In this article, we will learn more about the AI Generation node and how to set it up in Orchestrator, so your Flows can read, write, and classify text that fixed rules cannot handle.
What Is the AI Generation Node?
The AI Generation node takes data from earlier nodes in your Flow and uses an AI model to generate or transform text.
It is all-purpose. You write the prompt, you choose the model, and you name the results you want back.
2 very common use cases:
- Draft a personalized follow-up email from the meeting context and the account record;
- Read a chat transcript, decide whether a competitor was mentioned, and hand that answer back so the Flow can branch on it.
Why Use the AI Generation Node?
The AI Generation node handles language, which is the part fixed rules cannot do.
A Condition node checks a field. A matching node matches records.
When you need the node to read a paragraph, summarize a transcript, or write a sentence that sounds like a person, AI Generation node is the node that fills that gap.
How to Set It Up
On the canvas, select the plus icon at the point in the Flow where the node should run, then search for AI Generation and select it. The node's side panel opens with three settings, and only the first two are required.
-
Prompt. Tell the AI what to generate. Type
/to insert data from earlier nodes, such as the guest name, the matched account, or the meeting notes. - Model. Pick the model. Models are grouped by provider, and Anthropic, Google, and OpenAI are available. Each model shows its version.
-
Data Outputs (optional). Name the pieces of data you want back. Select Add Data Output, then fill in:
a. Result Name – the name the result is stored under, for example
Subject;b. Data Format – one of five formats: Long text, Number, Boolean, JSON Object, or JSON Array;
c. Explain what you expect for this Result – a description of the output. This is required, and the AI uses it to decide what to produce.
Giving the AI Access to Web Search and the Knowledge Base
The Access To section controls what the AI may search before it writes its results. Both tools are off until you turn them on.
- Web Search – the AI looks up public information online, such as what a company does or its recent news;
- Knowledge Base – the AI reads your workspace's Knowledge Base, such as your email-writing guidelines or product documentation.
When you turn on Knowledge Base, the AI can search every source in the workspace, including sources added later. To narrow it, select Edit Sources:
- All Sources – every source in the workspace, now and in future;
- Selected Sources – a fixed list you build with Add Source. At least one source is required.
Using the Results in Later Nodes
Every Data Output becomes a variable that downstream nodes can use. Once the AI Generation node runs, you can:
- Map
SubjectandBodyinto a Send Email node; - Branch a Condition node on a Boolean result, such as
competitorMentioned; - Write a result into a CRM field with a Create Record or Update Record node;
- Include it in a Slack Notification node.
A common pattern is classify, then branch: the AI reads a transcript and returns competitorMentioned as true or false, and a Condition node sends the true path to a competitive-intelligence Slack channel.
Credits
The AI Generation node consumes Chili Piper credits every time it runs.
There is no fixed cost per run. The amount depends on the model you select, the length of your prompt, and how much information you pass into the node. A short classification prompt costs considerably less than a long prompt that researches a company and writes a full email.
Two places show you the cost:
- Before the node runs, the credits badge at the top of the side panel indicates that the node can consume credits;
- After the node runs, the same badge shows the credits that run actually consumed.
Note: Your plan includes an annual credit allowance. To see your current balance and usage, go to the Billing Center.
Comments
0 comments
Please sign in to leave a comment.