My first job out of school was in defense operations research. Not at RAND, so maybe of limited relevance. But I found it less like modern data science and more like GOFAI. The work was mostly coming up with hand engineered models of military operations and then guesstimating the crucial variables since they were impossible to actually measure. I also observed that the analysis was rarely independent so it could inform policy in an unbiased way but instead was shaped to support particular policy choices.