AI Plan Summaries & Rollback Suggestions#

Reply CMP has two “contextual” AI features that appear at the right moment in your workflow — not as a chat panel, but as inline summaries generated from live operation output.

AI Plan Summaries (Provisioning)#

Where They Appear#

AI Plan Summaries appear at Step 4 of the provisioning wizard, in two moments:

  1. After Dry Run — when you click “Run Dry Run”, the Terraform plan runs and an AI summary appears below the terminal output.

  2. After Apply — when the apply completes (successfully or with failure), a summary appears.

What a Dry Run Summary Says#

The Dry Run AI summary tells you, in plain English, what the Terraform plan is about to do:

“This plan will create one Azure Virtual Machine (Standard_B2ms) and one managed disk (128 GB) in the eastus region. One network interface and one public IP address will also be created. No existing resources will be modified or deleted.”

The summary is intentionally brief (2–3 sentences) and designed for:

  • Confirming you are creating what you intended

  • Sharing with a teammate or change approver who is not reading raw Terraform output

  • Catching major surprises before applying (e.g., “this plan will delete 3 resources” when you expected only creates)

Output Types#

Moment

AI output type

Purpose

After Dry Run

Plan Preview

Summarises what will be created, modified, or deleted

After Apply (success)

Apply Summary

Confirms what was created, with resource IDs if available

After Apply (failure)

Error Explanation

Plain-language failure reason and suggested corrective action

AI Error Explanations#

When an apply fails, the AI Error Explanation panel shows:

  • A plain-language description of what went wrong (e.g., “The deployment failed because the service principal does not have the Contributor role on the target subscription.”)

  • A suggested corrective action (e.g., “Grant Contributor access to the service principal with Client ID [id] on the subscription [name].”)

  • The raw Terraform error output remains available for detailed investigation.

Note

AI Error Explanations are generated from the Terraform error output and may not always identify the root cause correctly. If the suggested action does not resolve the issue, review the raw Terraform output and consult your cloud provider’s documentation.

Warning

AI Plan Summaries simplify complex plans. For plans involving many resources, modifications, or destructions, always review the raw Terraform plan output before clicking Apply.

Rollback Suggestions (Change History)#

Where They Appear#

In the Discovery → Resource Detail → History tab. After selecting a specific change event (a diff showing what changed between two discovery snapshots), a Rollback Suggestion button appears.

Triggering a Suggestion#

  1. Open a resource in Discovery.

  2. Click the History tab.

  3. Click on a change event to expand the diff.

  4. Click “Get AI Rollback Suggestion”.

The AI reads the diff (before/after property values) and generates a rollback suggestion.

What a Rollback Suggestion Looks Like#

The suggestion provides provider-specific rollback commands.

Azure example:

# Restore VM size from Standard_D4s_v3 back to Standard_B2ms
az vm resize --resource-group myRG --name myVM --size Standard_B2ms

# Or via Terraform:
terraform import azurerm_virtual_machine.main \
  /subscriptions/.../resourceGroups/myRG/providers/Microsoft.Compute/virtualMachines/myVM
terraform apply

AWS example:

# Revert EC2 instance type from m5.xlarge to t3.medium
# Note: instance must be stopped first
aws ec2 modify-instance-attribute \
  --instance-id i-1234abcd \
  --instance-type "{\"Value\": \"t3.medium\"}"

The suggestion format uses the provider’s native CLI tool (Azure CLI / AWS CLI / gcloud) and, where applicable, a Terraform import approach.

Caveats#

Warning

Rollback suggestions are advisory only. They are generated from the discovered metadata diff and may not account for:

  • Dependencies on other resources

  • Changes to data or state inside the resource (e.g., database contents)

  • Additional steps required before the change is safe to apply

Always review and validate rollback suggestions in a non-production environment before applying to production.

Note

Rollback suggestions are not available if:

  • The resource was deleted

  • Fewer than 2 history snapshots exist for the resource

  • The diff involves a change type the AI does not have a template for