Artificial intelligence is transforming archival practice, offering new possibilities for processing, describing, and making accessible the vast holdings that no institution could manage manually. This keynote surveys the current state of AI adoption in archives - from handwritten text recognition and named entity extraction to audiovisual analysis and appraisal support - while insisting that technology alone is insufficient. Archivists must understand the tools they deploy, maintain accountability and transparency, and actively manage risks such as hallucination, bias, and data protection.