Resources

At a time when artificial intelligence is finding its way into nearly every industry, governments are now evaluating its implications for public administration.

Governments globally are experimenting with AI tools in a variety of ways, including service delivery, data analysis, and communicating with citizens at scale. Data from a 2025 OECD study found that government adoption of AI is highest in areas such as ‘public services, civic participation, and justice – areas with high transaction volume and direct citizen interaction. The Indian government, too, has identified AI adoption as a priority. In 2024, the IndiaAI Mission outlined seven pillars of work, including developing AI applications for India-specific challenges in healthcare, agriculture, and other sectors, and creating Large Multimodal Models using Indian data and languages.

As these conversations evolve, we wanted to explore a more specific question: can AI help make government data more useful for decision-makers on the front lines of public administration?

For us, the conversation is not about adopting AI for its own sake. It’s about understanding whether these tools can help strengthen governance capabilities and support better decision-making.

The Problem We're Working On

Through our work with the Government of Meghalaya, we are supporting efforts to institutionalise Evidence-Based Decision-Making (EBDM) to address key human development challenges across the state. This involves making evidence easily accessible to officials so decisions can be informed by a broader, context-specific knowledge base. 

Local administrators make consequential decisions every day about where to direct resources, which communities need urgent attention, and how to connect people with the services meant to reach them. However, the information that could support those decisions often exists out of easy reach, hidden behind numerous digital platforms.

Many administrators still rely on traditional approaches: manually pulling records, responding to problems as they become visible, drawing on experience and intuition. These approaches carry real value. Local officials often develop deep contextual knowledge of their communities. But without timely access to synthesised evidence, decisions become reactive rather than strategic.

This is especially visible in areas like maternal health. An officer may know that outcomes are poor in a particular area. Understanding why — whether the causes are linked to nutrition, immunisation coverage, service delivery gaps, school attendance, or access to welfare schemes — requires pulling information from multiple disconnected systems. In most cases, the data exists. The problem is making it usable at the point where decisions are made.

With this challenge in mind, we set out to develop a tool that would synthesize existing government data into visible, accessible insights.

Building the Tool

Although our team has extensive experience working with large datasets and government information systems, we needed different skills to build an interface that makes information accessible. That’s what led us to apply to Glific's AI Chatbot Accelerator 2026.

In April 2026, Sidharth Santhosh from our team joined a national cohort of nonprofit practitioners exploring how AI-enabled WhatsApp chatbots could support communities and programmes. Experts from Glific at Tech4Dev guided us as we considered the technical aspects of building a chatbot and the practical considerations that determine whether a tool proves useful in real-world settings. Tech4Dev’s deployment experience also helped us plan for practical constraints like data literacy.

The programme also reinforced the importance of an iterative approach. In the coming months, we will continue testing our assumptions, refining chatbot flows, and learning from real-world use cases to understand which interactions produce usable outputs for officials.

How the chatbot supports officials

The chatbot we are developing will help block- and district-level officials access and interact with administrative data through a simple interface. It synthesises existing government data, providing officials with a clearer picture of the challenges and opportunities within their jurisdictions.

Instead of manually navigating multiple portals and dashboards, an official can use the chatbot to instantly receive tailored, synthesised insights that connect information across various datasets.

By connecting siloed information, the chatbot helps reveal relationships between indicators and sectors and schemes. This reduces the effort required to move from collecting information, to understanding, then to action. In practice, this shifts the role of data from something that is reviewed periodically to something that actively supports day-to-day decision-making.

Government systems already contain a wealth of information. The challenge is making that information usable at the point where decisions are made. Tools like this could help shift the role of administrative data from a reporting requirement to a resource to support decision-making.

We are still in the early stages of this journey, with different pilots taking off. Over the coming months, we hope to learn as much from what doesn't work as from what does, and to contribute those lessons to a broader conversation about the uses of technology in governance and administration.