The AI Prompt node

The AI Prompt node makes a single-turn call to a language model and writes the result as the node's output. It is the simplest of the two AI nodes: one prompt in, one answer out, with no tools and no multi-turn conversation. This article is for workspace users with the Admin role or higher who build workflows in the Designer.

• What the Node Does
• Configuration Fields
• Choosing a Model
• Grounding on Knowledge
• Usage Credits

What the Node Does

Add an AI Prompt node from the AI section of the palette. When the step runs, AutomationAI sends your prompt — and an optional system message — to the model and captures the response. This is a single request and reply: the node does not call tools or carry on a conversation across turns. For a goal-directed node that reasons and uses tools, use the Agent node instead.

Configuration Fields

• promptTemplate — the prompt sent to the model. It supports {{ }} binding expressions, so you can fold in values from earlier nodes (see the article on passing data between nodes).
• systemMessage — an optional system message that frames how the model should respond
• model — shown in the node as Model (optional). Leave it blank, the default, to use the runner's default model; the field's placeholder reads Runner default. See Choosing a Model.
• maxTokens — the maximum length of the response, from 1 to 32768 tokens (default 1024)
• outputKey — the field name the response is written under in the node's output (default text)

You can also define explicit parameters bindings that the prompt template references. After the step runs, later nodes read the result at nodes.<id>.output.<outputKey>.

Choosing a Model

The step runs on a runner, which calls the AI provider configured on that runner. Each runner has a default model, configured in the runner's Key Vault along with its provider. See the article on choosing the runner AI provider.

• Blank model — the step uses the runner's default model. This is the recommended setting for most nodes.
• Named model — a request, not a requirement. The runner tries that model on its own AI provider, and if the provider reports that the model does not exist, it falls back to its default model once and continues. Any other error fails the step as usual.

Enter the model as a fixed value; model IDs differ by provider, for example gpt-5.4, openai/gpt-5.4, claude-sonnet-5, or a Foundry deployment name. Each run records the model that actually answered, and the step's detail in the run history shows it. When the runner fell back, the step shows a notice such as "Requested gpt-9 → ran on gpt-5.4 (not available on this runner)."

A named model requires runner version 2.2.0 or later. An older runner ignores the field and uses its default model. A self-updating runner picks up the current version automatically; see upgrading a runner to a new version.

Script nodes that call the runner's AI helper directly can request a model the same way, with the optional -Model parameter of Invoke-NodeAi in PowerShell, model: in Node.InvokeAi in C#, or model= in invoke_ai in Python. Leaving it out uses the runner's default model.

Grounding on Knowledge

You can ground an AI Prompt node on your knowledge base so the model answers with your own documents as context. In the node's grounding section, attach one or more Knowledge folders or documents; each appears as a chip you can remove. When the step runs, AutomationAI recalls the most relevant passages from the attached knowledge and supplies them to the model along with your prompt. For the full picture, see the article on grounding agents and AI Prompt nodes on knowledge.

Usage Credits

The AI Prompt node runs on your runner, which calls the runner's own AI provider, so its tokens are billed by that provider and do not consume Usage Credits. The step still records the model and token counts so you can review them in the run history. When the node is grounded on knowledge, AutomationAI searches your knowledge on each run, and that search consumes a small number of Usage Credits. See the article on understanding Usage Credits and billing for what is metered.