Well not all business problem require a ML solution. Sometimes a baseline solution works. There are cases where implementation of an ML algorithm will not yield effective results on the ground, in such a scenario the cost far outweighs the benefits accrued. Basically your ROI is negative.This happens due to many reasons for e.g poor data (in terms of many missing values) or your industry space is too unpredictable. A seasoned data scientist would know the impending pitfalls by seeing the data and by analyzing the business problem at hand. Hope I have clarified.