Converting an AI Playground run into a workflow

The AI Playground answers questions and completes tasks in conversations: you plan each step with Orion and start it yourself. When a conversation does something you want to repeat (on a schedule, from a webhook, or as part of a larger automation), CloudRadial AutomationAI lets you convert it into a new draft workflow, either as agent steps that keep the tools and company scope each step used, or as PowerShell scripts that AI Generate writes from the conversation. You can also convert a single run. This article is for workspace users who build automations (Admin or Owner Role).

• When to Convert
• Converting a Conversation
• Agent Steps or PowerShell Scripts
• Converting a Single Run from History
• What Carries Over
• What You Get
• Editing the New Workflow

When to Convert

The AI Playground is the fastest way to try something, but a conversation only runs when you plan and start it. Convert it into a workflow when you want to keep it: run it on a schedule, trigger it from a webhook, add steps around it, or hand it to a runner as a deployment. Converting saves you from rebuilding the goals and re-selecting the same extensions by hand.

Converting a Conversation

A conversation can be converted once at least one of its steps has succeeded. Anyone who can open the conversation can convert it, including a conversation someone else started.

1. Open the conversation in the AI Playground.
2. Select Convert to workflow at the top of the conversation.
3. Under How should the workflow do the work?, choose Agent steps or PowerShell scripts, then select Continue.

Only steps that succeeded are converted, in the order they ran. Failed or cancelled steps are left out.

Agent Steps or PowerShell Scripts

Agent Steps

Choose Agent steps to rebuild the conversation as a chain of agents. A dialog opens with the conversation's title as the workflow name. Pick a folder, adjust the name if you like, and select Create workflow. The new workflow opens in the Designer.

The workflow gets one Agent node per successful step, in order. Each node keeps its step's goal, model, extensions and company scope (one company, or all companies), and each node after the first receives the previous node's output, just as a follow-up in the conversation received the earlier result. A conversation with more than 20 successful steps converts its first 20.

PowerShell Scripts

Choose PowerShell scripts to have AI Generate write script steps instead. Workflows opens with AI Generate already filled in: the name is the conversation's title, and the description lists the successful steps in order, each with its company, its tools and a short excerpt of what it returned. Review and edit the description, then select AI Generate. Nothing is generated until you do.

Converting a Single Run from History

You can also convert one run on its own from AI Playground run history (/ai-playground/history), which lists each run, including each step of a conversation. You can start a conversion two ways:

• Use the Convert action on a run's row in the history grid
• Open a run to view its detail, then choose Convert to workflow in the run detail panel

Either way, AutomationAI opens the new-workflow dialog with a suggested name already filled in from the run's goal. Pick a folder, adjust the name if you like, and select Create workflow. The new workflow opens in the Designer.

The history list shows AI Playground runs only. Runs that came from a workflow, a playbook, or a designer node test are not listed here and are not convertible.

What Carries Over

Each Agent node is seeded from its run so it behaves the same way:

• Goal: the instruction the run was given becomes the agent step's goal
• Model: the same AI model the run used
• Extensions: the exact set of extensions the run was allowed to use. If the run had access to every tool, so does the workflow; if it was scoped to a specific set, that same scope is applied; if it used no tools, none are enabled.

When you convert a conversation as agent steps, each step's company scope carries over too. The runner does not: a workflow runs on the runner you deploy it to. You can review a run's extension scope on its detail before converting, so you can confirm what the workflow will inherit.

What You Get

Conversion creates a new draft workflow version with a minimal, valid graph: a Start node, the Agent nodes seeded from the conversation's steps (or a single Agent node when you convert one run), and an End node. A PowerShell conversion produces whatever script steps AI Generate drafts from the description. Nothing is published or deployed automatically. The workflow is a draft you can shape before it runs anywhere.

Editing the New Workflow

Once the workflow opens in the Designer, treat it like any other: rename it, edit each agent's goal or model, add nodes before or after the agents, then publish a version and create a deployment when you are ready. See Creating your first workflow in the Designer for the full authoring flow.