That depends on the starting point: How many data sources are there? What is the data quality? What infrastructure exists? Typically, such a project is structured in phases: data expedition and consolidation, exploratory analysis, AI model development, interface development, and production operations.

What matters is that each phase delivers standalone value. The exploratory analysis already yields insights before the first AI model is built. The consolidated data foundation is a gain even independent of the sales interface. Pragmatism beats perfection: better to start with available data and improve iteratively than to wait for the perfect data situation.