Fujitsu collaborated with a central government agency in Japan to run a demonstration experiment using its Takane large language model in public comment operations, automating tasks such as classifying opinions by support or opposition and summarizing them. The experiment, conducted in 2025, showed that advanced analysis tasks previously performed manually could be automated while maintaining quality, and officials at the agency confirmed the effectiveness of the system. Building on these results, Fujitsu is developing a generative Artificial Intelligence service that will support policy formulation and legislation, with the aim of offering it by fiscal year 2026, and is planning Artificial Intelligence workflows that integrate suitable models and tools throughout the legislative drafting process.
In conventional Japanese governmental public comment operations, officials read and classify submitted opinions, analyze trends, draft responses, and then consider policy, and this process can take over a month until the results are announced. In the demonstration experiment, Fujitsu applied Takane to approximately 120,000 characters of actual public comment data received by the participating central government agency, and tasks such as classifying opinions by support or opposition and summarizing them, which were previously performed manually, were automated and completed in about 10 minutes. When checking consistency between draft laws and individual opinions, Takane was able to correctly identify the relevant clauses in the draft law for over 80% of the opinions when both the draft law and opinions were input, indicating major potential for labor savings compared to manual linking. These outcomes suggest that reducing time spent on organizing opinions will allow administrative officials to focus more on substantive deliberation and policy reflection, which aligns with the goals of Evidence-Based Policy Making.
Public comment is a key mechanism for national and local governments to gather direct input from citizens affected by rules, but high-interest topics can generate many thousands of submissions that strain administrative capacity and risk undermining trust if not handled thoroughly. The central government agency involved in the pilot sees digital technology as essential to easing officials’ workloads while improving how public views are reflected in policy, and it selected public comment aggregation and analysis as a promising use case highlighted by the Digital Agency’s verification of generative Artificial Intelligence utilization. Fujitsu positions Takane, co-developed with Cohere Inc., as a core technology for such applications and plans to support government Artificial Intelligence infrastructure initiatives, such as Government Artificial Intelligence, by delivering services that leverage robust Japanese language capabilities. More broadly, Fujitsu ties this work to its corporate purpose of contributing to the Sustainable Development Goals and notes that it reported consolidated revenues of 3.6 trillion yen (US$23 billion) for the fiscal year ended March 31, 2025 from its portfolio spanning Artificial Intelligence, computing, networks, data and security, and converging technologies.
