This work discusses and solves the methodological problem of reducing the
probability of erroneous decisions when solving local problems by groups of intelligent agents.
Intellectual agents can be carriers of both natural and artificial intelligence. The basic method of
the new information technology is based on the theory of evolutionary decision reconciliation
developed by the authors. The method is built on the original use of genetic algorithms in which
intelligent agents perform the fitness function based on their knowledge and skills. The
technology is based on the knowledge model proposed during the work, the Rasch model, and
the mathematical apparatus of the Condorcet theorem. A theorem that allows finding the
conditions under which the probability of an erroneous solution tends to zero, is formulated and
proved. The results of using the theory for objects recognition from images by a group of neural
networks are presented. A significant decrease in the probability of errors of the first kind is
obtained in comparison with the classical methods. It is proposed to use the technology of
evolutionary decision reconciliation to solve local problems in medical diagnostics, banking, and
other fields.