The “show me first” pattern for high-risk operations

Why some tasks need extra caution

You’ve learned to write clear prompts with the three essentials and quality boosters. These work well for single items or low-risk changes. But consider this scenario: you need to adjust product descriptions across 50 items to emphasize outcomes and benefits instead of features and specifications. You write a prompt: “Update the Product Description element in all Product content items to focus on what our product helps customers achieve, rather than how it works.“ The AI Agent executes immediately. Then you discover it updated products you didn’t intend to change. Or the tone shift doesn’t match your brand voice. Or 50 turned out to be 75 items. What started as a targeted update becomes a broad change that takes extra time and resources to reverse. The “show me first” pattern prevents this. Instead of jumping straight to changes, you verify the scope before the agent acts.

High-risk operations that need the safety layer

Some operations carry significant consequences if executed incorrectly. Always use “show me first” for:

• Bulk updates – Changing content across multiple items simultaneously. The more items affected, the higher the risk.
• Publishing and unpublishing – Making content live or taking it offline affects what your audience sees. Mistakes here are immediately visible externally.
• Deletion – Removing content is irreversible. Deleted items may not be recoverable.
• Workflow changes affecting multiple items – The agent follows the workflow rules, but acting at scale means a small mistake can shift dozens of items into the wrong step.
• Schema changes – Modifying content types or elements affects your content structure. These changes cascade across all content using that type.

When you can skip the safety pattern:

• Single item operations where you’ve specified the exact content item by name
• Discovery tasks that don’t make changes (finding and listing content is inherently safe)
• Moving a single, named item to a different workflow step
• Simple, reversible changes to items you’re actively working with and can easily verify

Principle 6: Start with “Show me first”

For high-risk operations, verify what will be affected before making changes. The “show me first” pattern gives you control by breaking risky operations into verification and execution phases.

The four-step safety pattern

The safety pattern breaks risky operations into verification and execution phases.

1. Show me what will be affected – Use discovery verbs (Find, List, Show) to see which items match your criteria.
2. Review the results – Verify the scope matches your expectations. Check the number of items and confirm they’re the right ones.
3. Suggest the changes – Ask the AI Agent what it would do. Review the proposed changes before they happen.
4. Implement after verification – Once you’ve confirmed everything looks correct, tell the AI Agent to proceed.

This approach gives you two verification points: confirming which items will be affected, and confirming what changes will be made.

See exactly what changed

After you implement the agent’s changes, verify them before publishing. An expandable bar appears in the chat for each modified content item. With a quick view, you’ll see which elements were affected with highlighted differences without opening each content item individually.

Example: Content audit and cleanup

After using Kontent.ai for several months, you need to find outdated, stale content to streamline your content operations and prevent content inventory bloat before it happens. This is a common, high-value use case for the AI Agent. 1. Show me: “Find all Article content items in the News collection in the Draft workflow step that have not been modified in the last 30 days.” AI Agent returns: “Found 40 Article drafts that haven’t been modified in more than a month.“ 2. Review: Forty items. Based on your team’s monthly content pace, this number of articles seems right. To validate further, you ask the agent: “Provide direct content item links so I can review.”

3. Validate: You review a subset of the returned articles to confirm they match your expectations. 4. Suggest: “Suggest which articles we should refresh because they relate to AI, and which we should archive because they don't align with our current AI focus.” The AI Agent provides a summary of 15 articles to refresh and 25 to archive. 5. Implement: “Move the 25 archival recommendations to the Archived workflow step.” Because you verified the scope and reviewed the approach on sample items, you’re confident the bulk operation will work correctly. And for the 15 articles to refresh, you could assign them to team members or continue working with the AI Agent to update their content.

Integrating AI Agent into your workflows