid-task, provide a new intent before finishing the current one, or switch topics entirely. The **Manage Interruptions** feature gives you granular control over how your AI Agent handles these situations.

You can configure interruption behavior at three levels, with the more specific level taking precedence.

***

## Interruptions Hierarchy

| Level           | Precedence       | Where to Configure                                                             |
| --------------- | ---------------- | ------------------------------------------------------------------------------ |
| **Node**        | Highest          | Node **Instance** tab > **Advanced Controls** > **Interruptions Behavior**     |
| **Dialog Task** | Middle           | Dialog Builder > **more options** icon > **Manage Interruptions**              |
| **App**         | Lowest (default) | **Conversation Intelligence > Conversation Management > Manage Interruptions** |

> If no node or task-level settings exist, the app-level settings apply.

***

## App-Level Interruption Settings

Go to **Conversation Intelligence > Conversation Management > Manage Interruptions**.

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/KUEkqyD1q5ta9F2D/ai-for-service/automation/intelligence/images/bot-level-settings-manageinterruptions-window-1.png?fit=max&auto=format&n=KUEkqyD1q5ta9F2D&q=85&s=2760683ea277841f5eea39d95c9e4caf" alt="bot level settings manage interruptions" width="664" height="854" data-path="ai-for-service/automation/intelligence/images/bot-level-settings-manageinterruptions-window-1.png" />

<Note>When a user provides input after 15+ minutes of inactivity and the input is within the previous session context, it is treated as an interruption. If the user starts a new utterance, the old context is discarded.</Note>

### Allow Interruptions

| Option                                                            | Behavior                                                                                                                |
| ----------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| **Hold the current task and resume after the new task completes** | Switches to the new intent immediately. Requires a Resume Option (see below).                                           |
| **Discard the current task and switch to new**                    | Discards the current task and switches. Sends the user a notification message. Customizable via **Manage Response**.    |
| **Switch to new task without notification and discard current**   | Silently discards the current task and switches. No user notification.                                                  |
| **Continue the current task and add new task to follow-up list**  | Stays on the current task and adds the new intent to the `FollowupIntents` array. Customizable via **Manage Response**. |

### Do Not Allow Interruptions

Turns off interruptions at the app level. Can still be overridden at the task or node level.

### Allow the End User to Select the Behavior

Prompts the user to confirm whether to switch tasks. Customize the confirmation message via **Manage Response**. Requires a Resume Option.

***

## Resume Options

Resume options define what happens to the on-hold task after the interrupting task completes.

| Option                                                                   | Behavior                                                                                                                       |
| ------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------ |
| **Get confirmation before resuming**                                     | AI Agent asks the user (Yes/No) whether to resume the on-hold task. Customizable message.                                      |
| **Notify the user and resume**                                           | Resumes the on-hold task automatically and notifies the user. No confirmation required. Customizable message.                  |
| **Resume without any message**                                           | Directly resumes the on-hold task with no notification to the user.                                                            |
| **Always resume without message if the task ended in a single response** | If the interrupting task completed in a single response, resumes the on-hold task silently regardless of other resume options. |

**Example (Get confirmation before resuming):**

```
User: Can you book me a flight for tomorrow?
AI Agent: From which city are you flying?
User: Los Angeles
AI Agent: Where to?
User: By the way, what's the weather forecast for tomorrow?
AI Agent: Please enter the city for the forecast.
User: Los Angeles
AI Agent: Weather Forecast for Los Angeles — March 15, 25°C, Mostly Sunny
AI Agent: Should I continue with the task 'Book Flight'?
User: Yes
AI Agent: Enter the name of the destination airport.
```

***

## On Hold Quantity

Set the maximum number of tasks that can be held at once. Default: **1**. Range: 1-N.

When the limit is reached, new tasks are ignored regardless of interruption settings.

<Note>Tasks resume in reverse chronological order (most recent first). Limit on-hold tasks to 1-2 to avoid confusion.</Note>

***

## Node-Level Customization

1. Open the dialog task and select the node.
2. Go to the **Instance** tab > **Advanced Controls**.
3. Under **Interruptions Behavior**, select **Customize for this node** and configure settings.

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/DfcaQTTT1JeaWLu_/ai-for-service/automation/intelligence/images/node-level-customization.gif?s=bc8c4fe885af7e3b864d40ce1b4e4b72" alt="node level customization" width="516" height="740" data-path="ai-for-service/automation/intelligence/images/node-level-customization.gif" />

***

## Dialog-Level Customization

1. Open the dialog task.
2. Click the **more options** icon (top-right of Dialog Builder) > **Manage Interruptions**.
3. Under **Interruptions Behavior**, select **Customize for this task** and configure settings.

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/DfcaQTTT1JeaWLu_/ai-for-service/automation/intelligence/images/dialog-level-manage-interruptions.png?fit=max&auto=format&n=DfcaQTTT1JeaWLu_&q=85&s=3a0d9529f36bc4aa69899b7a5e2d9b53" alt="dialog level manage interruptions" width="1024" height="462" data-path="ai-for-service/automation/intelligence/images/dialog-level-manage-interruptions.png" />

***

## Behavior for FAQs

| Option                                                    | Behavior                                                                                |
| --------------------------------------------------------- | --------------------------------------------------------------------------------------- |
| **Respond to FAQs and resume the on-hold task** (default) | AI Agent answers the FAQ and then resumes the on-hold task.                             |
| **Execute the FAQs using Hold & Resume Settings**         | Treats FAQ intents like any other intent and applies configured Hold & Resume settings. |

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/DfcaQTTT1JeaWLu_/ai-for-service/automation/intelligence/images/faq-interruption-behavior-window.png?fit=max&auto=format&n=DfcaQTTT1JeaWLu_&q=85&s=adbcc491518609cf9c48160400d799ed" alt="FAQ interruption behavior" width="905" height="450" data-path="ai-for-service/automation/intelligence/images/faq-interruption-behavior-window.png" />

***

## Behavior for Ambiguous Intents

| Option                                                                                    | Behavior                                                                       |
| ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ |
| **Present ambiguous intents to the end-user and proceed as per Hold and Resume settings** | User selects a task; interruption settings are applied based on the selection. |
| **Do not present ambiguous intents and continue on the current task** (default)           | AI Agent ignores ambiguous intents and continues the current task.             |

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/KUEkqyD1q5ta9F2D/ai-for-service/automation/intelligence/images/ambiguous-interruption-behavior-window.png?fit=max&auto=format&n=KUEkqyD1q5ta9F2D&q=85&s=087dca19b818cbec43a5e8572f7c8f0b" alt="ambiguous interruption behavior" width="884" height="409" data-path="ai-for-service/automation/intelligence/images/ambiguous-interruption-behavior-window.png" />

***

## Behavior for Small Talk

| Option                                                          | Behavior                                                                              |
| --------------------------------------------------------------- | ------------------------------------------------------------------------------------- |
| **Respond to Small Talk and resume the on-hold task** (default) | AI Agent responds to Small Talk by putting the current task on hold, then resumes it. |
| **Execute Small Talk using Hold & Resume settings**             | Applies configured Hold & Resume settings to Small Talk.                              |

<img src="https://mintcdn.com/koreai-agent-management-platform-dev/y0lMHvFSXaTOPJ5Z/ai-for-service/automation/intelligence/images/smalltalk-interruption-behavior-window.png?fit=max&auto=format&n=y0lMHvFSXaTOPJ5Z&q=85&s=5df30ea70cebc3b0c8af564c6eeb672f" alt="small talk interruption behavior" width="896" height="409" data-path="ai-for-service/automation/intelligence/images/smalltalk-interruption-behavior-window.png" />

***

## Behavior for User Authorization

When a user provides unexpected input at an authorization prompt:

1. If **Small Talk** is detected: Small Talk response is shown, and the authorization prompt is repeated.
2. The Platform checks for any intents regardless of interruption settings:
   * **Single intent detected** — Asks the user to confirm discarding the current task and triggering the new one.
   * **Multiple intents detected** — Shows an ambiguity dialog with an option to ignore and continue the current task.
3. If **no intent** is found — User is re-prompted with the authorization link.


Built with [Mintlify](https://mintlify.com).