The dialogue builder is where you create conversation flows, or dialogues, for your advanced AI agents. Each dialogue is linked to a specific reply, and each reply is in turn linked to either a use case (for zero-training or agentic AI agents) or an intent (for expression-based AI agents).
This article contains the following topics:
Related articles:
Creating a new dialogue
In the dialogue builder, you can create a dialogue that contains the conversation flow your AI agent should follow when interacting with a user.
- In AI agents - Advanced, in the top-right corner, use the AI agent drop-down field to select the AI agent you want to create a dialogue for.
- In the main menu on the left, click Content > Use cases (for zero-training AI agents) or Content > Intents (for expression-based AI agents).
- Select the use case or intent you want to create a dialogue for.
- Select the Replies tab.
- Click Add reply.
The dialogue builder opens.
To move around within the dialogue builder, click and drag an empty area. To zoom in and out, use the plus (+) and minus (-) icons in the top-right corner, or press CTRL and scroll the mouse wheel.
- In the AI agent message block that is always populated at the start of a new dialogue, enter the text that your AI agent should greet a user with.
- To add additional message blocks, click the plus (+) icon.
You can add them before or after the initial AI agent message block, or anywhere else the plus icon (+) appears. To learn more about the available message blocks, see Available block types.
- To manage an added block, hover your mouse over the block, select the
options menu (
), and select one of the following options:
- Collapse branch: Collapses the branch starting with the selected block to make the overall dialogue easier to see.
- Move: Allows you to select another block to move the selected block underneath.
- Copy: Copies the selected block and the branch it begins, if applicable.
- Paste block: Pastes a copied block underneath the selected block.
- Paste branch: Pastes a copied branch underneath the selected block.
- Actions: Allows you to copy, paste, or remove an action associated with the selected block. See Creating and adding actions for advanced AI agents for more information about actions at the block level.
- Delete: Allows you to delete the selected block or the branch starting with the selected block.
- To link a block to another block, click the link icon (
) underneath a block and then select the block you want the first block to link to.
Note: A customer message block can’t be linked to another customer message block. - To validate that all blocks are correctly configured, click the options menu
(
) and select Validate dialogue.
- To test the dialogue without
affecting the live AI agent, do either of the following:
- Test the entire dialogue by clicking Test dialogue in the upper-right.
- Test a specific branch of the dialogue by hovering over the block
you want to start from and clicking the icon (
).
- To save your progress without affecting the live AI agent, click either of
the following:
- Save: Saves your changes.
- Save with note: Saves your changes and allows you to include a note about what you changed. This information can be viewed in the Version history pane. No logic check is performed when saving as a draft. If multiple people are working on the same dialogue at the same time, the person who saves last overwrites everything.
- To make your changes live to users interacting with the AI agent, click
either of the following:
- Publish: Publishes your changes.
-
Publish with note: Publishes your changes and allows you to
include a note about what you changed. This information can be
viewed in the Version history pane.
A logic check happens at this time. If you can't publish, fix any errors highlighted in red.
Available block types
Block types define where a user goes next within the conversation flow, including what information is asked for and expected.
Block type | Description |
AI agent message | This is a scripted message that the AI agent should say. If you’ve activated rich messaging, this block type supports HTML and Markdown formatting. |
Customer message |
This is where you define how the user might respond to an AI agent and how the dialogue should progress based on that response. For more information, see Using the Customer message block in conversation flows for advanced AI agents.
Note: It's not possible to have two
consecutive customer messages.
|
Generative replies | This is an AI-generated response to a user's
message. For more information, see About the Generative replies block in conversation flows for advanced AI agents. |
Conditional | This is a way to personalize responses in a
dialogue based on certain conditions, including parameters and
detected entities. For more information, see About conditional blocks in conversation flows for advanced AI agents. |
API integration | This is a way to connect to business-specific services to retrieve information that can be used during the conversation (for example, retrieving order status information from a logistics provider). Available only if you have integrations implemented. |
Carousel | This is a rich messaging type for displaying up to
10 different options with a visual card element. Available only
for certain CRMs and channels. For more information, see Adding carousels in conversation flows for advanced AI agents. |
Link to |
This is a way to pass users between dialogues. For example, you can create one escalation flow and link to it from other dialogues instead of creating an escalation flow as part of each individual dialogue. From this block, you can select the reply (dialogue) you want to link to. In the first field, select the use case or template you want to link to. In the second field, select the specific reply you want to link to within the use case or template. If you’ve integrated AI agents - Advanced with Sunshine Conversations, you can select the “Auto-select reply method” value in the second field. The effect that this value has depends on the type of AI agent:
Actions that link to another flow do not increase the automated resolution count. |
Availability | This is a way to respect your business's operating
hours during a conversation to manage user expectations and
check for agent availability before escalating. For more information, see Setting operating hours for advanced AI agents. |
Escalation | This is how you define the escalation path to a
human agent, including the message the AI agent sends to the
user and whether the escalation uses the messaging or email
channel. For more information, see About escalation strategies and flows for advanced AI agents. |