Events - Kore.ai Docs

Events fire when specific actions occur during a conversation or when conditions are met on active channels. You configure what the AI Agent does when each event fires. Go to Automation > Conversation Management > Events.

Event Actions

When an event fires, the AI Agent executes one of these actions:

Action	Description
Initiate a Task	Launches a selected Dialog task (Standard or Hidden). If the dialog isn’t published, the AI Agent shows an error. Preview errors using Debug mode.
Run a Script	Runs JavaScript using session, context objects, variables, and functions. Supports Debug mode.
Show a Message	Displays a simple or advanced message with full formatting and channel override support.

Multilingual behavior for Show a Message:

Messages can be defined per language per event.
Deleting a message for one language removes it from all languages.
Adding a message for one language copies it to all other languages.
Editing a message in one language affects only that language.

Event Types

Events fall into three categories: Intent, Conversation, and Sentiment.

Event	Category	Trigger
Intent not Identified	Intent	AI Agent can’t identify user intent.
Ambiguous Intents Identified	Intent	Two or more intents identified with similar confidence scores, or only low-confidence Knowledge Graph intents are found.
End of Task	Conversation	A dialog task or FAQ completes.
Task Execution Failure	Conversation	Error occurs during task execution (service call failure, unreachable server, agent transfer error, webhook failure, etc.).
RCS Opt-in	Conversation	User opts in to the RCS Messaging channel.
RCS Opt-out	Conversation	User opts out of the RCS Messaging channel.
Repeat Bot Response	Conversation	User requests the AI Agent to repeat its last response (primarily for voice/IVR).
Sentiment Events	Sentiment	User emotion matches configured tone criteria. Click + New Event to define custom sentiment events.

Intent Events

Intent not Identified

Fires when the AI Agent can’t understand the user’s intent. Configure one of:

Show a standard message from the standard responses library. Learn more.
Automatically run a dialog task — select the task from the dropdown.

Ambiguous Intents Identified

Fires when the NLP engine detects multiple intents with similar confidence. By default, the AI Agent presents the list of ambiguous intents to the user for selection. Conditions that trigger this event:

Two or more “definite” intents are identified.
Two or more “possible” intents are identified within the proximity threshold.
Only low-confidence Knowledge Graph intents are found and no other engine identifies an intent.

You can override default behavior with a custom Dialog Task that calls koreUtil.getAmbiguousIntents() to retrieve intents and confidence scores. Learn more. To configure:

Go to Conversation Intelligence > Events > Intent Events.
Click Configure to enable the event.

If not enabled, the list of ambiguous intents is shown to the user.

Select a configuration option:
- Present all qualified intents to the end-user for disambiguation — Default behavior. Learn more
- Automatically run a Dialog Task — Select a dialog with custom disambiguation logic. Learn more
Click Save & Enable.

When ambiguous intents are detected during testing, the Debug Log shows: Multiple intents are identified — Ambiguous Intents Identified event is initiated.

Interruption behavior: When this event is enabled, interruption follows node-, dialog-, or app-level settings — except that if Continue the current task and add a new task to the follow-up task list is selected, the interrupting task isn’t added to the follow-up list. Learn more

Variable Namespaces

Go to More Options > Manage Variable Namespaces at the top of the Events screen to associate variable namespaces with events. This option appears only when Variable Namespace is enabled for the AI Agent. Learn more.

Conversation Events

End of Task

Fires when the AI Agent is no longer expected to send or receive messages. The context is updated with the end reason and the completed task name (FAQs use FAQ as the task name). Client-side implementations (BotKits, RTM, Webhook channels) can use the end-of-task reason flag to drive follow-up actions.

Scenario	End of Task Flag
Reached the last node of the dialog	`Fulfilled`
Task canceled by the user	`Canceled`
Error in task or FAQ execution (no Task Failure Event, no hold tasks)	`Failed`
Linked dialog completed without returning to parent	`Fulfilled_LinkedDialog`
FAQ answered successfully	`Fulfilled`
Event executed via Run Script or Show Message (no tasks on hold)	`Fulfilled_Event`
Error in executing an event via Run Script or Show Message (no tasks on hold)	`Failed_Event`
User declines to resume an on-hold task (no other task on hold)	`Canceled`

Task Execution Failure

Fires when an error occurs during dialog task execution, including:

AI Agent execution errors
Service call failures or unreachable servers
Agent transfer node errors
Knowledge Graph task failures
Webhook node failures
Unavailable sub-dialog
Errors parsing the AI Agent message

Defaults: Always enabled with the Show Message action. Can’t be disabled. This app-level behavior can be overridden per task from the dialog task settings. Learn more

RCS Opt-In / Opt-Out

Fires when a user opts in to or opts out of the RCS Messaging channel. Configure a follow-up response to confirm the user’s action.

Repeat Bot Response

Fires when specific utterances (predefined or custom-trained) request the AI Agent to repeat its last response. Applies to voice channels: IVR, Audiocodes, Twilio Voice, and SmartAssist Gateway.

The Repeat Bot Response event uses the NLU multilingual model for non-English languages.

Priority behavior:

Conflict Scenario	Behavior
Intent vs. Repeat event	Intent takes priority.
Sub-intent vs. Repeat event	Sub-intent takes priority.
FAQ vs. Repeat event	FAQ takes priority.
Group node sub-intent vs. Repeat event	Group node sub-intent takes priority.
Mid-conversation (no conflict)	Repeat event triggers.

Example: In a Flight Booking AI Agent, a user on IVR says: “Sorry, I can’t hear you. Can you please repeat it again?” after a booking confirmation. If the session is active and the End of Task event has a follow-up message (for example, “Is there anything else I can help you with?”), the Repeat Bot Response event repeats both the confirmation and the follow-up together. To configure:

Go to Conversation Intelligence > Events.
Click Repeat Bot Response Event to configure it.
Click Manage Utterance to review or add pre-trained utterances.
Add utterances as needed, then click Train.
After training, configure preconditions:
- Channels — Add voice channels (IVR, IVR Audiocodes, Twilio Voice, SmartAssist Gateway).
- Context Tags — Add context objects to scope the event to specific dialog tasks. Learn more.
Set the Event Configuration:
- Repeat Only Last Bot Response (default) — Plays a filler message before repeating. Default filler: “Sure, I will repeat it for you.” Click + Add Filler Message, select the IVR channel, enter the filler text, and click Done.
- Auto-generate Response — Uses LLM and Generative AI to generate the repeated response. Requires the Advanced NLU model. Click Enable Now to activate.
  When Auto-generate is enabled, filler messages aren’t used.
Expand Advanced Settings and configure:
- Repeat Attempts Limit — Number of retry attempts (1-10, default: 5).
- Behavior on Exceeding Repeat Attempts — Choose End of Dialog or Initiate Dialog (select which task to redirect to).
Click Save & Enable.

Sentiment Events

Sentiment Events detect a user’s emotional state and trigger specific behaviors — such as transferring to a live agent when a user is frustrated. The NLP interpreter analyzes user utterances for tone and stores scores in the context object. Use these scores in conditional transition statements to drive dialog flow.

Introduced in v7.0. Not supported in all languages. Learn more.

Configure a Sentiment Event

Go to Conversation Intelligence > Events > Sentiment Events. Click + New Event.

Parameter	Description
Event Name	Unique name for the event.
Event Trigger	When to check sentiment: On Task Completion or After Every User Utterance.
Emotion Configurations	Select one or more emotions: anger, disgust, fear, sadness, joy, positive. For each, choose Session or Message scope and define a score range (-3 to +3).

Session-level tone — Aggregated across all messages in a session.
Message-level tone — Calculated per individual user message.
When multiple emotions are selected, all conditions must be met. To trigger on any single emotion, create separate events.

Sentiment Event Flow

Tone scores update with every user message, and sentiment events are continuously evaluated. When an event’s conditions are met:

Initiate a Task — The current task is discarded and the AI Agent switches to the configured dialog.
- Other implicitly paused tasks are also discarded.
- Tasks on hold via Hold and Resume settings are resumed per those settings.
- If the dialog is unavailable, a standard response is shown.
Run a Script — The script runs and task execution continues. Script errors display a standard error response.

Priority rules:

Sentiment events take precedence over direct intent invocation.
When multiple sentiment events match simultaneously, the highest-precedence event (by defined order) wins.

Behavior when running a script:

Scenario	Behavior
Small Talk detected	Sentiment executes first, then small talk.
Dialog detected	Sentiment executes first, then dialog.
Fallback flow	Sentiment executes first, then fallback logic.
Entity Node	Sentiment executes first, then error prompt is shown.
Confirmation Node	Sentiment executes first, then confirmation prompt.
On-Intent Message Node	Sentiment executes first, then moves to the else condition.

Behavior when initiating a task:

Scenario	Behavior
Small Talk detected	Sentiment executes first, then small talk.
Dialog detected	Sentiment executes first, then the configured dialog runs to completion.
Fallback flow	Sentiment executes first, then configured dialog runs.
Entity Node	Sentiment executes first, then connected dialog runs, ending both dialogs.
Confirmation Node	Sentiment executes first, then connected dialog runs, ending both dialogs.
On-Intent Message Node	Sentiment executes first, then connected dialog runs, ending both dialogs.

Reset Tone

Sentiment values reset at two points:

Start of every user conversation session (default).
After a sentiment event fires:
- If a script ran: values reset after successful execution.
- If a dialog task was triggered: values transfer to the new dialog’s context and reset in the original global context.

Tone Analysis

The NLP interpreter scores user utterances on six emotions. Access scores from the Context object or use them to configure sentiment events.

Tone Types

Tone	Description
`angry`	Anger detected.
`disgust`	Disgust detected.
`fear`	Fear detected.
`sad`	Sadness detected.
`joy`	Joy detected.
`positive`	General positivity of the utterance.

Tones aren’t mutually exclusive — an utterance can score high for fear and mild for sadness simultaneously.

Tone Score Scale

Scores range from -3 to +3:

Score	Meaning
+3	User definitely expressed the tone.
+2	User expressed the tone.
+1	User likely expressed the tone.
0	Neutral — tone not expressed or suppressed.
-1	User likely suppressed the tone.
-2	User suppressed the tone.
-3	User definitely suppressed the tone.

Positive values = tone explicitly expressed. Negative values = tone explicitly negated. Examples:

“I am happy about this news” → positive joy score
“I am not happy about this news” → negative joy score

Score Calculation

Scores are calculated from the base tone value plus any modifiers (adverbs or adjectives that amplify or reduce intensity). Examples:

“I am extremely disappointed” → higher angry score than “I am disappointed”
“I am not disappointed” → negative angry score

The tone analyzer compiles base tones per emotion and calculates:

Current node score → stored in message_tone
Session average score → stored in dialog_tone (reset at end of each session)

Context Object Variables

Variable	Scope	Description
`message_tone`	Current node	Array of tone emotions and scores for the current dialog node.
`dialog_tone`	Session	Array of average tone emotions and scores for the entire session.

Both variables return arrays of objects with tone_name, count, and level fields. Key/value pairs are returned only when a tone is detected. A level of 0 indicates a neutral tone. Always handle positive, negative, zero, and undefined values when reading tone variables.

Examples

message_tone
tone_name: positive, level: 2
tone_name: disgust, level: -2
tone_name: angry, level: -2

dialog_tone
tone_name: angry, level: -3
tone_name: sad, level: -3
tone_name: positive, level: 3
tone_name: joy, level: 3

Sentence: “I don’t think that this is a good idea and I am not happy with how it came out, especially because of your attitude.”

"dialog_tone": [
  { "tone_name": "joy", "count": 1, "level": 0.67 },
  { "tone_name": "sad", "count": 1, "level": 0.5 },
  { "tone_name": "angry", "count": 1, "level": 0.5 }
]

Sentence: “This is a great idea! I’m super excited already.”

"dialog_tone": [
  { "tone_name": "joy", "count": 1, "level": 3 },
  { "tone_name": "sad", "count": 1, "level": 2.8 },
  { "tone_name": "angry", "count": 1, "level": -3 }
]

Sentence: “This was a funny and casually well-written book, a good read. But it’s a little frustrating because it abandons the narrative without finishing it.”

"dialog_tone": [
  { "tone_name": "joy", "count": 1, "level": 1.5 },
  { "tone_name": "sad", "count": 1, "level": -1.5 },
  { "tone_name": "angry", "count": 1, "level": -1 }
]

Using tone scores in transition conditions:

if context.message_tone.angry > 2.0
    then goTo liveAgent

For more information, see Context Object.

Extending Tone Detection with Custom Words

Add custom words to tone concepts during concept training to extend the emotion vocabulary. Concept name syntax: ~tone-<tonename>-<level>

<tonename> — one of the 6 tone types listed above.
<level> — a number from 1 to 7, where 1 = -3, 4 = 0 (neutral), 7 = +3.

Only add base tone words. Intensity modifiers like very or extremely are handled automatically.

Examples:

Word	Concept	Meaning
freaking	`~tone-angry-7`	Very strong anger (+3)
Yikes!	`~tone-angry-5`	Mild anger (+1)
please	`~tone-angry-4`	Neutral tone (0)
Thanks!	`~tone-angry-1`	Not angry at all (-3)

​Event Actions

​Event Types

​Intent Events

​Intent not Identified

​Ambiguous Intents Identified

​Variable Namespaces

​Conversation Events

​End of Task

​Task Execution Failure

​RCS Opt-In / Opt-Out

​Repeat Bot Response

​Sentiment Events

​Configure a Sentiment Event

​Sentiment Event Flow

​Reset Tone

​Tone Analysis

​Tone Types

​Tone Score Scale

​Score Calculation

​Context Object Variables

​Examples

​Extending Tone Detection with Custom Words

Event Actions

Event Types

Intent Events

Intent not Identified

Ambiguous Intents Identified

Variable Namespaces

Conversation Events

End of Task

Task Execution Failure

RCS Opt-In / Opt-Out

Repeat Bot Response

Sentiment Events

Configure a Sentiment Event

Sentiment Event Flow

Reset Tone

Tone Analysis

Tone Types

Tone Score Scale

Score Calculation

Context Object Variables

Examples

Extending Tone Detection with Custom Words