Data Readiness Checklist for AI Pilots

Short, actionable items to confirm you can run a timeboxed pilot.

• Data sources identified: List each source and owner.
• Access path: Confirm where data will be extracted from and who grants access.
• Sample availability: Can you get a representative sample (weeks of data) within days?
• Labeling plan: Do you need labels? If yes, outline how many examples, who labels, and estimated time.
• Data quality checks: Run basic checks for missing values, malformed records, and duplicates.
• PII/regulatory review: Identify personal or regulated info and required safeguards or approvals.
• Versioning & provenance: Record which data snapshot you’ll use and how you’ll reproduce experiments.
• Storage & compute: Confirm storage location and compute capacity to run experiments within timeframe.
• Data retention & deletion plan: How long will pilot data be kept and how will it be disposed?
• Fallback dataset: If real data isn’t available, do you have a synthetic or anonymized alternative to run an initial test?

Use this checklist before you begin the pilot and again at the mid-pilot review to confirm no drift in readiness.