Is your data ready for AI? A simple checklist
By Rafael · Updated · 3 min read
The short answer
AI is only as good as the data behind it. Before you build, check five things: the data exists, it's accurate and current, it's somewhere you can reach, you're allowed to use it, and someone owns keeping it right. If any answer is no, fix that first. It's quicker than fixing an AI project that gives confident wrong answers.
Why data comes first
AI doesn't know your business. It works from what you give it: your documents, records, price lists and past examples. If those are out of date, scattered or wrong, the AI will be confidently out of date, scattered or wrong too.
In February 2025, Gartner predicted that through 2026, organizations will abandon 60% of AI projects that aren't supported by AI-ready data. In its survey of 1,203 data management leaders, 63% said their organization either didn't have the right data management practices for AI or wasn't sure.
A simple example: an AI assistant that answers customer questions from last year's price list will quote last year's prices, politely and confidently. The fix isn't a better AI. It's an up-to-date price list.
The five-question check
- Does the data exist?Is the information the AI needs written down anywhere, or does it live in people's heads?
- Is it accurate and current?Would you hand it to a new employee as the truth today? Check a sample of 20 records or documents.
- Can you reach it?Is it in one place, or spread across inboxes, spreadsheets and old folders? AI works best from one trusted source.
- Are you allowed to use it?Does it include personal, confidential or customer data? Check your policy and contracts before it goes near an AI tool.
- Does someone own it?Who keeps it right when things change? Without an owner, good data goes stale within months.
AI data readiness check What will the AI do? ______ What information does it need? ______ 1. It exists, written down: Yes / Partly / No 2. It's accurate and current (20 checked): Yes / Partly / No 3. It's in one place we can reach: Yes / Partly / No 4. We're allowed to use it with AI: Yes / Partly / No 5. Someone owns keeping it right: Yes / Partly / No Owner: ______ Any "No"? Fix that before building. Any "Partly"? Fix it during the first weeks of the pilot. Biggest gap: ______ Fix by: ______
Quick fixes that make the biggest difference
- Pick one source of truth for each kind of information, and retire the copies.
- Clean the most-used records first. Fixing the top 100 often matters more than the next 10,000.
- Remove duplicates and archive outdated documents so the AI can't find them.
- Write down who updates what, and when.
- Take personal details out of anything you use for testing.
Your data doesn't have to be perfect. It has to be good enough for one clear job. Test that with a set of real examples before you scale, as in our guide to AI pilot exit criteria, and check data readiness alongside the other five areas in Pilotwise.
Common questions
- What does AI-ready data mean?
- Data that exists, is accurate and current, can be reached in one place, is allowed to be used with AI, and has an owner who keeps it right, for the specific job you want AI to do.
- Does my data need to be perfect before I use AI?
- No. It needs to be good enough for one clear job. Start with the information that job needs, fix the biggest gaps, and test with real examples.
- Can AI clean up my data for me?
- It can help, for example by spotting duplicates or tidying formats, but a person who knows the business has to decide what's correct. Check its work.