Historical diaries and daybooks serve as fascinating windows into the past yet analyzing them is generally a time-intensive affair requiring database or text encoding skills. The “Daybooks to Data” project describes an ongoing project to develop a custom AI research assistant specifically designed for historical diary exploration and analysis.  Building on proof-of-concept work with Josephine Bellamy's 1930s Dorchester daybooks, I am working on custom GPT tools that combine natural language processing with knowledge of historical diary conventions to assist researchers in extracting data and insights from these primary sources. The end goal is a set of custom GPTs that apply analytical functions trained on patterns of English-language daybook writing such as named entity recognition, relationship mapping for social network analysis, geographic and temporal pattern detection, and thematic identification and data visualization.  This work-in-progress will report on some important issues including how we validate AI-generated insights against traditional close reading, what standards should guide the development of AI tools for scholarly adoption and what kinds of tools can serve academic researchers, students and community historians alike. I look forward to feedback on and discussion of AI's role in historical research.