Get Your Data Ready Before an AI Project

Imagine a company that decides to put an AI assistant on top of its contracts, price lists and customer records. The project starts well. Then the team discovers it cannot find the right files.

The problem: "The AI is only as good as what we give it"

An AI project reads your information. If that information is scattered, out of date or locked inside systems that do not talk to each other, the result will be unreliable, however good the tool is.

These problems usually stay hidden until someone tries to use the information for something new. Four of them are worth checking first.

Where it goes wrong

• Scattered files: the same kind of document sits in shared folders, personal drives, email attachments and chat apps, and nobody can say which place is the real one.
• Unclear ownership: nobody is responsible for a given set of documents or records, so nobody can say whether it is complete or correct.
• Outdated versions: several copies of a price list or policy exist, and the old one is as easy to find as the new one. An AI tool cannot tell which one you meant.
• Systems that cannot be connected: information sits in an older system with no way to export it or link it to anything else. It has to be copied by hand.

The turning point: review readiness before you build

Before choosing a tool, look honestly at what the project will depend on. A readiness review does this in a structured way, and it ends with a short list of fixes in order, not a long report.

1. List what the project needs. Name the documents, records and systems the use case relies on, and where each one lives today.
2. Find the owner. For each source, name one person who is responsible for keeping it correct. If no one can be named, that is the first finding.
3. Check the version. Confirm which copy is current, and decide what happens to the others: archive, merge or remove.
4. Test the connections. Ask whether each system can share its information in a usable form, or whether it is a closed box.
5. Order the fixes. Put them in sequence: what unblocks the most, what is quick, what needs a decision from management, and what can wait.

What changes

You get a short list of fixes in a sensible order, each with an owner. Some are small, such as agreeing where the current price list lives. Others need a decision, such as replacing a system that cannot be connected.

The project then starts on firmer ground. Surprises show up on paper rather than halfway through the build, and the fixes help the business even if the AI work comes later.

Tidy information is not a luxury for AI. It is the raw material.

What this means for your company

Pick the use case you most want to start with, and ask four questions. Where does the information live? Who owns it? Is it the current version? Can the system share it?

If any answer is unclear, a readiness review is the sensible first step. It is part of my AI readiness and strategy service, which includes a data and systems readiness review and a prioritised plan. It is the same approach I apply in the AI solutions I have delivered for large enterprises and public-sector organisations over the past year.

If this sounds familiar and you would like to talk it through, the WhatsApp button on this page is the quickest way to reach me.