This paper presents an analytical model for choosing optimal investment schedules for the development of new systems under various types of risk. Two modes of risk reduction are considered. In the first mode, risk is reduced by gathering information through prototype building or sequential development, where risky parameters are assumed to have unknown but fixed values. The second mode involves an increase in systems development and usage skills through experience and learning, which may reduce the development cost and increase the acceptance of the system among the potential users. The second mode of risk reduction changes the true values of the parameters. Starting with a conceptual multi-dimensional framework for analyzing systems risk, a dynamic decisiontheoretic model for guiding the investment process is developed. The model specifies the level of investment in development activities at any stage, depending on the information gathered from prototypes or parts of the actual system developed to that point. Some properties of global and myopic investment policies are derived. The sensitivity of the level of investment to the accuracy of information is characterized. Experience and learning effects are considered in a simple two-period setting, where familiarity with the development process in the first period reduces the cost of developing the remaining part of the system in the second period. Extensions, testing, and implementation of the model are discussed.