© IE Online Media Services Pvt Ltd
Tags:
Journalism of Courage
“With ChatGPT coming into the mainstream, AI became democratised. Since it offered an easy explanation, people started asking questions,” said Prakash Kumar, chief executive officer at Wadhwani Centre for Government Digital Transformation (WGDT).
In India, where a vast majority of the population continue to remain digitally distant even with rapid smartphone adoption, Kumar’s organisation is driving a voice-first approach for citizen engagement through its Unified Citizen Engagement Platform (UCEP). This is an AI-powered initiative that is representing a shift in how government services can reach even the most underserved communities across the country.
Kumar, who is the former founding CEO of GST Network, in conversation with indianexpress.com, threw light on the challenges and opportunities in India’s digital governance. “Nearly 40 per cent of our population who are unable to access it (emerging technologies) may have a phone, but they don’t know how to access an application. They are literally excluded,” Kumar explained, highlighting the gap between device ownership and digital literacy.
According to him, this exclusion is aggravated when we consider the recent Supreme Court judgment from April 2024 that mandated that while the government can digitise its service, it must ensure accessibility for all regardless of their digital literacy levels. The ruling shows the legal push for inclusive digital solutions.
Kumar said that UCEP’s voice-first design addresses this challenge.
Unlike traditional chatbots that require typing skills and app navigation, the platform allows citizens to speak naturally in their local languages. “There are lots of chatbots available which you can use, but then you need to know how to open an app and how to type. How many can? I come from the Hindi belt. I can’t type in Hindi. That’s unfortunate. I can only type in English. But I can speak Hindi,” Kumar noted, emphasising the natural advantage of voice interaction for India’s diverse linguistic landscape.
The platform has been designed to strengthen the citizen-government engagement in local language across a variety of contexts. Some of the early use cases include assisting citizens seeking information, checking eligibility for schemes, lodging grievances, and tracking application status as well as government outreach use cases such as issuing advisories to farmers, nudges to parents for child immunisation, or reminders to taxpayers to file returns.
Collecting beneficiary feedback on government schemes and programmes and offering counselling and guidance to students is another use case. It could also be instrumental in running large-scale public campaigns and awareness drives.
As of now, UCEP supports 12 Indian languages – Hindi, English, Marathi, Malayalam, Telugu, Punjabi, Gujarati, Tamil, Kannada, Bangla, and Odia. Moreover, the AI platform also has the ability to adapt to local dialects.
When asked about the underlying technology that powers the platform, Kumar said that UCEP is powered by a secure, government-grade AI stack that has been purpose-built for citizen engagement at scale. “At its core, it combines generative AI, conversational intelligence, intelligent retrieval, and context-aware orchestration to deliver natural, human-like interactions,” he said.
Kumar explained that the platform is cloud-native, hyper-scalable, multi-tenant, and built entirely on open-source technologies, enabling population-scale conversations to run simultaneously at low cost. “Its open architecture and API-driven design allow ministries to easily ingest documents, integrate databases, and configure new services without rebuilding from scratch,” he noted.
When it comes to regional language, UCEP seamlessly connects with Bhashini for speech-to-text and translation, ensuring coverage across all 22 official Indian languages and local dialects. According to Kumar, since it is hosted on the IndiaAI infrastructure and designed to be compliant with the Digital Personal Data Protection (DPDP) Act, 2023, UCEP ensures enterprise-grade data privacy and security while remaining flexible and future-ready.
Kumar envisions significant monetary benefits across various government schemes. For instance, only 50 million hectares are covered under PM Fasal Bima Yojana out of 150-200 million hectares of farmland. Proactive outreach through UCEP could substantially increase insurance coverage, protecting more farmers from crop losses due to natural calamities.
Beyond financial metrics, the platform addresses critical social challenges. Kumar highlighted potential applications in healthcare: “A large part of our children are anaemic or stunted. Can we give them automated reminders to mothers for vaccinations? Check out – this is the ration which is given to parents of such children.” Such interventions, delivered in local languages, could significantly impact child health outcomes.
Addressing concerns about AI reliability and misinformation, Kumar emphasised UCEP’s constrained approach to information retrieval. “The tool is trained first on government data; it cannot use its Internet knowledge. Large language models are trained on all kinds of information. So, we use RAG and train on government data, and it will answer only from that data,” he explained. The system’s temperature is set to zero, and it’s programmed to acknowledge limitations rather than generate potentially incorrect responses.
The Wadhwani Centre’s approach extends beyond technology deployment to capacity building. Over the past four years, they have trained 9,600 government officers through workshops and converted training programmes into online courses available on the government’s IGOT platform, reaching over one million officers and staff. This comprehensive approach ensures that technology adoption is supported by adequate human capacity.
The organisation also provides Chief AI Officer-led teams to six major ministries – agriculture, education, skilling, labour and employment, women and child development, and electronics – helping develop 15-20 project roadmaps per ministry for AI implementation.
Kumar advocates for India to develop indigenous AI capabilities rather than relying solely on Western models. “Since they are trained on western data or largely English-speaking world data… they may not have full information about our systems, and they say they may be biased against us,” he argued, highlighting both technical and strategic considerations for long-term AI sovereignty.
His message to young technologists reflected this broader vision: “Look at the problem which is afflicting a large number of people in our country and try to solve it, and there are enough people in the government who can help you in identification.”
As India continues to build upon its digital public infrastructure foundation, voice-first AI platforms like UCEP represent a promising pathway toward truly inclusive digital governance, ensuring that technological advancement reaches every citizen, regardless of their digital literacy or linguistic background.
UCEP is yet to be officially launched. The team at WGDT are closely working with the Government of India on it.
