The increase in competition due to the rise of generic drugs forces pharmaceutical companies of branded drugs to optimize their sales resources. Which physicians are most likely to switch and why?
Which physicians should be called and how many times to prevent them from switching the companies’ branded drug to a generic drug?
Given a sales force of particular size, a specific number of calls that can be made a day per salesperson and a specific number of physicians, how many times should each salesperson call with each physician to prevent brand switching and increase the sales of a particular branded drug?
The ideal AI solution had to be explainable yet also accurate. Physicians were segmented and the optimal number of calls to make was defined per segment. Explainable AI has been used to ensure explainable yet accurate brand switching predictions. Interpretable methods such as GA2M were combined with Shapley explanations to provide accurate and explainable brand switching predictions.
The development of an AI solution requires a balance between software development and analysis. Delivering short-term business insights on the one hand and working towards a scalable and repeatable AI solution by building an AI pipeline is challenging.
Delivering a scalable AI solution that can easily be put into production and delivers business value requires constantly iterating between exploration (feedback on business value) and exploitation (integration). Constantly cycle between an analysis phase gathering and identifying relevant business insights, followed by an integration phase in which the analysis that produced the relevant business insights is integrated in a robust and scalable AI pipeline.