Embodied conversational agents (ECAs) are virtual characters using verbal and non-verbal communication for Human-machine interaction. The aim of our research is to create an ECA-based user interface for assistive technologies targeting older adults with cog-nitive impairment. Our design methodology is a co-design living lab approach, collecting design guidelines through questionnaires, focus groups and user trials. In this paper, we report on the results of the first phase of this iterative design process. We developed Louise, a semi-automatic ECA prototype that aims to compensate, through attention monitoring , for a user's attentional disorders by performing autonomous prompting, i.e., calling the user to regain his or her attention in case he or she got distracted. We evaluated the performance of Louise with a group of experts in assistive technologies and collected their feedback. Louise's simple attention estimator is more than 80% accurate. The system got quite positive reviews from users.