DAIJOBU runs Claude Code workshop for Diet secretaries

The workshop grew out of a gap DAIJOBU and Sugiura identified in how political offices work. Assembly offices handle many repetitive tasks — scheduling, meeting records, list management, petition sorting — but the information involved often includes personal data or unpublished material, leaving staff unsure how far AI can be used. A pre-event survey of 17 respondents showed 14 had used ChatGPT but only 3 had tried Claude Code or similar tools, and the most common worry was how much information to entrust to AI.

The session built its content around those concerns. Participants worked through nine uses in political offices, including adding meeting invitation emails to calendars, sorting petitions, preparing for meetings, updating lists, and drafting thank-you emails. For petition sorting, a fictional petition was used to show AI organizing items by field, urgency and responsible desk while people made the decisions and replies.

DAIJOBU's background in software testing and vulnerability assessment shaped the data-handling guidance. The company teaches a four-layer security approach in its corporate training, and at the workshop it applied that thinking to secretarial work: for example, letting AI read but requiring human confirmation for writing and human action for sending and deleting, replacing names and contact details with numbers before handing petitions to AI, and keeping highly sensitive personal information such as medical history out of AI. The organizers framed the message as people setting the destination while AI handles the driving, and suggested three things to try immediately: creating one routine, turning writing rules into a skill, and starting a CLAUDE.md file.