AI in Public Sector: Japan & UK Tackle Productivity

News

Commentary

In an unstable and complex geopolitical and economic context, countries are striving to optimize their resources and capabilities to maintain, improve, or sustain public services with increasingly reduced budgets. This necessitates doing more with less—a challenge ideally suited for Artificial Intelligence. This appears to be beginning in many countries. Japan was the first to position itself at the forefront of this new paradigm, applying AI to administrative tasks within its public administration and as assistance in managing cities and prefectures, leading to a stance that favors the deregulation of AI.

Japan’s commitment to artificial intelligence in the public sector was formalized in April 2023, when Prime Minister Fumio Kishida met with Sam Altman, CEO of OpenAI, to explore the adoption of ChatGPT for bureaucratic tasks. Unlike other countries that opted for temporary moratoriums, the Japanese government expressed its intention to actively study the use of generative AI in ministries and agencies, provided that its implementation does not negatively impact public employment. Digitalization Minister Taro Kono then voiced his desire to "actively consider" these tools, positioning Japan at the forefront of institutional AI adoption.

In addition to Japan, the United Kingdom is joining this group of countries implementing AI in their public sector to enhance productivity and capabilities. Indeed, a hackathon organized by PwC and supported by Microsoft has been held, in which over 100 government officials explored ways to apply AI to improve efficiency in areas such as policing, public contract management, and customer service. According to UK Foreign Secretary Jeremy Hunt, the goal is to increase productivity and save up to 38 million hours per year in police bureaucracy through the use of AI and other early interventions.

The collaboration between OpenAI and the UK government has recently intensified. In March 2026, both parties signed a memorandum of understanding to accelerate the adoption of AI in public services and the private sector. The agreement entails the deployment of advanced models in public administration, infrastructure development, and technical information exchange. OpenAI has already developed a chatbot based on its API for the GOV.UK portal, enabling small businesses to access guidance on commercial regulations. Additionally, its technology supports internal tools such as "Humphrey", an AI assistant designed to reduce administrative burden in Whitehall, and "Consult", which automates the classification of public responses to government consultations.

However, productivity does not depend solely on the implementation of advanced technologies such as AI, but also on the improvement of basic IT infrastructure, investment in capital, process optimization, and personnel management. According to FT, the productivity challenge in the National Health Service (NHS), where AI technology is already being used to streamline processes such as radiograph analysis and clinical note-taking, remains significant. Nevertheless, it is noted that some hospitals still operate with paper records and outdated IT infrastructure, hindering the effective deployment of AI. NHS leaders emphasize the importance of addressing fundamental issues—such as staff burnout, social care needs, and investment in basic IT systems—before fully realizing the benefits of AI.

This contrast between the sophistication of AI solutions and the obsolescence of underlying infrastructure is a phenomenon we have observed in other domains. The effectiveness of language models, no matter how advanced, ultimately depends on the quality of input data and the capacity of institutional systems to integrate them into real-world workflows. A medical image analysis model will have limited impact if healthcare centers lack standardized digitalization systems.

The UK financial sector, one of the most dynamic in the world, is also embracing the study of AI to optimize its operations, as indicated by the Bank of England. According to its report, 72% of surveyed companies are using or developing ML (Machine Learning) applications, and these are becoming increasingly common across various business areas. This trend is expected to continue, with an estimated 3.5-fold increase in the average number of ML applications over the next three years. The insurance and banking sectors are anticipated to experience the greatest growth in the use of these applications. However, their implementation is not without risks, including those observed by consumers related to bias, data representativeness, and privacy, while for companies, the primary risks are the lack of usability and interpretability of ML applications, due to entrenched reliance on legacy IT systems (software and operations).

The experience of the UK financial sector illustrates a paradox we have previously highlighted when discussing the implementation of AI in the insurance sector: organizations with complex legacy systems face barriers that are not only technical but also organizational and cultural. The “lack of usability and interpretability” noted by the Bank of England refers to the difficulty of integrating machine learning models—whose internal logic is opaque even to their developers—into business processes that have traditionally relied on explicit rules and documented traceability.

Japan’s commitment to deregulation and the United Kingdom’s emphasis on public-private collaboration represent two distinct models for integrating AI into public administration. However, both share a fundamental premise: the digital transformation of the public sector cannot be limited to the adoption of isolated tools, but requires a comprehensive review of infrastructure, processes, and staff competencies. As the NHS itself has noted, AI does not replace the need for investment in basic systems; rather, it presupposes them.