From AI Models to the Harsh Reality: My Company's Struggling Database

In the early stages of implementing AI, I aimed to automate supplier payments using a seven-year SQL data history from my company, but I quickly encountered data inconsistencies such as supplier name variations, which complicated the analysis.

This led us to simplify the model by focusing on the remaining reliable columns, but even then, data misclassifications, like profit & loss accounts, hindered progress, showing that the data was insufficient for a reliable model.

Despite not achieving full automation, the project led to a complete restructuring of our database, highlighting the importance of high-quality data, a realization that now shapes my consulting approach to AI readiness in companies.