The Add-On AI Features are smaller AI helpers for building projects and reviewing conversations. They all use one model. Under Administration > Settings, the Add-On AI Features tab sets the LLM Configuration, Model and Reasoning Effort for all of them, and for AI Insights. Each add-on is switched on or off for your tenant separately.
What an LLM Configuration is and where other features choose theirs are covered in LLM Integrations. How the LLM Configuration, Model and Reasoning Effort fields behave is covered in Creating an LLM Configuration.
📝 Note: The add-ons stay disabled until the Add-On AI Features tab is filled in. While a control is disabled, its tooltip says the Add-On AI Features have missing information. A control is hidden, rather than disabled, only when its add-on is switched off for the tenant or you are viewing a published version.
The add-ons
| Add-on | Where it appears |
|---|---|
| Utterance generation | An intent's Utterances, Generate with AI |
| Text Response Generation | A Response node, the sparkle button beside + |
| Conversation summarization | Conversation Details, Generate Summary |
| LiveChat summarization | A Live Chat Transfer node, Transfer the conversation summary to Live Agent |
| Generate project with AI | The project cards |
| Generative coding assistant | The Code node |
| Ask LLM Node | The Ask LLM node. It is switched on or off with this group, but its model is chosen on the node. |
Utterance generation
In an intent's Utterances, click Generate with AI.
- Description (required): the intent in a few sentences or keywords, up to 500 characters.
- Number of Utterances: Default: 10 (range 1-30).
For an order-tracking intent, for example:
The customer wants to know where their order is, whether it has shipped, or when it will arrive.
In FAQ import
When you import FAQs into intents, turn on Generate Utterances with AI to add generated utterances to each intent. Default: 5 per intent (range 1-10). With the switch on, the file can have up to 150 rows instead of 1000.
Generate project with AI
On the project cards, click the Generate project with AI icon.
- Name (required): follows the usual project name rules.
- Primary Language (required).
- Description: up to 200 characters. For example:
A support assistant for an online electronics store: order tracking, returns and delivery times. - Number of Intent(s): Default: 10 (range 1-10).
- Number of Utterance(s) per Intent: Default: 3 (range 1-5).
Text Response Generation
On a Response node, click the sparkle button next to + (Add alternative Texts with AI). Each click writes one alternative text. The button appears once the node has at least one text, and it is hidden while SSML is on.
📝 Note: The AI writes alternative texts only when you click the button while designing the flow. At runtime, the node picks one of its texts in the order set on the node (Random or Sequential), without using AI.
Conversation summarization
In Conversation Details, click Generate Summary. The button does not appear if the conversation already has an AI Insights Summary value or a summary has already been generated. It is disabled while masking is in progress. See Conversation Details.
LiveChat summarization
On a Live Chat Transfer node, select Transfer the conversation summary to Live Agent. The summary is generated at the moment of the transfer.
Generative coding assistant
In the Code node, the assistant writes JavaScript from a description you give. It has no model selector of its own: it uses the configuration and model set on the Add-On AI Features tab.
Generative Q&A
Generative Q&A is not available in projects created now. Projects that already use it keep it. For answers from your own documents, use a knowledge base with the RAG Agent node. See Creating a Knowledge Base.