Before launching an AI project, look at your data
Many teams want to test AI quickly, which is healthy. The risk appears when you start with the tool before looking at the raw material: available data, business rules, exceptions, file quality and control points.
The question isn’t only technical
An AI assistant can answer, classify, extract or propose a summary. But if the source data is incomplete, contradictory or poorly named, it only makes the problem more visible.
Before choosing a solution, you need to know which sources exist, who updates them, which columns are truly reliable and which decisions rely on this information.
A short diagnostic avoids bad starts
The right reflex is to take a representative sample: a few files, a CRM export, a ticket stream, a monthly report. In it you look for duplicates, empty fields, unstable formats and implicit business rules.
This step may seem simple. Yet it changes the rest of the project, because it shows what can be automated now, what needs cleaning up, and what requires a business decision before any development.
AI comes after the scoping
Once the useful data is identified, you can choose a narrow scope: qualification support, draft generation, document search, consistency checking or summarising customer feedback.
The project becomes clearer for the teams. They know what the tool can do, what it must not do, and where to keep a human validation.
Next step
Turn this reference point into a concrete project
If this topic resonates with a situation in your organisation, a short diagnostic lets us look at the process, the available data, the risks and the right initial scope.