An enterprise invests in AI while the archives contain inconsistent codes and duplications. I propose to improve first definitions and responsibilities on data.
In the example, the same customer appears with different identifiers. Sophisticated analysis does not automatically correct every problem at the source.
The result should be compared with reliable information. The preparation may be less spectacular than a demonstration, but it makes the process more understandable. Technological progress also depends on organizational work that remains humane and verifiable.