Abstract A concept of Z-number was introduced by Prof. Zadeh as a formal construct to describe imprecision and partial reliability of real-world information. Computation with Z-numbers inherits an important problem of fuzzy arithmetic – increase of entropy that damages informativeness of results. This becomes especially important for computation with a large number of Z-numbers which takes place in many practical problems of decision analysis. In this paper we propose an approach to compute a sum of a large number of Z-numbers with a predefined level of informativeness of a result. An example is used to show validity of the approach.