Hi,
Are you familiarized with this kind of problems? First of all: To build a classification model you need a dataset of examples with the actual label, so the model can learn how to classify future unseen examples. This is, using a dataset where each row describes a customer (an example) and each column the attributes used for defining that customer, you need a bunch of objects with an additional class attribute that contains the value "yes" or "no" (for example) indicating whether the customer represented in that row bought the project. With this input data you can build the model (it could be a decision tree) using the known information. Then, with the classification model that you have built, you can predict if new customers are potential buyers.
Do you have that prior information?
Noelia