Colsubsidio is a company with more than 60 years in the market, currently has more than 16,000 employees. One of the current problems for the organization is the leak that arises in terms of knowledge, and the losses that occur in different factors, money, resources and time are invested during the training process. This company doesn´t have action plans, contingencies or knowledge management models, the absence of these prevents the retention of the knowledge that the employee has acquired during the course of the organization. The study that was carried out sought in the first instance to consolidate the data of the different information systems to structure a consolidated database that groups the historical socioeconomic and organizational information of the workers who have and are working in the company. With the information collected, algorithms were created that allowed to identify clusters and key variables that have a high incidence in the rotation of personnel, with the help of these algorithms, predictive models were constructed that allowed calculating the probability of retirement of active personnel in the company. The results provided by the model helped to propose and design a model for the transfer of knowledge that can be used in people who have a high probability of withdrawal, so that in this way the knowledge leakage can be reduced