But can it really capture all the locality-specific factors that are needed for good advice, the diversity of the farms and farmers, their labour constraints, their soil types, their skills and their market access? 

Let us assume that the ICRISAT app can work perfectly well – perhaps it will. Even then, it can be used to illustrate my second point. The app envisions to automatically guide farmers through the farm season. But do we really want this? Do we want farmers that merely follow commands, farmers that function like factory hands? Or do we want farmers who have the knowledge and skills to question such advice? Farmers that understand why some advice makes sense and then follow it, but who also have the knowledge and confidence to readjust, question and resist bad advice. One may say that this line of argument underestimates the capacities of farmers or that apps merely support decisions.