it communicates the outputs from it in such a way that colleagues with a nontechnical background can understand them and have confidence in them.

it justifies the extensive investment required for an ML project.

it helps to hide the uncertainties surrounding analytic results, if the data is not reliable.