Define your methodology

It may be a good idea to state ahead of time the methodology you plan to use. The degree to which you will specify it will depend on the technical level of the people you are having the discussion with. You may describe your plan's methodology at a high level with statements like "a classification model will be trained…," or maybe you are required to be more specific: "Method XYZ will be used to scale the features… then an SVM with a linear kernel will be trained…"; make sure that you explain the reasons you have for proposing your methodology.

A word of advice: Don’t use this as an opportunity to brag about your technical knowledge; you may come across as pedantic or as someone who doesn’t know how to communicate with stakeholders. Keep in mind that technical details are often not important for most of the nontechnical business people; they want the solution and they do not care about your fancy methodologies or how advanced and complicated your deep convolutional neural network (CNN) is. Keep in mind one of the golden rules of human communication: Know your audience, and if they understand and require technical details, then provide them; otherwise, keep the discussion at a high level.

Of course, it will be impossible to anticipate every detail of the procedure you will actually follow when producing the model. The proposed methodology, rather than a complete plan, will be just a sort of initial guide for what you will do, and when you get your hands on the data you may decide to proceed in a different way.