Telecommunications companies face a growing challenge with the increase in the number of requests for cancellation of residential Internet services (churn), directly affecting their profitability. Identifying the main causes of this evasion is essential to developing efficient customer retention strategies . Personalized customer service emerges as a vital component in this context. This study, based on data from a telecommunications multinational, proposes a heuristic for agent recommendation, using machine learning to improve customer retention, suggesting the appropriate agent based on success stories. All of our models presented an F1-Score equal to or greater than 90%.