Automatic phonemic transcription tools now reach high levels of accuracy on a single speaker with relatively small amounts of training data: on the order of 100 to 250 minutes of transcribed speech. Beyond its practical usefulness for language documentation, use of automatic transcription also yields some insights for phoneticians. The present report illustrates this by going into qualitative error analysis on two test cases, Yongning Na (Sino-Tibetan) and Tsuut’ina (Athabaskan). Among other benefits, error analysis allows for a renewed exploration of phonetic detail: examining the output of phonemic transcription software compared with spectrographic and aural evidence. From a methodological point of view, the present report is intended as a case study in Computational Language Documentation: an interdisciplinary approach that associates fieldworkers (“diversity linguists”) and computer scientists with phoneticians/phonologists.