Conversation Management

This document covers the core conversation management features of the AI Agent: standard responses, interruption handling, small talk, and alert tasks.

Feature	Description
Standard Responses	Pre-defined NLP interpreter messages triggered by events, user inputs, or system conditions.
Manage Interruptions	Configure how the AI Agent handles mid-task interruptions such as FAQs or small talk.
Small Talk	Enable casual, non-task conversations to make interactions feel more natural.
Alert Tasks	Set up proactive notifications the AI Agent sends based on triggers or schedules.

Standard Responses

Standard responses are pre-defined NLP interpreter messages the AI Agent displays based on events, user inputs, or system conditions. All defaults can be customized, and channel-specific overrides can be added. Go to Automation AI > Conversation Intelligence > Conversation Management > Standard Responses. For each category, the trigger condition is listed alongside the default response(s). The NLP interpreter randomly selects from multiple responses at runtime.

Response Categories

Category	Description
Statements	Responses triggered by AI Agent or user actions — for example, task cancellation or invalid input.
Queries	Questions the AI Agent asks for more information — for example, Which field do you want to change?
Errors & Warnings	Messages displayed during errors — for example, End date cannot be before the start date.
Questions	General questions posed to the user that don’t fit other categories.
Choices	Yes/No prompts — for example, Would you like to receive push notifications?
Greeting	Initial responses when the app starts or when a user returns after inactivity.

Statements

Responses triggered by actions from the AI Agent or user.

Message	When It Appears
When the user confirms to close the conversation in a language and switch to another language	User ends a conversation in one language and starts the next in a different language.
When only field prompt is present in the OOB	Internal field prompt for action and alert tasks. Not recommended for customization by enterprises.
When language preference is set	User switches language using cheat lang or picks from language options.
When task is not accessible to user	Developer tries to access an app task they don’t have permission for.
User confirms alert setup	An alert is set up successfully (Alert Task only).
User discards/abandons the current activity	User says discard, or the task fails due to max retries, invalid LoV, or cURL error.
User implies to end the conversation	User says bye, good night, etc. Shown when the Follow-up Intents list is presented at dialog end.
Location provided by user is invalid	AI Agent cannot validate input for a Location-type entity.
Amending an Alert field	User says Yes to amend alert parameters (Alert Task only).
Amending a task which has no fields	Task has no entities to amend (Alert, Action, and Information Tasks).
Ask at least one optional field	User did not provide any optional field value when at least one is required ( only).
Topic retry limit exceeded	Developer mode only. Dialog looped through the same nodes beyond the allowed limit.
Universal fallback response	AI Agent cannot generate a response for any reason, or maximum node execution limit exceeded.
Server returned authorization failure	Authorization failure occurred during script or API execution.
User selects a valid option from a list	User selects an action/alert/info task from the listed options.
User selects an action	User selects an action task; AI Agent confirms and asks for required parameters.
Alert is identified, ask for alert time	Displayed when Trigger Interval is set to At a specific time.
Alert is identified, ask for alert time interval	Displayed when Trigger Interval is set to Every X minutes.
User selects an alert	User selects an alert task; AI Agent confirms and requests required parameters.
Task discarded	User types discard mid-task.
App choosing the default account	Shown during alert amend when AI Agent cannot parse the user’s response.
Confirmation Dialog for Task consolidated	AI Agent confirms captured parameters before executing an Action/Info task.
Confirmation Dialog for Alert consolidated	AI Agent confirms all details before setting up an alert (if confirmation is enabled).
Confirmation Dialog for Alert consolidated without fields	Alert has no fields; AI Agent confirms before execution (if confirmation is enabled).
The alerts will be muted	Developer has muted alert notifications.
Confirmation Dialog when alerts are muted	Shown when Mute alert notifications is enabled for an RSS alert.
Alert frequency description	Confirmation of alert query frequency for web-service alert tasks.
Not understanding the input at confirmation	User’s response during alert frequency confirmation does not match the expected input.
No results found for the search string	Typeahead field returned no results.
Successful task setup	Alert task set up successfully.
I am not sure what you mean	One or more low-confidence intents found. Usually shown as FAQ suggestions.
Ambiguity in intent recognition	AI Agent failed to identify user intent; prompts user to rephrase.
User has selected daily alert — ask for frequency or specific time	Both frequency interval and specific time options are available.
App needs to present filter options	Multiple filters available during alert setup.
App needs to present filter options with exit option	Same as above, but includes an option to skip.
User finished configuring filters	Filter configuration for an alert task completed.
User asks Who am I	User types Who am I.
User asks for help	User types Help; AI Agent lists available tasks.
Skipped values	Part of confirmation message for Action/Info tasks — shows skipped (optional) parameters.
Ask user to change dependent field	A dropdown depends on another field, but the Platform cannot fetch data from that field.
When there is no data	User chooses to change account but responds no to the AI Agent’s confirmation.
When date given in invalid format	User enters an unrecognizable date format. Valid formats: YYYY-MM-DD, DD-MM-YYYY, 1 Jan 1982, Next Friday.
Follow-up sentence post task discard	AI Agent confirms task was discarded and offers to restart.
Confirmation Dialog for task with skipped fields	Confirmation before executing an Action/Info task when some optional fields were not provided.
Listing all tasks	Displayed when user types Help and no task matches.
Listing all alerts	Displayed when user types Help and the app has only alert tasks.
Listing all FAQs	Displayed when user types Help — includes Knowledge Graph collections.
Listing all alerts after tasks	Shown when app only has alert tasks and intent is not identified.
Listing all alerts after tasks and FAQs	Shown when all task types are configured and user types Help.
Lists all actions and FAQs	No matching task found; lists actions and FAQs.
Lists all alerts and FAQs	No matching task found; lists alerts and FAQs.
When app has no tasks	Developer tries to run an app without any tasks.
When a server call fails	Webhook or service node call failed; task is discarded.
User prompt when a task is discarded due to server call failure	API call did not complete successfully.
Specific app is not available	App is not yet published.
Field title and value in confirmation	Part of the confirmation step — shows captured field values for alert, action, or info tasks.
Recursive dialog found	AI Agent recursively called the same dialog and exceeded the limit (100).
Loop limit reached for a node	Same node invoked repeatedly beyond the allowed number of times.
Message component doesn’t have a message	Message or Service node returned empty content.
User says thanks	User thanks the AI Agent.
User says OK, fine, yes	Shown after task completion when the user acknowledges positively.
User says no, nope	Small talk response to negative acknowledgments.
User says great, awesome	Shown after task completion when the user responds positively.
User says that helped, that was useful	Shown after positive feedback (great, cool, fair enough, that helped).
User says I am done	Shown when user indicates they are finished (done, that’s all, that’s it).
Successfully added filter	A filter was added to an alert.
After successfully adding a filter, showing filter options again	Shown when a filter is applied to an alert (Alert Task only).
User gave invalid input for filter	Filter name doesn’t match any configured filter.
Resume paused task	User chooses to resume on-hold task. Shown only when Resume option is Notify the user with a message that the on-hold task is being resumed.
Notify the user that we have canceled the paused task	User chose not to resume the on-hold task during confirmation.
Notify the user that we have canceled the current task to switch to another	Hold option is Discard the current task and switch to new.
When an interruption is identified but hold is not allowed	New intent detected during task execution; new intent added to follow-up list (current task continues).
Service call timed out	Third-party API call (webhook or service node) timed out.
Dialog has no transitions	Dialog task has only an Intent node and no further flow defined.
Manage account	User says manage accounts, add phone number, delete account, etc.

Queries

Questions the AI Agent asks users for more information.

Message	When It Appears
Ask if user would like to switch to a new language (multiple detected)	Multiple new languages detected mid-conversation; user asked to select preferred language.
Ask if user would like to switch to a new language (single detected)	A different language detected; user offered option to switch or continue.
Follow-up sentence post task execution or discard	Task completed or discarded; user offered option to restart.
Confirmation before discard	User types Discard or Discard All during task execution (applies to select clients).
User types Change or similar words	User changes their mind during alert filter setup; alert setup is discarded.
User prompted to enter text to search	Typeahead field prompt (Action, Alert, Info tasks).
Ask the user to enter alert frequency	Frequency can be interval or specific time (Alert only).
Ask the user to enter alert interval	Frequency is an interval, e.g., every 4 hours (Alert only).
Ask the user to enter alert time	Frequency is a specific time, e.g., daily at 9:00 am.
Alert identified — ask for frequency	Prompted based on schedule options defined by the alert developer.
Prompting user to enter name for alert instance	User installs alert in Kore messaging rooms.
User needs to select from a static dropdown	Filter field type is static dropdown.
User needs to select from a search results list	Filter field type is dynamic dropdown.
Request for email address for webhook instructions	Webhook alert setup — developer email needed for configuration instructions.
Query to get the alert instance description	User prompted to provide a description during alert setup.
Tell me the alphabet or unique word for the action	Displayed before a list of options when no custom template is used.
I am not sure what you mean after an invalid choice	User’s follow-up doesn’t match any of the initially identified ambiguous alerts.
Asking user to pick an alert	A list of alerts is shown; user prompted to select one.
Asking user to pick an alert after an error	User selected an invalid option from the alert list.
When user responded with invalid answer to confirmation dialog	User said they want to amend but the AI Agent didn’t understand what — confirmation repeated.
Typeahead field prompt (first field in task)	First field in task is a typeahead type.
Ask user what to rename the alert	User chose to change/amend the alert name.
After user amends a parameter — ask if anything else to change	User amended a parameter; AI Agent asks if further changes are needed.
No results found for search string — ask to try again	Typeahead field query returned no results.
Follow-up sentence post discard — Part 2	At alert confirmation, user said no twice.
Listing ambiguous tasks	User prompted to pick a task from a list; response led to ambiguity.
Ask if user wants to continue the paused task	Shown after the interrupting task completes (Resume option: Get confirmation before resuming).
Ask if user would like to switch	Hold & Resume is set to Allow the end user to select the behavior; AI Agent confirms before switching.

Errors & Warnings

Messages displayed when errors or validation issues occur.

Message	When It Appears
Agent transfer fallback response	Agent unavailable, wait time exceeded, or agent transfer failed.
User exceeds maximum retries for an entity (task on hold)	User exceeded max retries while another task is on hold.
Dialog task associated to FAQ is not available	The dialog linked to a FAQ was suspended, deleted, or not assigned to the user.
User needs to authorize or re-authorize	End-user authorization or re-authorization is required. The authorization URL is appended dynamically.
User authorization has failed or is yet to authorize	Task requires authorization; AI Agent provides an authorization link.
User’s intent recognized, but authorization fails	Service call needs authorization and no account has been authorized yet.
User authorization failed more than 3 times	User entered wrong credentials three times in a row.
User provided value is not unique	User’s input causes ambiguity when selecting from a partially matched intent list.
App did not find any options based on user’s choice	Option set is empty, or no account credentials are available.
Discarding task — no options found for list or lookup	User input not found in service call response (dynamic dropdown or typeahead field).
Task identified, but server call failed	Kore account for the third-party API expired or failed.
User enters invalid email	Invalid email address entered for an alert, action, or info task.
User enters invalid country	Invalid country name entered.
User enters invalid URL	Invalid URL entered.
User enters invalid phone number	Invalid phone number entered.
User enters invalid number	Non-numeric input entered for a number field.
Invalid date format	Invalid date format entered.
Missing date	No date provided for a date field.
Ask user to enter time for a field	Time missing for a DateTime entity in a dialog task.
Ask user to enter time	User provided only a date for a DateTime entity (e.g., scheduled appointment).
Invalid time — ask to try again for a field	User’s time input for a DateTime entity is missing or invalid.
Invalid time — ask to try again	User entered an invalid time value (e.g., Aug 29 at 25).
Missing date/time	Error prompt for DateTime entities in Actions and Alerts.
You have chosen a wrong choice	Sub-response used when a user selects an invalid option from a list.
Cannot understand your request	AI Agent fails to identify intent; lists available tasks and FAQs.
App is selected, user asked for reauthorization	API call failed due to expired credentials.
Listing all tasks after wrong choice	User selected an invalid option; app has only Actions/Dialogs/Info tasks.
Listing all alerts after wrong choice	User selected an invalid option; app has only Alert tasks.
Listing all FAQs after wrong choice	User selected an invalid option; app has only FAQs.
Listing all FAQs — not understanding request	App has only a Knowledge Graph and fails to understand intent.
Listing all actions — not understanding request	App has only Action tasks and fails to understand intent.
Listing all actions and alerts — not understanding request	App has only Action and Alert tasks and fails to understand intent.
Listing all actions and FAQs — not understanding request	App has only Actions and a Knowledge Graph and fails to understand intent.
Listing all alerts — not understanding request	App has only Alert tasks and fails to understand intent.
Listing all alerts and FAQs — not understanding request	App has only Alert tasks and a Knowledge Graph and fails to understand intent.
Listing all tasks — not understanding request	App fails to understand intent; shows all available tasks.
Listing all tasks and alerts after wrong choice	User selected an invalid option; app has tasks and alerts.
Listing all alerts, tasks, and FAQs after wrong choice	User selected an invalid option; app has all task types.
Lists all actions and FAQs after wrong choice	App has only Actions and a Knowledge Graph; user selected an invalid option.
Lists all alerts and FAQs after wrong choice	App has only Alerts and a Knowledge Graph; user selected an invalid option.
Task discarded — user exceeded maximum allowed attempts	User entered an invalid entity format five times.
Entity component in dialog has no error prompt	Entity node has no error prompt defined.
Entity component in dialog has no message attribute	Entity node has no message prompt defined.
User input exceeds maximum sentence or word limit	Input exceeds the allowed character or word count.

Questions

General questions the AI Agent poses to users.

Message	When It Appears
When user input matches and frequency is asked	User input led to ambiguity while setting an alert frequency.
Ask user to input alert frequency	Prompt for frequency at which the alert task will run.

Choices

Yes/No prompts used during alert and task setup.

Message	When It Appears
Search results displayed (type-ahead search)	User input triggered a typeahead search in an alert filter (Alert Task only).
Confirming alert frequency	App confirms current alert trigger frequency with the user.
Repeat confirmation of default alert frequency	User input doesn’t match the expected frequency response.
Confirmation to set up alert with frequency	Confirm the trigger frequency for an alert.
Prompting user to set up filters	App asks whether the user wants to configure alert filters.
Prompting user to set up filters as soon as alert is identified	Same as above — triggered immediately on alert identification.
Asking user to set up push notifications	User wants to amend the alert notification setting.
Repeat — asking user to set up push notifications	User’s response to notification setting doesn’t match yes/no.
Amending the description	User wants to change the alert description.
Amending the frequency	User wants to change the alert trigger frequency.
Amending notification settings	User wants to change whether notifications are enabled.
Prompting user to choose days of the week	User selecting Weekday/Weekend for trigger frequency.
Query to choose interval for frequency	User prompted to select a frequency interval.
Query to send webhook setup instructions via email	App checks if developer wants webhook configuration instructions emailed.
Do you want to change the alert name?	User wants to amend the alert name.
Query to choose operator while setting up a filter	User prompted to select an operator for an alert filter.
Options to pick frequency for alert notification days	All interval options available (Daily, Weekday, Weekend).
When the user needs to choose from provided choices	User prompted to pick from a dropdown in alerts or actions.
Confirm if user wishes to reuse a pre-saved context	Prompted when a pre-saved context is available for reuse.
Show possible intervals for alert schedule	User entered a specific time, but only intervals are valid.
Show closest frequency for alert schedule	User’s input doesn’t match available frequencies; nearest option shown.
Show closest time for alert schedule	User’s input doesn’t match available times; nearest option shown.
When changing a field	User selected No during alert parameter confirmation.
Prompting to set up filters after alert identified	User provides input to set up an alert that has filters configured.
Ask if setup instructions should be sent via email (first prompt)	First prompt in alert setup is to ask about emailing setup instructions.
While choosing options	Shows options from the Platform for a typeahead entity in an action or alert.
After selecting task — ask about schedule	Alert selected from list; AI Agent asks how often the user wants to be notified.
Wrong choice for task schedule	User entered invalid frequency for an alert; AI Agent asks again.
After selecting task — ask for field value	First field in an action or alert is a dropdown.
User says no for alert setup confirmation	User chose not to proceed with the alert configuration; offered to amend.
After selecting task — user selects wrong choice in dialog	User entered an invalid choice from a non-LoV list.
User enters wrong choice for alert frequency	Alert frequency value doesn’t match options defined by the developer.
User enters wrong value for a dropdown	Input doesn’t match any value in the List of Values.
List ambiguous LoV values	Multiple partial matches for a List of Values entity; user asked to select.
User enters invalid choice for an action	User’s response doesn’t match the presented task list.
Ambiguous date	User entered an ambiguous date (e.g., 6/9); both interpretations shown.
Ambiguous airport	User entered an airport name that matches multiple airports (e.g., London).
Ambiguous location	Geocode service returned multiple location possibilities.
More than one account to pick from	User has multiple third-party accounts; asked to select one.
Intent recognized, but more than one account to pick from	Same as above, when intent is identified.
Custom lookup returned more than one value	Multiple results returned from a custom list query.
Chosen list is empty after selecting a valid choice	Dependent dropdown has no options because the parent field value returned empty results.
User selects an invalid choice (dependent dropdown)	Same as above; user’s value for the dependent field is invalid.
Present follow-up tasks at end of dialog	Shown at end of dialog when Follow-up Task setting is Yes, ask the user to select a follow-up task.
Ask user to select entity to amend (ambiguous entity value)	Multiple entities share the same value during dialog execution; user asked to clarify which to amend.
Max on-hold quantity exceeded — ask user to cancel a task	User tried to put more tasks on hold than the configured maximum.
User provides intent when prompted for String/Description entity	User input matched an intent when prompted for a string/description entity; AI Agent asks how to proceed.

Add and Edit Responses

To add a new standard response:

Select a category tab and hover over the target situation.
Click Add Response.
In the Add Channel Overrides dialog:
- Select a channel (or Any Channel for all channels).
- For channel-specific responses, select a task and optionally a field.
In the Response editor, choose a mode:
- Simple Mode — Plain text with optional formatting (bold, italics, hyperlinks, lists).
- Advanced Mode — JavaScript for custom channel-specific messages. Use response.message for the default message string and response.choices for the options array.
Click Save.

Channel Override Examples

Slack:

var message = {};
var attachments = [];
var attachment = {
    'text': "Exception occurred during API execution",
    'color': '#FFCC00',
    'attachment_type': 'default'
};
attachments.push(attachment);
message.attachments = attachments;
print(JSON.stringify(message));

Facebook:

var message = { text: response.message };
var buttons = [];
for (var i = 0; i < response.choices.length; i++) {
    var choiceItem = response.choices[i];
    if (choiceItem.length > 20) {
        choiceItem = choiceItem.substring(0, 18) + '..';
    }
    buttons.push({
        content_type: 'text',
        title: choiceItem,
        payload: choiceItem
    });
}
message.quick_replies = buttons;
print(JSON.stringify(message));

SMS:

var message = {};
var indexArray = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z'];
var text = response.message + '\n';
for (var i = 0; i < response.choices.length; i++) {
    text = text.concat(indexArray[i] + ') ' + response.choices[i] + '\n');
}
message.text = text;
print(JSON.stringify(message));

Email:

var message = {};
var text = response.message + ' ';
for (var i = 0; i < response.choices.length; i++) {
    text = text.concat('<u>' + response.choices[i] + '</u><br>');
}
message.text = text;
print(JSON.stringify(message));

Customizing Ambiguity Display with `getChoices()`

Use getChoices() to retrieve ambiguous intent choices and render them in a custom format (for example, as buttons):

var info = getChoices();
var message = {
    "type": "template",
    "payload": {
        "template_type": "button",
        "text": "Did you mean",
        "subText": "You can now customize response for did you mean",
        "buttons": []
    }
};
for (var i = 0; i < info.length; i++) {
    message.payload.buttons.push({
        "type": "postback",
        "title": info[i],
        "payload": "payload1"
    });
}
print(JSON.stringify(message));

Manage Interruptions

Users rarely follow a linear conversation path — they may switch topics, ask unrelated questions, or provide a new intent mid-task. Manage Interruptions gives you granular control over how the AI Agent handles these situations. Go to Automation AI > Conversation Intelligence > Conversation Management > Manage Interruptions.

Interruptions Hierarchy

Settings at more specific levels override broader defaults.

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.

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.

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

Disables 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

Defines 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. Message is customizable.
Notify the user and resume	Resumes the on-hold task automatically and notifies the user. No confirmation required. Message is customizable.
Resume without any message	Directly resumes the on-hold task with no notification.
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 settings.

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.

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

Node-Level Customization

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

Dialog-Level Customization

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

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.

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.

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.

Behavior for User Authorization

When a user provides unexpected input at an authorization prompt:

If Small Talk is detected: Small Talk response is shown, and the authorization prompt is repeated.
The Platform checks for 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.
If no intent is found — User is re-prompted with the authorization link.

Small Talk

Small Talk enables the AI Agent to engage users in casual conversation. AI Agents that respond to social inputs like How are you? create a positive impression and improve engagement. Go to Automation AI > Use Cases > Small Talk.

Intents and FAQs take precedence over Small Talk. If both are detected, the intent or FAQ is handled first. Small talk is language-specific — configure it separately for each enabled language.

Features

Feature	Description
Custom Small Talk	Build a personality and train the AI Agent for casual conversations around your areas of interest.
Nested conversations	Support follow-up queries during a conversation for more engaging interactions.
Emoji identification	The Platform recognizes emojis in user utterances and responds accordingly.
Intent-aware	Interjections like hello at the start of an utterance are not treated as small talk if an intent follows — for example, Hello, I need to book a flight triggers the book-a-flight intent, not Small Talk.

Default Small Talk (Greetings Group)

When you create an app, a Greetings group is generated automatically. The following messages are included:

User Query	When It Appears
User says Hi, Hello, Hey, etc.	Shown when the user greets or enters the app name.
How are you?	Shown for greetings like how are you, how’s it going, wassup, etc.
Who are you?	Shown when the user asks who the AI Agent is.
User says great, awesome, etc. (no task in context)	Shown after task completion when the user responds positively.
User says that helped, that was useful, etc.	Shown after positive feedback.
User says no, nope, etc. (no task in context)	Small talk response to negative acknowledgments.
User says I am done, That’s it for now, etc.	Shown when the user indicates they’re finished.
User implies to end the conversation (bye, good night, etc.)	Shown when Follow-up Intents are presented at dialog end.
User says thanks	Shown when the user thanks the AI Agent.
User says OK, fine, yes, etc. (no task in context)	Shown after task completion when the user acknowledges positively.

When you open Small Talk in an existing app, you are prompted to migrate Standard Response greetings to Small Talk. This migration is permanent — greetings are removed from Standard Responses and cannot be retrieved if you proceed.

Set Up Small Talk

Step 1: Add a Group

Go to Automation AI > Use Cases > Small Talk.
Click New Group.
Enter a Group Name and click Proceed.

Groups are evaluated in the order they are added.

Step 2: Add Query–Response Pairs

In the Small Talk editor, enter the User Query and Bot Response.
Press Enter or click Add to save the pair.

Additional options per query:

Alternate questions — Additional phrasings for the same query. These are pattern-based.
Alternate responses — Multiple responses; the Platform picks one at random at runtime.
Child queries — Follow-up questions available only after the parent is answered. Up to 3 levels of nesting are supported.

Use the drag handle to reorder queries. Add specific queries before generic ones (for example, how is your day today? before how are you).

Step 3: Configure Channel-Specific Responses (Optional)

Hover over a response and click the Settings (gear) icon.
In the Manage Response window:
- Add a default message for all channels.
- Click + Add Response > New Prompt Message for channel-specific responses.
- Select a Channel, then enter a response in Simple or Advanced (JavaScript) mode.
Click Save.

Import Small Talk

To import an existing Small Talk dataset:

Click the more (ellipsis) icon in the Small Talk view.
Select Import.
Download the sample file if needed to verify formatting (JSON or TSV).
Select your file and click Next.

Warning: Importing replaces all existing Small Talk data.

Small Talk Context

Mark sections of user input patterns with _ to capture contextual data. This data is stored in context.smallTalk.matchData and can be used to personalize responses. Example pattern: I am from _~location Where are you from?

// context.smallTalk
{
  "pattern": "I am from _~location Where are you from?",
  "matchData": {
    "_0": "chicago"
  },
  "previousMatchData": []
}

Multiple tokens are positional (_0, _1, _2, etc.) and may not be contiguous if an optional token is absent. Example pattern with optional token: how is the _[ climate temperature ] { at _~location } ~time

User says “how is the temperature in London today” → _0: temperature, _1: London, _2: today
User says “how is the climate today” → _0: climate, _2: today (_1 is absent)

Access parent-level context from child nodes via:

context.smallTalk.previousMatchData[i].matchData._0

Context persistence: The context is cleared when:

The session ends.
A new primary Small Talk intent is detected.
Any non-Small Talk intent is triggered.

Training

After making changes to Small Talk, click Train in the Small Talk editor to retrain the assistant.

Importing a Small Talk file triggers training automatically.

Key rules:

Only 3 levels of child nodes are allowed.
Duplicate top-level questions across groups are not allowed.
Questions are evaluated in the order added within a group; groups are evaluated in the order created.

Alert Tasks

Alert Tasks proactively deliver information to users by polling a service on a real-time or scheduled basis. Users can configure alerts to trigger at specific times or when defined conditions are met. Go to Automation AI > Use Cases > Alert Tasks.

Example Use Cases

Time- or location-based promotional offers
Scheduled delivery updates
Inventory level changes
Critical issue notifications
Weather alerts
Cross-system status changes

Create an Alert Task

Open the AI Agent.
Navigate to Automation > Use Cases > Alert Tasks.
Click Create Alert Task.
Configure the General Settings:

Field	Description
Task Name	The name displayed in the Platform.
Display Name	The name displayed in the Store. Defaults to Task Name.
Connection Type	Webhook — app pushes data to a Kore.ai endpoint. RSS — Kore.ai polls an RSS feed. Webservice — Kore.ai polls a web service using end-user credentials.
Connection Mode	For Webservice only: REST or SOAP.
Input Preconditions	Intent is detected only if specified context tags are present. Available via More Options.
Context Output	Context tags set when this intent is initiated. Available via More Options.

Advanced Settings

Field	Description
Task Install Confirmation Message	Message shown when the task is successfully set up.
Turn Off Confirmation Messages	Yes — skips NLP execution confirmation. No — prompts the user before running.
Search Keywords	Keywords users can use to find this task in the marketplace.
Task Demo Link	Link to display a task demo icon in the Store.

Authorization

Define how end-users are authenticated to receive alert notifications. Not required for Webhook connection types.

On the API Request tab, expand the Authorization section.
Select an existing Authorization Provider or click Create New.
Select an Authorization Type:

Type	Description
Basic Auth	Standard username/password collection over SSL.
OAuth v2 Password Grant Type	Custom authorization for non-standard web service auth flows.
OAuth v1	Protected resource access via API without exposing user credentials.
OAuth v2	Modern OAuth with specific authorization flows for web apps and services.
API Key	Token-based identification/authentication for incoming app requests.

API Requests

Configure the API request that fetches data for the alert. The configuration depends on the Connection Type:

Connection Type	Configuration Guide
Webhook	Configure a webhook endpoint to receive pushed data.
REST	Define REST API parameters, initializers, pre/post-processors, and response samples.
SOAP	Define SOAP API parameters and the request envelope.
RSS	Define the RSS feed URL and polling parameters.

You can define API Parameters, an optional API task Initializer, API Request pre/post-processors, and expected Response Samples.

Bot Responses

After configuring the API request, define how alert results are displayed to the user.

Component	Description
Report Template Definition	Displays payload data as a table, contact card, or formatted report. Not applicable to Webhook, Action, or RSS tasks.
Message Definition	Defines the look and feel of the alert message delivered to the end-user based on the webhook payload.

Alert Settings

Define when and how the alert fires. Not applicable to Webhook tasks.

Setting	Description
Filters	Limit alert results by criteria (name, account, amount, text match, etc.). Only matching results trigger a notification.
Schedule	Set polling frequency and schedule options presented to users during setup.
Advanced Settings	Configure the initial alert message set, alert reminders, and duplicate message prevention.

Version Namespaces

Associate Variable Namespaces to scope which variables are used during task execution. This option is only available when Namespaces are enabled.

View Logs

View logged events for published alert tasks to troubleshoot request/response behavior.

In Alert Tasks, click the Settings (gear) icon for the task.
Click View Logs.

Each log entry is a JSON object containing the request, response headers and body, and execution time.

{
    "requestLogs": [{
        "name": "Request 1",
        "request": {
            "endpoint": "https://www.example.com/rss/{channel}.xml",
            "pathFields": {
                "channel": "latestnews",
                "lastRun": "2025-10-09T10:20:45Z",
                "now": "2025-10-09T10:26:37Z",
                "today": "2025-10-09"
            },
            "payloadFields": {},
            "headerFields": {
                "If-Modified-Since": "Thu, 09 Oct 2025 10:00:00 GMT",
                "If-None-Match": "\"example-123456\""
            }
        },
        "response": {
            "status": 200,
            "headers": { ... },
            "body": "Example.com Latest News\n..."
        },
        "executionTime": "4860 ms"
    }],
    "alertsData": [{
        "uxmap": {
            "channel": "default",
            "body": "<% var newLink = title; print(newLink); %>"
        },
        "payload": "{ ... }",
        "response": {
            "channel": "default",
            "body": "Example.com launches new AI-powered feature\n"
        }
    }]
}

Refresh the Logs page to view the latest entries.

Modules

Platform Services

References

​Standard Responses

​Response Categories

​Statements

​Queries

​Errors & Warnings

​Questions

​Choices

​Add and Edit Responses

​Channel Override Examples

​Customizing Ambiguity Display with getChoices()

​Manage Interruptions

​Interruptions Hierarchy

​Allow Interruptions

​Do Not Allow Interruptions

​Allow the End User to Select the Behavior

​Resume Options

​On-Hold Quantity

​Node-Level Customization

​Dialog-Level Customization

​Behavior for FAQs

​Behavior for Ambiguous Intents

​Behavior for Small Talk

​Behavior for User Authorization

​Small Talk

​Features

​Default Small Talk (Greetings Group)

​Set Up Small Talk

​Step 1: Add a Group

​Step 2: Add Query–Response Pairs

​Step 3: Configure Channel-Specific Responses (Optional)

​Import Small Talk

​Small Talk Context

​Training

​Alert Tasks

​Example Use Cases

​Create an Alert Task

​Advanced Settings

​Authorization

​API Requests

​Bot Responses

​Alert Settings

​Version Namespaces

​View Logs

Standard Responses

Response Categories

Statements

Queries

Errors & Warnings

Questions

Choices

Add and Edit Responses

Channel Override Examples

Customizing Ambiguity Display with `getChoices()`

Manage Interruptions

Interruptions Hierarchy

Allow Interruptions

Do Not Allow Interruptions

Allow the End User to Select the Behavior

Resume Options

On-Hold Quantity

Node-Level Customization

Dialog-Level Customization

Behavior for FAQs

Behavior for Ambiguous Intents

Behavior for Small Talk

Behavior for User Authorization

Small Talk

Features

Default Small Talk (Greetings Group)

Set Up Small Talk

Step 1: Add a Group

Step 2: Add Query–Response Pairs

Step 3: Configure Channel-Specific Responses (Optional)

Import Small Talk

Small Talk Context

Training

Alert Tasks

Example Use Cases

Create an Alert Task

Advanced Settings

Authorization

API Requests

Bot Responses

Alert Settings

Version Namespaces

View Logs