India’s artificial intelligence ambitions are forcing a rethink of the National Informatics Centre’s (NIC) role, from a provider of digital infrastructure to a potential backbone for AI-led governance. But getting there will require more than incremental upgrades.
According to former NIC Director General Neeta Verma, the shift demands structural changes across infrastructure, capacity, and governance, not just technology adoption.
“Only about 10% of AI pilots in the public sector make it to scale. The reasons are not technological, but systemic,” she says in an interaction with AIM.
That gap between experimentation and execution is where NIC’s next phase will be defined.
From Digital Backbone to AI Backbone
For decades, NIC has powered India’s digital public infrastructure, with systems spanning data centres, networks, and cloud platforms across nearly every district.
This scale, combined with long-standing institutional trust, gives it a natural advantage in leading AI adoption across government. But Verma argues that existing strengths alone are not enough.
Moving to an AI-first role will require NIC to evolve from building infrastructure to enabling shared intelligence systems across ministries and states.
Despite rising interest in AI, most government pilots fail to move beyond the experimental stage.
Verma, the first woman to serve as NIC Director General, held the position from 2016 to 2022. During her tenure, she led the development of key digital platforms, including MyGov, and oversaw the expansion of national data centres and the launch of India’s national cloud.
She also spearheaded the creation of NIC CERT and later served as Chief Advisor (IT) to the Election Commission of India, contributing to the management of the 2024 Lok Sabha elections.
Based on her experience, Verma points to three recurring issues.
First, data readiness is often overestimated. Government datasets are fragmented, inconsistent, or not usable at scale. Second, regulatory and accountability frameworks remain unclear, especially when AI systems influence decision-making. Third, ownership is fragmented. Policy, operations, and technology teams often work in silos, making integration difficult.
Just as critically, many AI pilots are not embedded into core workflows.
“Most AI pilots don’t fail because the technology falls short. They fail because the system is not ready to absorb them,” she says.
Building Shared AI Capabilities
To address the infrastructure concerns, NIC will need to prioritise shared AI capabilities as digital public infrastructure.
Instead of building isolated solutions, Verma emphasises reusable models, platforms, and tools that can be deployed across departments.
This approach would reduce duplication, improve efficiency, and accelerate adoption.
Alongside infrastructure, capacity building will be crucial. AI deployment at scale depends on whether officials across levels, not just central teams, can understand and use these systems effectively.
She also calls for expanding initiatives like NIC’s Centre of Excellence for AI in Governance, particularly at the state level, where implementation gaps are most visible.
MeghRaj Needs a Redesign, Not an Upgrade
At the core of NIC’s infrastructure challenge is MeghRaj, India’s national cloud.
Built in 2014 to host applications and digital services, MeghRaj was not designed for AI workloads, which are significantly more compute-intensive.
“AI workloads are GPU and accelerator-intensive. This cannot be addressed through incremental upgrades,” Verma says.
Instead, she proposes a parallel AI compute layer alongside existing cloud systems.
This would include a centralised infrastructure for high-performance tasks such as model training, distributed systems at state or regional levels for low-latency inference and shared, pooled compute accessible across departments.
Such a model would move NIC away from fragmented deployments towards a more efficient, scalable architecture.
Initiatives under the IndiaAI Mission are already expanding access to high-performance compute, but integrating these into NIC’s broader ecosystem will be key.
Verma also highlights the importance of sovereign AI infrastructure, particularly for handling sensitive government and citizen data.
Where AI Will Deliver First
In terms of impact, AI adoption is likely to be uneven.
Citizen-facing services such as conversational interfaces, multilingual platforms, and voice-based systems are expected to see faster gains due to lower risk and clearer usability benefits.
Back-end applications, including welfare targeting or decision support, will take longer. These systems require stronger validation frameworks, clearer accountability, and greater trust.
Verma emphasises a crucial design tenet for incorporating AI into governance AI systems should leverage existing digital public infrastructure instead of replacing it.
“The principle is simple. AI should sit on top of APIs, not replace them,” she says.
Core systems should remain rule-based and auditable, while AI handles interaction layers and orchestration.
This layered approach reduces risk while allowing innovation at the edges.
From Pilots to Embedded Systems
The larger shift, according to Verma, is from isolated AI experiments to embedded, shared capabilities within government systems.
This transition will define whether AI remains a set of disconnected pilots or becomes foundational to governance.
For NIC, that means going beyond infrastructure provisioning to enabling adoption across institutions, workflows, and geographies.
Your reaction