> ## Documentation Index
> Fetch the complete documentation index at: https://docs.os.default.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI steps

> Send a prompt to an LLM from a workflow and use its response in later steps.

AI steps let a workflow call a language model directly, so you can generate or transform text
without leaving the workflow.

<Note>
  Add a step in the **Workflows** builder with **Add Node**. Many fields accept data from the
  trigger or an earlier step: type `{{` to open the variable picker.
</Note>

## AI Prompt

Sends a prompt to a language model and returns its response.

| Field              | What it does                 | How to use it                                                                                                                     |
| ------------------ | ---------------------------- | --------------------------------------------------------------------------------------------------------------------------------- |
| **Model**          | Which language model to use. | Pick a model from the list, for example an OpenAI, Anthropic, Google, or xAI model.                                               |
| **Dynamic prompt** | The prompt to send.          | Write the prompt. Type `{{` to insert workflow data into it, for example the lead's company name or an earlier enrichment result. |

For example, prompt the model to summarize a lead's enriched company description into one
sentence, then use that sentence in a Slack notification or CRM field.

If the request fails (for example, an invalid model or an empty prompt), the step does not stop
the workflow. Instead it returns a failed result so a later **Multi-Branch** step can react to
the failure rather than the whole run erroring out.

The response text is available to later steps, along with whether the request succeeded and,
if it failed, an error message you can log or act on.

{/* verified against product:
   AI Prompt node label, config fields, and soft-failure behavior
     → apps/default-api/.../workflows/registry/nodes/ai/ai-prompt.ts
   model picker sourced from the curated Vercel AI Gateway catalog (OpenAI, Anthropic,
     Google, xAI models) per the node's own doc comment; provider/vendor names kept generic
     here since the specific catalog changes independently of this doc.
   prompt field placeholder confirmed as "Type {{ to insert a data reference..."
     → apps/default-api/.../workflows/registry/config.ts (AI_PROMPT_INPUT_MODE)
   NOTE: registry-driven labels may false-flag in the lint. */}