In 2025 the Historical Archives of the European Union (HAEU) took the plunge into the world of artificial intelligence (AI) for archival purposes with a pilot study on archival description. The study involved the use of AI to convert files of scanned documents into structured, searchable datasets. The project started with a set of relatively well-structured files from the EEC fonds, namely the agendas of the Council of Ministers.
An external company was engaged to carry out the work based on their expertise and experience in working with European Archives.
As a first step, archivists worked with the external provider to extract information from the agendas. Despite some teething problems in the first round handling different versions of the agendas and summaries, the initial exercise permitted the Archives to set down some guidelines for the company on how to move forward. The pilot study was completed with data extracted from 10,000 pages, comprising approximately 400 files for the years 1972, 1973, 1975 and 1976. The work involved turning electronic files into machine-readable text and extracting key fields to produce a search interface for researchers.
According to Mary Carr, HAEU archivist who oversaw the project, “the potential benefits of extracting structured metadata from archival records for our researchers cannot be underestimated. The plan is to progress further in this direction in 2026”.
While for the moment AI is particularly beneficial for extracting repetitive data, a future challenge will be to see its efficacy with more varied information, such as those found in the private deposits. “We will need to weigh up the pros and cons during this experimental phase and work with archive AI specialists in making this technology more adaptable to the range of files we have in our deposits,” Carr explained. Positive results would permit the HAEU’s archival team to dedicate its time to quality controls and historical and scientific context, leaving the more mundane work to the machines.