In a recent speech, Microsoft CEO Satya Nadella evoked the lofty power of artificial intelligence (AI) to accelerate progress, prosperity, and standards of living. “This technology reaches everyone in the world,” he said. Generative AI takes an idea born in tech communities such as Silicon Valley and Tel Aviv—the ability of algorithmic engines to synthesize vast troves of data to generate text, images, and more— and gives it to governments across Asia to turbocharge service delivery to citizens.

Myriad use cases and opportunities are already percolating. These include supporting public workers to better perform administrative or research tasks and transforming the citizen experience in areas such as health and education. They also include helping people interact with government in Mandarin, Hindi, Bahasa Indonesia, or any of thousands of regional dialects, removing language barriers and improving countless lives.

The fundamental value of generative AI is to serve as a human “co-pilot.” Within the business of government, this doesn’t mean replacing the role of public employees, but instead augmenting their efforts with high impact and at low cost. This might mean accelerating workers’ ability to find the information they need: for example, quickly searching laws, regulations, and previous reports on a topic to locate an answer or drive new policy directions. AI tools can also help summarize meeting notes or streamline the process of drafting a standard piece of ministerial correspondence.

The Government Technology Agency of Singapore (GovTech) is harnessing the power of generative AI using Microsoft Azure OpenAI service to complete these types of routine tasks. This AI-powered assistance might allow team members to branch out into practical fieldwork, interact with citizens directly, or focus on the more human and strategic aspects of their role.

“It’s about freeing up time for the more value-added human aspects of what it means to work in government,” says Marcus Bartley Johns, Microsoft’s regional director of government affairs and public policy in Asia. With public-sector budgets perpetually under pressure—and Japan and Korea in particular confronting aging populations and a shrinking workforce—he argues that the benefit is less about governments saving money and more about optimizing time so that employees can focus on the most rewarding and high-impact work, and governments are “getting the most out of the people they do employ.”

Good public service delivery runs on high-quality interaction, whether it is a social worker in touch with vulnerable families or a call center operator explaining how to obtain a driver’s license. Already, generative AI technology is inspiring integrated translation tools that enable local communities to understand what government services are available to them in a way that simply wasn’t possible before.

For example, in the Indian state of Haryana, villagers in Biwan are using Jugalbandi, a new WhatsApp-based chatbot, to do everything from apply for pension payments to access university scholarships. The solution “understands” questions in multiple languages, whether spoken or typed, retrieves information on relevant programs—usually written in English—and relays it back in the local dialect. This saves users considerable time, compared with conducting a web search and navigating a maze of links.

Abhigyan Raman, a project officer at AI4Bharat, an open-source language-AI center based at the Indian Institute of Technology Madras, describes the solution as a personalized agent or “chatbot plus plus.” He explains: “It understands your exact problem in your language and then tries to deliver the right information reliably and cheaply, even if that exists in some other language in a database somewhere.”

Introduced in April, the solution now covers 10 of India’s 22 official languages and 171 of approximately 20,000 government programs. Integrations with speech-recognition applications such as Mobile Vaani, an interactive voice platform that supports farmers, are also being considered.

India’s language complexity makes it a test bed for multilingual AI solutions everywhere. Smita Gupta is a lawyer at Bengaluru-based collective OpenNyAI, which aims to extend the chatbot to domestic workers and garbage collectors. “If you can solve and build for India, you can solve and build for the world,” she says.

Within Asia, generative AI also offers a window to a future of personalized, human-centered delivery in health and education. Imagine an AI engine transcribing a conversation between a doctor and a patient, identifying a potential diagnosis, and recommending a prescription, which the practitioner then reviews. This allows the doctor to spend more time talking to the patient—and less time staring at a screen. Similarly, in education, generative AI can help teachers streamline class preparation and quickly assemble lesson plans. Students might gain a high-quality automated tutor that intuitively guides them to improve their work.

In Taiwan, CoolE Bot is a chatbot introduced by the Ministry of Education that enables K–12 students to practice speaking English without the anxiety of standing up in class. Launched in December 2022, and based on the Azure OpenAI Service and other Microsoft technologies, the chatbot assesses pronunciation, accuracy, and fluency. If a student is ever stumped, they can click on a button and the chatbot will suggest a question to keep the conversation flowing.

So far, about 30,000 students a month are using the solution, which is designed to assist Taiwan in its goal to become bilingual in Chinese and English by 2030. “We want to help our students quickly enhance their English skills to compete with other countries,” says Howard Hao-Jan Chen, an English professor at National Taiwan Normal University.

Early concerns about generative AI in education centered around fears students would cheat or miss out on key steps in their learning. However, Microsoft’s Bartley Johns cites the successful incorporation of calculators into math teaching as evidence that learning and assessment methods adapt. “There are lots of positive opportunities here, and I haven’t spoken to anyone in education in Asia who thinks generative AI won’t be used in the universities and schools over the longer term,” he says.

These are early days in the development of generative AI. Regulation is nascent across Asia. However, with great power comes community obligation. It is vital the public sector develops and deploys generative AI responsibly, in ways that protect privacy and data security and foster citizen trust.

The first question for governments is to what extent do existing laws and regulations apply. The importance of AI operating within legal guardrails was addressed by Japan’s ruling Liberal Democratic Party in a recent AI national strategy white paper, while the Officer of the Privacy Commissioner in New Zealand recently published helpful guidance about how to comply with privacy law when using generative AI. Novel issues might include a requirement for the public sector to provide basic transparency about the AI models that it relies on. Partnering with trusted cloud service providers allows governments to leverage existing privacy and security architectures rather than start from scratch.

Inclusion is the second key challenge for a technology billed as benefiting everyone in the world—one that invites a natural, conversational interaction when the nearest government service center may be hundreds of miles away. This starts by extending mobile broadband connectivity, an area where governments in Asia have made progress but wide disparities between urban and rural areas remain. Use of personal devices also needs to be democratized and expanded.

Nonetheless, Microsoft’s Bartley Johns is optimistic about the transformative potential of generative AI as it is placed in millions, and ultimately billions, of hands. “There is a real leapfrog opportunity here, and that’s what we’re hearing from governments across Asia,” he says. “The underlying technologies are here today.”

This content was produced by Microsoft. It was not written by MIT Technology Review’s editorial staff.