How to Use AI Effectively to Review Risky Terraform Plan Changes

Terraform plan output is often hundreds of lines long, burying critical changes like database replacements among routine updates, making manual review error-prone. A more reliable approach involves exporting the plan as structured JSON using 'terraform show -json' and filtering it with jq to isolate only delete and replace operations before passing it to an AI model. Rather than pasting raw plan text into a chat, engineers can feed a compact, pre-filtered summary to the AI along with a directed prompt requesting risk assessment, replacement triggers, and a recommended action per resource. This structured input produces concise, scannable AI responses instead of lengthy essays that may gloss over dangerous changes. The method was demonstrated with a sample plan where an AI correctly flagged a database replacement as high-risk due to potential data loss if final snapshots were disabled.