How it works — ARNIE AI

You describe what you want

It starts in plain language. Type a request the way you'd say it to a teammate — a policy to enforce, a config to change, a node to remediate. No YAML, no module names. ARNIE's job is to understand intent, so the interface is simply your words.

Input · natural language

ARNIE generates the playbook

ARNIE translates your intent into a real Ansible playbook — the actual tasks, modules, and parameters. Route generation through the AI provider you choose. The output isn't pseudo-code; it's a runnable artifact, ready for review.

AI · intent to Ansible

03 · The human gate

You approve — the line that never moves.

Here is where ARNIE stops. Every generated change waits for a human. You see the full playbook and exactly what it will affect. Approve it, edit it in place, or reject it. Nothing proceeds without a person saying yes — not optional, not configurable away, not skippable. This gate is the core of how ARNIE is built: an operator you can trust because it never acts alone.

Approve Edit Reject

The change is pushed to Git

On approval, the playbook is committed to your repository — versioned, attributed, part of your history before it ever runs. If you need to know later what changed and why, the answer is already written into source control. The audit trail starts here, automatically.

Versioned · pushed to Git

Ansible Automation Platform executes it

The committed playbook syncs to AAP and runs through your real pipeline — the same automation backbone you'd use by hand, now driven by an approved, AI-generated change. AAP applies it to your live cluster. This is the moment intent becomes infrastructure.

Execution · AAP to cluster

The result returns to you

The outcome comes back to the chat where it began — what ran, what changed, what succeeded. The loop closes. You asked in plain language, a human approved, your pipeline executed, and the result is reported back where you started. Then ARNIE is ready for the next request.

Result · back to chat

This is the loop, closed.

Most AI tools generate and stop. ARNIE carries the request the whole way — generation, approval, execution, result — and brings it home. The closed loop is the hard part, and it's the part that matters. It's powered by RMCP, the engine shared across the BLCKBX platform.

Read the RMCP story on blckbx.io →

A sentence goes in.
Live infrastructure comes out.