Skip to main content

​
Definition

In the context of configuring a generative AI conversational agent, actions refer to the concrete operations that the agent can perform beyond simple text generation. They enable the agent to do something before, during, or after a call. An action is an explicit instruction that the agent can execute, generally triggered by an identified user intent or a particular context in the conversation. Examples of actions in an AI conversational agent:
  1. Call an external API → Retrieve an order status, check a calendar, query a database
  2. Trigger a workflow → Create a CRM ticket, send a support request, launch an internal procedure
  3. Send an email or SMS → Appointment confirmation, client follow-up, summary delivery
  4. Create / modify a database entry → Update a client profile, record a response
  5. Redirect to a human → Transfer to an advisor when needed, escalate a complex case
  6. Perform a calculation or business verification → Estimate a quote, validate eligibility, compare two products
Relationship with text generation: In a generative agent, AI can identify when an action should be taken from natural dialogue, then trigger the corresponding action via an orchestration mechanism (functions, plugins, or specific modules). Why it’s important: Actions transform a conversational agent from a simple “assistant that talks” into a true autonomous and useful agent, capable of acting in the real or digital world, not just responding.

​
Usage

To effectively respond to the entire communication lifecycle, the action architecture has been structured around three strategic moments.
1

Before the Call

During the ringing interval, you can trigger preliminary actions such as identifying the caller or retrieving essential contextual information to prepare the agent for the interaction.
2

During the Call

Once the call starts and throughout its duration, you can equip your agent with numerous actions.These capabilities allow you to precisely configure how the agent accesses additional information, manages call transfers, determines call termination conditions, and optimizes interactions with IVR systems and hold messages.It is in this phase that actions reveal their maximum potential, offering a multitude of automation possibilities. This flexibility can sometimes complicate design, but our platform has been optimized to make this configuration as intuitive as possible.
3

After the Call

Once the communication is complete, post-call actions can be executed to ensure necessary follow-ups: automatic sending of confirmation emails, synchronization with your management system, updating databases, and other integration processes.

​
First steps

​
Actions During the Call

The following actions are available during the call.

​
Call Transfer

The transfer action allows for the intelligent redirection of a call to another agent or an external number. This strategic feature can be configured to execute at different moments during the interaction.

​
Use Cases

  • Transfer to a specialized agent when a particular request is identified
  • Redirect to an external service outside of business hours
  • Automatic escalation to technical support for complex issues

​
Call Transfer Action Parameters

When configuring a transfer action, please provide the following information:
FieldDescription
Action NameExplicit identifier such as “Transfer” or a more specific name according to your needs
DescriptionPrecise definition of the purpose and mechanism of this transfer action
SIP ConnectionSelection of an existing SIP connection for the transfer (optional)
Maximum Ring DurationMaximum ringing duration (in seconds) before interrupting the transfer and resuming the call by the agent (Min 5s, Max 45s, Default 20s)

​
SIP Connection

If you have one or more existing SIP configuration(s) (see SIP Connection), you can use them as a destination for the transfer action instead of a standard phone number. Custom Headers You can configure additional SIP headers specific to this transfer:
  1. Click on Add to add a new header.
  2. For each header, specify:
  • Header name: Name of the SIP header.
  • Header Value: Value associated with this header.

​
Transfer Timing Restrictions

You can precisely define the time slots allowing transfers:
  1. Enable the option Restrict transfer to time slots
  2. Select eligible days (Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, Sunday)
  3. Define the availability hours (e.g., 09:00 - 17:00)
  4. Click on Add to validate this time slot.
You can set multiple daily time slots by selecting new time ranges and clicking Add again.

​
Adding Transfer Numbers

Two methods are available for adding transfer numbers:
Manual Addition
  1. Select the country from the dropdown menu (country code)
  2. Enter the phone number
  3. Add a description to identify this number (e.g., “Technical Support”)
Import
Allowed File
SpecificationDescription
Accepted Format.csv file (max. 5 MB)
Required StructureSee the table below
File Format
ColumnTypeRequiredDescription
phoneTextYesE.164 format (e.g., +33612345678)
descriptionTextNoDescription of the person or service
Download our standardized import template to facilitate your data preparation.

​
Maximum Ring Duration

You can adjust the maximum wait duration before considering the transfer failure. The default setting is 20 seconds (Min 5s, Max 45s).

​
Best Practices

  • Set time slots aligned with the actual availability of your teams
  • Use clear descriptions for each transfer destination
  • Conduct thorough transfer path tests to ensure a smooth experience

​
Workflow

This feature allows for the development of sophisticated integrations with your external systems during the course of your communications.

Configure Workflows During the Call

See how to configure workflows

​
Hang Up

The hang-up action allows for controlled termination of a communication under specifically defined conditions. This feature can be implemented at different stages of the interaction.
A default hang-up mechanism (not visible in this interface) is already integrated to avoid interaction loops while maintaining user experience quality. There is already a default hang-up action (not visible here) so that my agent does not loop while not hanging up on the user.
Only one Hang Up action can be configured for an agent. Adding a custom hang-up action overrides the default hang-up configuration.

​
Use Cases

The hang-up action is particularly relevant in the following scenarios:
  • When all conversation objectives have been achieved
  • Upon detecting interaction with an automated system (machine/IVR)
  • When the conversation reaches a stalemate without progression

​
Configuration Parameters

Configuring a hang-up action requires defining the following parameters:
ParameterDescription
Action NameUnique and descriptive identifier of the action
Action DescriptionDetailed explanation specifying the goal and functioning of the action
Minimum Number of AttemptsCalibration of the minimum attempt threshold before activation (scale from 1 to 5)

​
Description

Specify in natural language the precise conditions under which the agent should terminate the communication.

​
Attempt Configuration

The configuration control allows you to adjust the minimum number of attempts before triggering (Min 1, Max 5).

​
Best Practices

  • Use precise descriptions detailing the exact termination conditions
  • Configure an appropriate number of attempts suitable for your operational context
  • Conduct thorough testing before deployment in production
  • Optimize parameters based on conversation analysis and feedback

​
RAG

The integration of RAG (Retrieval Augmented Generation) data sources allows you to enrich your agent with specific knowledge to deliver contextualized and precise responses.

​
Required Fields

FieldDescription
Action NameUnique and descriptive identifier of the action
Action DescriptionDetailed explanation specifying the objective and functionality of the action
Message before executionConfigure the message said by the agent before executing the action
Data Source (RAG)Selection of a data source to enrich the interaction

​
Description

Specify in natural language the exact circumstances under which the agent should activate this functionality.

​
Message before execution

Enable or disable the message that the agent will say before executing the RAG action. You can add an example phrase to be spoken if the option is enabled.

​
Data Source (RAG)

The dropdown menu allows you to select from the data sources configured in your environment.

RAG

Configure your first RAG database

​
Keyboard Interaction

This feature allows the agent to emit tones by simulating key presses on a telephone keypad, thus facilitating interaction with interactive voice response systems .
Only one Keyboard Interaction action can be configured for an agent.

​
Required Fields

FieldDescription
Action NameUnique and descriptive identifier of the action
Action DescriptionDetailed explanation specifying the objective and functionality of the action

​
Description

In most implementations, the default description proves adequate, but you can enrich or customize it using natural language suited to your operational context.
For optimal management of IVR systems, it is recommended to use the Wait and Keyboard Interaction actions together.

​
Wait

This feature allows the agent to temporarily suspend its responses, which is essential during holds (welcome music) or to detect IVR systems and wait for the complete announcement of menu options.
Only one Wait action can be configured for an agent.

​
Required Fields

FieldDescription
Action NameUnique and descriptive identifier of the action
Action DescriptionDetailed explanation specifying the objective and functionality of the action

​
Description

In most implementations, the default description proves adequate, but you can enrich or customize it using natural language suited to your operational context.
For optimal management of IVR systems, it is recommended to use the Wait and Keyboard Interaction actions together.
To ensure an optimal experience, judiciously limit the number of actions per communication and ensure their seamless integration into the conversational flow.

​
Actions After the Call

The following actions are available after the call:

​
Email

The email sending action allows for the automated transmission of electronic messages following a telephone interaction.

​
Use Cases

This feature is particularly suited to the following scenarios:
  • Automatic sending of a summary of the conversation
  • Notification to relevant teams following a customer interaction
  • Structured transmission of specific requests identified during the call
  • Automated escalation for issues requiring specific intervention

​
Configuration Parameters

Creating an email sending action requires defining the following parameters:
ParameterDescriptionRequired
Action NameUnique and descriptive identifier of the action✓
SubjectSubject of the email to be sent✓
ToEmail address(es) of the main recipient(s)✓
CcEmail address(es) of recipients in copy
BccEmail address(es) of recipients in blind copy
Conditioned by ClassificationsTrigger criteria based on call categorization

​
Subject Configuration

You can dynamize the subjects of your communications by leveraging the contextual variables of the call. Use the {} icon located to the right of the field to access available variables.

​
Conditional Configuration

To optimize the triggering of the email sending action, you can use the Conditioned by Classifications parameter:
  • Select one or more relevant classifications from the dropdown menu
  • The email will only be sent if the call matches the defined classification criteria
  • This feature is particularly effective in targeting specific interaction scenarios

​
Best Practices

  • Precisely configure triggering conditions to avoid excessive email sending
  • Test your configurations with test email addresses before deployment in production

​
Workflow

This feature allows for the development of sophisticated integrations with your external systems after the end of your communications.

Configure Workflows After the Call

See how to configure workflows