Knowing one’s own negotiation style and recognising the partner’s negotiation style are fundamental prerequisite to negotiate effectively. Since novice negotiators often fail to state these styles correctly, a dedicated training to support the identification of negotiation styles is required. In this paper, an AI-based training approach is presented. As a first step, software agents acting as training partners have to be designed. We implemented a genetic algorithm to shape the agents’ behaviour. As a first implication, the designed agents act according to their intended negotiation styles without the need for extensive datasets of previous negotiations to specify those styles.