Instructional strategies employing complex problem-solving examine how learners represent the problem space and argue for decisions. Unlike problems with few solutions (e.g., multiple choices), instructors struggle to evaluate solutions to ill-structured problems quickly and consistently across students. This prevents instructors to provide timely feedback to learners, who may continue to engage in erroneous thinking. Despite the availability of several software packages to assess the learners' works as maps, several barriers still prevent the wide-scale adoption of such assessment systems by instructors. We examine three open questions and identify opportunities from Artificial Intelligence to develop the next generation of assessment systems.