What Data Do You Need Before Implementing AI? A Non-Technical Guide

Before implementing AI, you need three things: data that actually records the process you want to improve, enough of it to show patterns (often just a few hundred examples), and access to it in a usable form.

You Need Less Data Than You Think

Stories about training giant foundation models on billions of examples make business owners believe AI requires massive datasets. In reality, fine-tuning or RAG retrieval requires only your existing SOPs, PDFs, and SQL tables.

The 4-Point Data Readiness Checklist

• Existence: Is the information recorded digitally anywhere?
• Access: Do you have API or database query access to export it?
• Quality: Are the records consistent, or filled with missing fields?
• Permissions: Is customer PII sanitized and privacy-compliant?