The Digital Agency has published the results of its first usage survey on "Government AI Gen'nai," a generative AI environment being trialed this fiscal year among roughly 180,000 national government employees ^[1]. The survey drew 6,439 responses, with 71.7% of users saying they were satisfied with the tool and 54.4% reporting that it saved them at least one hour of work per week ^[1]. The agency estimates the average time saved per user at about 1.83 hours, or roughly 110 minutes, per week ^[1]. Some respondents said that even when overall working hours did not change, time freed up by the AI allowed them to focus on more substantive tasks, improving the quality of their work ^[1]. Others raised concerns about the accuracy and reliability of AI responses, and some said they did not know how to use the tool effectively; 31.2% of respondents said they rarely or never use Gen'nai ^[1].

In response, the Digital Agency said it will expand an AI agent usage environment that began on October 20, 2026, incorporating feedback on response accuracy and file input/output, and will launch a "Gen'nai Library" within this fiscal year to let employees share effective use cases and prompts across ministries ^[1]. The agency also plans to promote wider adoption among employees who have not been using the tool, while using a set of standards issued in October 2026 to guide appropriate and safe use ^[1]. A second survey is planned for around January 2027 to assess progress ^[1].

Separately, the Digital Agency began providing a prototype "Diet Answer Drafting Support AI App" within Gen'nai to the same roughly 180,000 government employees, starting the same day ^[2]. The tool is designed to help officials search and organize past answers and reference materials and generate draft responses for Diet questions, a process the agency described as highly time-consuming and demanding of precision ^[2]. It was developed with input from 102 experienced officials across 16 ministries ^[2]. The prototype is being tested during the 222nd extraordinary Diet session, with full rollout planned for the ordinary Diet session starting in January, pending results ^[2]. The agency stressed that the AI does not replace employee judgment: officials must verify AI-generated content against source materials for accuracy and consistency, given the risk of "hallucinations," and final policy decisions and answers remain the responsibility of officials and ministers ^[2].

Source credit 出典：「ガバメントAI 源内 利用状況アンケート調査（第1回）の結果を公表しました」（デジタル庁）（https://www.digital.go.jp/policies/genai/f9ed1b1c-b22f-4466-8e50-c1c9adf9f8eb）、PDL1.0（https://www.digital.go.jp/resources/open_data/public_data_license_v1.0）を加工してJapan News (j.apan.news)が作成 出典：「ガバメントAI 源内で「国会答弁作成支援AIアプリ」のプロトタイプ版を提供開始します」（デジタル庁）（https://www.digital.go.jp/policies/genai/b736d2d7-b03b-43f0-9932-eba7da7f8142）、PDL1.0（https://www.digital.go.jp/resources/open_data/public_data_license_v1.0）を加工してJapan News (j.apan.news)が作成 This article is Japan News's own English summary of the Japanese original. It has been translated and edited by Japan News; it is not an official translation, and the Digital Agency did not write or review it.