Gen AI

Gen AI blocks use AI to produce content inside a workflow: write a reply or a summary from a prompt or from your Knowledge Base, or pull structured details out of free text. To hand the conversation itself to an AI agent, use the AI Agent blocks instead.

Find them under Gen AI in the workflow builder’s toolbar.

Summarise and Label also appear in the toolbar but are not available yet. The AI Agent block also sits under Gen AI in the toolbar; it is covered with the other AI Agent blocks.

How the Generate block works

Generate writes text with AI and saves it in a variable. Under Generate From, choose the source:

• AI Model: the model follows your prompt. Use it to summarise a ticket, draft a reply, check whether a question was answered, or turn text into JSON.
• Knowledge Base: the block searches your Knowledge Base with a question you set and writes an answer following your response instructions.

You also choose:

• Output: plain text, or JSON when later blocks need structured fields.
• Inputs: what the AI can see: the latest reply, the conversation history (and how many messages of it), internal replies and attachments.
• Model: which model to use. See model comparison.

Generate has a success exit and an error exit, so you can handle the case where the AI call fails.

For example, a ticket workflow can use Generate with an AI Model prompt to summarise the images a customer attached, then pass the summary to an AI agent.

How the Extract block works

Extract reads a message or a variable and pulls out the entities and intents you choose, saving each into its variable. Use it when the details you need are buried in free text, for example a date, an order number and a reason in one email.

Extract is more flexible than the Capture block in Listen: Capture works on the customer’s next reply, while Extract works on any text already in the workflow and can pull out several values at once.