In the domain of naval command and control and warfare, the complex and dynamic nature of the environment makes that a warfare officer, who has to compile a tactical picture of the situation, has to deal with a large number of tasks in parallel. This study aims at investigating the use of a cognitive model of visual attention as part of the design of a software agent that supports such officers. In order to support, agents can take over part of the parallel tasks. However, a problem is how to determine an appropriate work division between agent and its user: due to the rapidly changing environment, such a work division cannot be fixed a priori . This results in the need for dynamic task allocation, determined at runtime. For this purpose, two approaches exist, namely human-triggered and system-triggered dynamic task allocation . In the former case, the user can decide up to what level an agent should assist him. But especially in alarming situations the user does not have the time to think about proper task allocation . In these situations it would be better if the agent determines the allocation of tasks. Hence a system-triggered dynamic task allocation is desirable.