AI Prompt Step in Approval Path for Confluence

On this page

1 What the AI Prompt step can do
2 How it works
3 When to use it
4 How to add an AI Prompt step to an approval path
5 Usage
6 AI Prompt step fields
7 Recommended placement in the approval path
8 Improve AI results with the definition description

AI Prompt Step Documentation

The AI Prompt step adds AI-powered analysis to your Approval Path for Confluence workflows.

When an approval reaches this step, Approval Path can send selected Confluence page data, approval history, previous decisions, comments, page metadata, and related approval context to the AI provider configured by your Confluence admin. The AI then returns a structured analysis that approvers can read directly on the approval timeline.

You can use the AI Prompt step to help approvers understand a page more quickly, identify risks, verify policy compliance, summarize content, detect missing information, or prepare a recommendation before a human decision is made.

The AI Prompt step is advisory. It does not automatically approve or reject anything, and the approval path continues even if the AI analysis fails or is skipped.

What the AI Prompt step can do

The AI Prompt step can help your approval workflow by generating:

a short summary for the approver
a risk assessment
a policy compliance check
a decision recommendation
a briefing for an external reviewer
anomaly detection based on similar past approvals
a custom analysis based on your own prompt

The result is displayed as a structured card on the approval timeline. It can also be reused later in the workflow by Automation steps, HTTP steps, webhooks, and email message templates.

How it works

When an approval reaches an AI Prompt step, Approval Path for Confluence:

Collects the relevant approval context
Reads the Confluence page data allowed by the tenant’s data-sharing settings
Applies privacy and anonymization rules
Sends the prompt and selected context to the configured AI provider
Stores the AI response on the approval step
Displays the result on the approval timeline

The AI call runs in the background. While the analysis is being generated, approvers see an Analyzing… state. Once the result is ready, it appears automatically on the timeline.

The step is non-blocking. If the AI provider is unavailable, the budget is reached, or the prompt cannot be processed, the approval can still continue normally.

When to use it

Use the AI Prompt step when approvers need extra context before approving a Confluence page. Common use cases include:

reviewing documentation before publishing
checking whether a page follows an internal policy
summarizing long pages for busy reviewers
identifying missing sections in technical, legal, or process documentation
preparing external reviewers with a self-contained briefing
detecting unusual approval patterns compared with similar past page approvals
helping approvers understand previous comments and decisions

How to add an AI Prompt step to an approval path

To add an AI Prompt step:

Open an approval path definition.
Click Add step.
Select Prompt from the step picker.
Enter a step name.
Choose a template or start from a blank prompt.
Adjust the prompt text if needed.
Add conditions or label conditions if needed.
Save the definition.

Usage

Use the Usage tab to monitor AI consumption.

The Usage tab shows:

tokens used this month
configured monthly budget
percentage of budget consumed
total executions
usage by approval path definition

When the monthly token budget is reached, AI Prompt steps fail. The step conveys a clear, actionable message: the admin needs to either raise the budget or wait until next month.

A failed AI Prompt step does not fail the approval. The path continues; the AI's analysis is advisory, not gating.

AI Prompt step fields

When configuring an AI Prompt step, you can set the following fields:

Field	Description
Name	The name of the AI Prompt step. This is also used when referencing the result later in automations, HTTP steps, or message templates.
Start from a template	Lets you choose a built-in template, a custom template, or a blank prompt.
Prompt	The actual instruction sent to the AI provider.
Condition	Optional condition that controls whether the step runs.
Label condition	Optional label-based condition.

AI Prompt step names must be unique within the same approval definition. Use clear names such as:

Page summary
Risk assessment
Compliance check
Publishing recommendation
External reviewer briefing

These names are easier to reuse later in automation templates.

Recommended placement in the approval path

Place the AI Prompt step where its analysis is most useful. For example:

Put a Summary for approver step before a page owner approval
Put a Risk assessment step before a compliance or legal review
Put a Policy compliance check step before publishing controlled documentation
Put a Decision recommendation step after reviewers have already added comments
Put an External reviewer briefing step before an email approval step

If one AI Prompt step should use the result of another, place them in sequence. Example:

AI Prompt step: Page summary
AI Prompt step: Publishing recommendation
User approval step: Documentation owner approval

The second AI Prompt step can use the result of the first one, provided the first step is placed earlier in the same approval definition.

Improve AI results with the definition description

The approval path definition description is sent to the AI as the business purpose of the approval. A clear definition description can improve the quality of the analysis. For example:

This approval is used before publishing customer-facing product documentation. Reviewers should check whether the page is accurate, clear, complete, and safe to publish externally.

This gives the AI more context about what the approval is meant to decide.