A few years ago, while travelling with my family, we stopped at a small scenic town in Gujarat for tea. The shop owner, a middle-aged woman, was trying to make a digital payment to her supplier. She frowned at the screen, tapped a few times, then turned to her son and said, “Aa su lakhyu che? Mane samjhatun nathi” what’s written here? I don’t understand.
The incident stayed with me.
It said more about India’s digital divide than any data point or graph can ever represent. We are a country of near-universal mobile connectivity, but not of universal understanding. Our biggest digital divide today is not about who has a smartphone, but who the smartphone truly speaks to.
The data also speaks a language which is contrasting.
Today, over 800 million people are online and this number is only growing. Estimates suggest that by the end of FY2026, there will be more than a billion people who are online in India making us the second largest digital economy in the world. However, there is a catch. 98% of these online users consume content in their local language.
Yet, many digital interfaces, from government portals to apps, remain English-first. Imagine what it actually means. The app is there, the connectivity is stronger than ever, but is not understandable. When a mother in Surat hesitates to order groceries online or a farmer in Varanasi can’t navigate a loan application, the promise of “Digital India” remains unfulfilled.
Bridging the gap with AI
This is where the next leap in India’s digital journey must come from. Not faster networks. Not cheaper devices. But intelligence, intelligence that adapts to people, not the other way around.
Artificial intelligence, especially in natural language processing, is finally capable of closing the widest gap in our digital economy. The gap of comprehension. According to a 2024 IAMAI–Kantar report, over 600 million Indians identify local language as their only medium of digital comfort. Combine that with the fact that India has 22 official languages, 19,500 mother tongues and speech variants, and it becomes clear that the country’s biggest digital barrier is linguistic, not technological.
Generative AI now enables what policymakers once considered impossible, interfaces that speak the user’s language, understand their intent, and respond in dialects as diverse as Bundeli, Kutchi, Awadhi or Marwari. The technology is no longer aspirational; it is operational. India’s language models are being trained on the world’s largest non-English user base, giving us a unique advantage: we are not just participants in the global AI race; we are its most valuable training ground.
But beyond the excitement lies the real question. What does this solve?
AI can finally make digital public infrastructure inclusive by design.
Consider UPI. It processes more than 14 billion transactions a month. But imagine the next version where a user can simply say, in Bhojpuri or Garhwali, “Pachaas rupaye bhej do Shivani ko,” and the system completes the task without a single tap. No menus. No English. No fear of “Aa su lakhyu che?”
AI-powered voice interfaces alone have the potential to bring 150–200 million new users into the formal digital economy over the next three years. These are not just incremental users; they are the very citizens currently excluded by text-heavy, English-heavy design.
The stakes are even higher in public services. A farmer loses opportunities not because he lacks a smartphone, but because government advisories often reach him in a language he cannot interpret. The cost of this comprehension gap is massive, India loses an estimated $10–12 billion annually in productivity due to complex, inaccessible digital systems.
While the case for multilingual adoption is clear, many brands struggle to act on it. One challenge is fragmentation. Supporting multiple languages, dialects, and formats across apps, websites, customer support, and marketing feels operationally complex. Another is fear of quality and compliance, especially in regulated sectors like BFSI, where mistranslation can carry real risk.
Many companies still treat local languages as a “regional add-on” rather than a core growth lever. Others underestimate the cultural layer, assuming translation alone is enough, when true adoption requires transcreation, context, and tone.
AI can address these challenges, but only when implemented thoughtfully. Models must be trained on Indian linguistic realities, governed with strong human oversight, and embedded directly into enterprise workflows. When done right, multilingual AI reduces cost, improves trust, and unlocks scale rather than adding complexity.
For AI to truly democratise digital India, three things must happen:
- Interfaces must become invisible. Menus, tabs, and forms need to fade. The system should understand speech, intent, and context. That is how you build comfort for the next half-billion users.
- AI must be culturally grounded. Not just language models, but cultural models, understanding idioms, metaphors, accents, hesitation, silence. Technology that understands the way India speaks.
- Trust must be earned, not assumed.AI systems that power public services must follow the highest standards of privacy, transparency, and explainability. A tool is only empowering when the user feels safe using it.
The digital India we must build
The woman in Gujarat was not struggling with technology. She was struggling with translation. Her world was digital. But the interface was not built for her. If AI can help bridge that gap, between access and understanding, between intent and action, then India will unlock a digital dividend no country has ever seen. A billion people engaging meaningfully with digital systems is not just an economic outcome; it is a societal transformation. The next big story of India will not be about how many people came online, but how many people felt at home online. And that is a story only inclusive AI can write.
The article has been written by Nakul Kundra, Co-Founder of Devnagri AI
