We propose a new approach to capture and interpret high-level optimization intents from users using natural language such as plain English. Such interpretation aims to obtain a complete self-service and hands-free approach, where the user can specify optimization problems in high-level terms, with minimal or inexistent (optimization) domain specialist intervention. This framework should reduce the cost of implementation and deployment of new optimization problems, and streamline the operations. It can levarage modern Large Language Models, aied by domain specialists.