AI Actions

AI Actions are the mechanism through which SetGet AI makes changes to your workspace. Instead of executing changes immediately, the AI always proposes an action first and waits for your explicit confirmation. This keeps you in full control of what happens in your workspace.

What are AI actions?

An AI action is a structured proposal to create, update, or modify something in your workspace. When you ask SetGet AI to do something that involves changing data — like creating a work item, updating a status, or adding a comment — the AI generates an action card that shows exactly what will happen.

You then decide whether to:

• Confirm — Execute the proposed action as-is.
• Modify — Ask the AI to adjust the action before confirming.
• Reject — Cancel the action entirely.

This confirmation flow is a safety mechanism. It ensures the AI never makes unintended changes, even if it misinterprets your request.

The action confirmation flow

Here is how the flow works step by step:

Step 1: You make a request

You type a natural language request in AI Chat that involves a workspace change. For example:

"Create a high-priority bug in the WEB project titled 'Dropdown menu clips on small screens'"

Step 2: AI proposes an action

The AI parses your request, identifies the type of action, and generates a structured action card:

Action: Create Work Item
Project: WEB
Title: Dropdown menu clips on small screens
Type: Bug
Priority: High
State: Backlog (default)
Assignee: Unassigned

The action card appears in the chat as an interactive element with Confirm and Cancel buttons.

Step 3: You review the action

Before clicking anything, review the details:

• Is the correct project selected?
• Are the properties accurate?
• Is anything missing that you want to add?

Step 4: You decide

Decision	What happens
Confirm	The action is executed immediately and reflected in the workspace
Cancel	The action is discarded; nothing changes in the workspace
Modify	You send a follow-up message to adjust the action

Step 5: Confirmation feedback

After confirming, the AI provides feedback confirming the action was executed successfully:

"Done. Created WEB-156: 'Dropdown menu clips on small screens' with High priority in the Backlog state."

If the action fails (for example, due to a permission issue), the AI reports the error and suggests alternatives.

Action types

SetGet AI supports the following action types:

Create work item

Creates a new work item in a specified project.

Property	Required	Description
Project	Yes	Target project for the new item
Title	Yes	Work item title
Priority	No	None, Low, Medium, High, Urgent
State	No	Initial state (defaults to Backlog)
Assignee	No	Member to assign
Labels	No	One or more labels to attach
Due date	No	Target completion date
Description	No	Detailed description of the item
Estimate	No	Effort estimate (if estimation enabled)

Update work item properties

Modifies one or more properties of an existing work item.

Property	Can be updated	Notes
Title	Yes	Rename the work item
State	Yes	Move to a different workflow state
Priority	Yes	Change urgency level
Assignee	Yes	Reassign to a different member
Labels	Yes	Add or remove labels
Due date	Yes	Set or change the target date
Estimate	Yes	Update the effort estimate
Cycle	Yes	Add to or remove from a cycle
Module	Yes	Add to or remove from a module

Add comment

Posts a comment on an existing work item.

Property	Required	Description
Work item	Yes	The item to comment on (by ID)
Comment text	Yes	The content of the comment

TIP

When the AI proposes adding a comment, you can review the exact text before it is posted. This is useful when you ask the AI to draft a status update or technical note.

Modifying a proposed action

If the AI's proposed action is close but not quite right, you do not need to cancel and start over. Instead, send a follow-up message in the same thread:

Original proposal:

Action: Create Work Item
Project: WEB
Title: Dropdown menu clips on small screens
Priority: High
Assignee: Unassigned

Your follow-up: "Also assign it to Sarah and add the label 'ui-bug'"

Updated proposal:

Action: Create Work Item
Project: WEB
Title: Dropdown menu clips on small screens
Priority: High
Assignee: Sarah
Labels: ui-bug

The AI replaces the pending action with an updated version for your review.

Rejecting actions

Click Cancel on any action card to reject it. The action is discarded and nothing changes in your workspace. You can then:

• Rephrase your request for a different action.
• Ask the AI a question instead.
• Close the thread entirely.

Rejected actions are still visible in the thread history for reference, but they are clearly marked as canceled.

Action history

Every confirmed action is recorded in the AI action history, providing an audit trail of all AI-assisted changes.

Viewing action history

1. Open the AI panel from the sidebar.
2. Click Action History (or navigate to Settings > AI > Action History).
3. Browse the list of past actions.

Action history details

Each entry in the action history shows:

Field	Description
Timestamp	When the action was confirmed and executed
Actor	Which workspace member confirmed the action
Action type	Create, Update, or Comment
Target	The work item or entity that was affected
Details	The specific properties that were set or changed
Thread	Link to the AI chat thread where the action was proposed
Status	Success or failure

Using action history for auditing

Action history is useful for:

• Accountability — Know who confirmed which AI-driven changes.
• Debugging — If something looks wrong, trace it back to the AI action that caused it.
• Compliance — Maintain a record of AI-assisted modifications for regulatory requirements.

WARNING

Action history records only confirmed actions. Rejected or canceled actions are visible in their respective chat threads but do not appear in the global action history.

Permissions and actions

AI actions respect your workspace permissions. The AI can only propose actions that your account is authorized to perform:

Role	Allowed actions
Admin	All action types across all projects
Member	Create and update items in projects they belong to
Viewer	No actions available — AI chat is read-only analysis
Guest	Limited to projects they have been invited to

If the AI proposes an action you do not have permission to execute, the confirmation will fail with a permission error. The AI will explain what happened and suggest contacting a workspace admin.

Best practices

1. Always read the action card — Even for simple requests, verify that the AI selected the right project, assignee, and properties.
2. Use item IDs for updates — When updating existing items, reference them by ID (e.g., WEB-42) to avoid ambiguity.
3. Be cautious with bulk actions — Bulk operations affect multiple items. Review the full list before confirming.
4. Modify rather than reject — If an action is mostly correct, send a follow-up to adjust it rather than starting over.
5. Check action history regularly — Periodically review the action history to ensure AI-driven changes align with team expectations.

Related pages

• SetGet AI Overview — General AI capabilities and architecture
• AI Chat — Using the thread-based AI conversation interface
• AI Credits — Credit consumption for actions
• Work Items — Understanding the items AI can create and modify
• Security — Workspace roles and permissions