This paper proposes a modelization of complex social systems based on the integration of two computational methodologies: fuzzy logic and agent-based modelling. In particular, the objective of this work is to present a methodology to take into account the ambiguity of verbal interactions in learning processes. To this aim, we revise and fuzzify a classical computational model developed by March in 1991 describing how learning processes develop within organizations. We introduce fuzziness in the model in two ways: first, we propose a representation of the judgments of individuals involved in the learning process based on fuzzy sets theory, second, individual preferences are aggregated through fuzzy linguistic connectives. The results obtained through simulations show that learning processes based on verbal interactions make the organization able to better absorb the shocks produced by environmental turbulence.