The purpose of this study is to significantly reduce the lack of liquidity of small and medium-sized enterprises in Mexico City, by reducing credit risk management through a model based on Machine Learning to establish credit policies based on their characteristics, behavioural patterns, and payment capacity of customers without compromising the financial situation of the business. For the development of the research, an exhaustive review of the state of the art of credit risk management of companies and banking institutions worldwide was carried out, identifying the best practices of credit risk management, as well as successful cases and cases that were not so successful or failed in their application. Subsequently, a compilation of machine learning models was carried out to analyse the variables that are considered, mathematical calculations, company classification, and linear or multivariate regressions and to apply them to SMEs in Mexico City. Finally, some tests were carried out with a sample of clients of a specialised collection agency to determine the accuracy of the proposed machine learning model and to quantify financially the savings obtained in implementing the proposal.