Choose a safer first AI experiment.
Evaluate one repetitive task. Get a practical recommendation, three reasons based on your answers, the principal risk, and an experiment plan you can download or print. No email required.
Nine questions. One task. A plan you can keep.
Your task label and answers stay in this page until you download them. Reloading clears them. This tool does not ask an AI service to process your answers.
See an example of what you get
Illustrative example: drafting an internal meeting summary using public example notes. Suppose inputs, quality rules, review, and reversibility are available; the task is weekly, harm is rated low, and judgment is medium.
Result: Assist, don't automate. Medium judgment remains with a person, even when inputs and quality rules are available.
Next experiment: compare a draft against a human-written version, check accuracy and reviewer effort, and keep every external action human-controlled.
This is an illustration of the output, not a measured customer result or a claim of time savings.
This screening logic is a practical, unvalidated heuristic based on your answers. It is not an AI-readiness certification, security audit, legal/compliance assessment, or proof that a task should be automated. It does not promise time or cost savings. Never enter confidential, customer, health, payment, password, API-key, or other sensitive data.