CubeSat missions are constrained by the limited resources provided by the platform. Many payload providers have learned to cope with the low mass and power but the poor telemetry allocation remains a bottleneck. In the end, it is the data delivered to ground which determines the value of the mission. However, transmitting more data does not necessarily guarantee high value, since the value also depends on the data quality. By exploiting fast on-board computing and efficient artificial intelligence (AI) algorithms for analysis and data selection one could optimize the usage of the telemetry link and so increase the value of the mission. In a pilot project, we attempt to do this on the Puerto Rico CubeSat, where science objectives include the acquisition of space weather data to aid better understanding of the Sun to Earth connection.