India's AI ambitions face a critical challenge as GPT 5 achieved only 45% accuracy on a human-curated benchmark covering 11 Indic languages, including Prime Minister Modi's mother tongue Gujarati. According to ETHRWorld, this performance gap highlights the massive linguistic divide that threatens to limit AI's democratization across India's diverse population. The country has nearly two dozen official languages and more than a hundred dialects, creating barriers that could prevent AI from becoming a medium of inclusion and empowerment for all Indians. Voice recognition emerges as particularly critical, with South Asia using voice for everything - from business calls and WhatsApp voice memos to speech-based payments and coding tools.
The AI for Inclusive Societal Development roadmap examined a range of informal worker personas, including electricians, plumbers, healthcare aides, artisans, kirana store workers and tour guides. Across these groups, recurring challenges included limited digital readiness, fragmented skilling pathways and barriers to technology adoption. The findings suggest that the success of emerging technologies may depend as much on accessibility and usability as on technological capability. According to UNESCO's India AI Readiness Assessment, BHASHINI currently supports 22 Indian languages and has been integrated across a growing number of public-facing digital services. The platform was established to help bridge digital, literacy and language divides through multilingual and voice-enabled experiences, with potential to improve accessibility for more than 200 million citizens. The report identifies affordability, localisation and inclusivity as key assumptions for large-scale AI adoption over the next decade.
According to Vivek Prakash, CEO of AI and coding education platform Codingal, healthcare, agriculture and education represent the biggest opportunities in AI. As reported by Business Standard, each sector combines scale, local data depth, and genuine urgency that global models have not been trained for. The healthcare sector faces challenges including doctor shortages, diagnostic burden, hospital data, medical transcription, radiology, and insurance claims. Agriculture presents opportunities in weather forecasting, pest alerts, crop advisory, mandi prices, credit and insurance services. Education requires multilingual learning solutions, addressing teacher shortages and personalized tutoring needs. These sectors are particularly critical as AI moves into schools, hospitals, courts and public service, where language failures have consequences. Indian Railways' RailMadad platform exemplifies this approach, using BHASHINI's multilingual capabilities to enable citizens to raise grievances and seek information in their preferred languages.
To address the data scarcity challenge, Project Vaani is collecting speech data from approximately one million people across 773 districts, covering 86 languages and spanning all 22 states. The initiative aims to create a large-scale speech dataset exceeding 150,000 hours, helping address one of the longstanding challenges in developing language technologies for diverse linguistic contexts. As noted by Sandeep Chinchali, co-founder of Poseidon, Bengali makes up less than 0.1% of web text, creating significant data scarcity for Indic language models. Speech data requires accurate transcription, longer clips, varied acoustic environments, demographic and regional variants, as well as careful human review before improvement. A Stanford team warned that quality has become a key challenge when trying to scale such endeavors, raising ethical questions in a sector with a long history of poor pay and exploitation.
NITI Aayog's AI for Viksit Bharat roadmap estimates that AI-associated opportunities could contribute between $1 trillion and $1.4 trillion towards India's economic growth aspirations by 2035. The report argues that AI-driven productivity improvements, innovation and technology services could play an important role in helping India move closer to its long-term development goals. However, the same policy thinking highlights that adoption cannot be assumed simply because technology exists. The challenge is especially visible among segments of the workforce that sit outside formal technology ecosystems. For enterprises pursuing AI-led transformation, adoption is increasingly emerging as a business challenge rather than a purely technical one. The report emphasises that trust, disclosures, grievance redressal pathways and user confidence are important enablers of long-term AI usage, with trust becoming part of the infrastructure required for adoption. India's digital transformation was built on expanding access, but its AI journey may be defined by how successfully it converts that access into adoption as AI becomes embedded across public services, workplaces and everyday interactions.
