Due to technical advances and the changing political environment sensor management has become increasingly knowledge intensive. Aboard navy ships however, we see a decrease of available knowledge, both quantitative and qualitative. This growing discrepancy drives the need for automation of sensor management. Since the goal of sensor deployment is to have a complete and accurate operational picture relative to the mission we propose a three-stage sensor manager, where sensor task requests are generated based on the uncertainty in the (expected) objects’ attributes. These tasks are assigned to available and suited sensors, which in turn are fine-tuned for the task at hand. When trying to reduce the uncertainty in the classification solution one must first define how the classification process actually works. We discuss why the classification process needs to be automated as well and show how such classification algorithms will most likely work in the future.