AI-Assisted Transcription of Handwritten and Audio Data in Fisheries

Natural resource agencies routinely collect handwritten and audio field data that must be transcribed before analysis, creating a persistent bottleneck in data management. Recent advances in artificial intelligence (AI) may help automate portions of this work, yet their utility in routine fisheries applications remains largely untested. We evaluated commercially available AI-assisted workflows against standard manual transcription. The evaluation considered total processing time and initial transcription accuracy. Across the AI pilot datasets, initial accuracy averaged 97% for handwritten records and 95% for audio recordings. After quality control, the AI-assisted workflows reduced total processing time by approximately 37% for handwritten transcription and 49% for audio transcription. The larger operational datasets produced reductions of approximately 30% and 57%, respectively. Initial accuracy for the larger handwritten dataset declined to 82.3%, but all identified errors were corrected during quality control. These findings show that commercially available AI can improve a time-intensive aspect of natural resource data management without specialized programming or custom software. Incorporating AI-assisted transcription into existing workflows can reduce administrative workload and move field observations into management and research applications sooner.