The interview was conducted by Ula Rutkowska (Senior Researcher, Apolitical) and edited by Christina Obolenskaya (MSc in International History, LSE and Communications Intern, Apolitical). Dr Preet Deep Singh was named on Apolitical’s Government AI 100 2025.

Artificial intelligence is more than a tool — it’s a vital skill for today’s public servants. In this interview, Dr Preet Deep Singh highlights the importance of AI knowledge for civil servants, from understanding its capabilities to asking the right questions. He shares insights on hands-on learning, tailored training, and how AI can streamline workflows and support smarter policy creation.

This conversation explores why AI training matters and how it equips public servants to lead in the future of governance. Read the full interview for practical tips and strategies.

How do you use AI to help improve the public service sector in India?

There are 1.4 billion people in India, and it’s impossible to deliver government services to everyone without scaling up. One option is hiring more staff, but another is deploying agents across the country. However, a large portion of India isn’t typing first like many of us. Many use voice notes on WhatsApp, which presents a challenge due to India’s 22 official languages and 800 dialects. Even if we hired 5 million agents, they’d need to understand all these dialects. The solution lies in AI.

AI became crucial, especially after India’s 2024 Presidency of Global Partnership on AI. Effective AI deployment became a priority. At the same time, fake news and deepfakes are major concerns with the use of AI. In a large population like ours, we are still experimenting on how we verify truth because truth might have a different definition and in some cases, it might not be the same as reality.

In government, many processes require strict deadlines. For example, if we have a meeting, minutes must be issued within 24 hours. But that requires someone to listen, create drafts, and double-check the information. Can we use AI to automate dull tasks or those that are time-consuming or dangerous?

AI can also improve government efficiency. For instance, when evaluating a proposal, like building a new highway, instead of coordinating with multiple colleagues, we could ask AI to simulate expert discussions. This saves time and provides insights we might not have considered. Scheduling is another challenge—AI can help solve that by organizing time more effectively. Since India is large and works in silos, integrating AI-based thinking into government planning—across economics, administration, health, and more—is key.

In my role, I bridge the gap between technology and government. I translate technical language into something legal teams and decision-makers can understand. I help senior officials and stakeholders grasp how AI can be used effectively, whether for regulation, implementation, or innovation. I connect the dots between government and technology. I’m not the one driving the initiative, but I help others navigate it. My job is to be a translator between two worlds.

Can you discuss some possible application areas for AI?

One project is with the Indian Railway Catering and Tourism Corporation (IRCTC), the Indian railway ticketing service. We've created a bot that allows users to book tickets using voice commands in multiple languages, including English, Hindi, and Gujarati. Another project involves a WhatsApp bot for district administration, where residents can report issues, and the bot tracks and updates the status. In the bot, I can say that I am a resident of a so-and-so place, and the road outside my house has not been cleaned, and I want to get it cleaned. This automatically gets filled into a data frame dashboard of the collector and goes to the relevant department municipality, and it is tracked and then an update is sent back right so you no longer have to actually go to the collector's office.

These pilots, when scaled, can unlock a lot of efficiency. For example, the IRCTC bot can handle ticket bookings without users needing to click through multiple screens, and the district administration bot can streamline the process of reporting and resolving local issues.

How important is AI training for the 21st-century public servant?

It's probably an 11 on a scale of 1 to 10. AI can save a lot of time, which is invaluable for civil servants who are always short on time. For example, AI can help with grammar checks in letters, saving significant time. If I can save half an hour of a public servant’s time that would have gone into reviewing grammar mistakes, I have unlocked a lot for the nation.

In addition, AI unlocks learning. For instance, AI can help civil servants understand complex laws or interact better with stakeholders. If everyone is using AI, it's essential to keep up. Not using AI is like bringing a knife to a gunfight; it's crucial to stay competitive.

What are the greatest challenges governments face in AI adoption?

One challenge is understanding the difference between a regular predictive model, like Netflix's recommendation system, and generative AI, which creates something new. Another challenge is understanding the limitations of generative AI. For example, if you ask GPT to write a government letter without context, example, or iteration, it might not meet your expectations.

We need to navigate between the two extremes of what is genuinely a limitation of the technology and what is a limitation at our end. Accountability remains with us, so understanding these nuances is essential.

What does effective AI training for governments look like?

I host online courses, on the government portal that is part of the Karamyogi Capacity Building Commission, and in-person trainings. People are more open to asking questions in person, especially without their seniors present. For basics and introduction, online courses with detailed handouts work well. For more complex questions around AI use, in-person training with peers is better.

People need to experiment with AI and ask specific questions to learn effectively. For example, many people want to know where ChatGPT is running, whether it's on their phone, and if it's accessing their data. These are silly questions, but they are important for understanding. Once people are introduced to the idea, they can ask more specific questions in person.

What are your thoughts on building a national AI strategy?

In bureaucracies, work often follows a pyramid structure. The person at the top comes up with an idea and delegates it to the next level. The second person passes it along to a third, who creates something. The second person then reviews and revises it, sending it back for further changes. This back-and-forth continues, with each level revising and improving upon the work.

I want AI to handle the foundational work, providing a solid base for me. With my team, we can then build on that foundation and create something meaningful. While policy need not be outsourced, AI can certainly support the process by providing a strong starting point.