Many AI projects stall for the same reason: the model works, but the data it needs is scattered, out of date or open to the wrong people. Before you pick a use case, it’s worth answering six questions about your data.
1. Do you know where the data lives?
List the systems that hold the information the use case needs: the CRM, the ERP, file shares, email, spreadsheets. If nobody can draw this map, start there. It’s the cheapest step and it often surfaces quick wins on its own.
2. Is there one version of the truth?
If the same customer has three different addresses in three systems, an AI tool will repeat the confusion with confidence. Decide which system is the source for each kind of record, and sync the others from it.
3. Who is allowed to see what?
An assistant that answers questions over your documents should only use what the person asking is allowed to see. That depends on permissions being right in the first place. Review access on your most sensitive libraries before connecting anything to them.
4. Does the data stay where it should?
Know where each service you use processes data, and whether that meets your regulator’s rules. Building on services inside your own Azure tenant keeps that question simple.
5. Can you measure the result?
Pick a number before you start: hours spent on a task, time to answer a customer, forecast error. Without a baseline, a pilot can’t prove it worked, and it won’t get funding to scale.
6. Who will own it?
Models and dashboards need owners, just like any other system. Agree who checks the quality and who decides when it needs changing.
Next step
If you can answer four of these six with confidence, you’re ready for a pilot. If not, the gaps are your first project. Our data review works through these questions with your team.