India’s AI Advantage
Just over two years ago, I had the opportunity to meet Honourable Prime Minister Narendra Modi and deliver a keynote at a conference of Chief Secretaries, attended by nearly 200 of India’s most senior IAS officers. I outlined a vision for how India could move beyond being an AI-services nation to becoming a true AI product nation, one that leads in economically and socially critical sectors.
My central argument was clear - India cannot follow the West in investing heavily in expensive, consumer-focused Large Language Models (LLMs). Instead, it must focus on a different class of enterprise AI, homegrown, sovereign, task and industry-specific, and compact models. These models can drive products from the Indian companies to solve real-world problems.
These models go beyond language. They are built to work with multimodal data, sensor, seismic, weather, and industrial inputs, grounded in India’s own data, knowledge systems, and workflows. What we call Small Language Models (SLMs) is a subset of this broader enterprise AI approach. The opportunity is significant. The global SLM market is expected to grow from under $1 billion in 2025 to over $5 billion by 2032. Gartner predicts that by 2027, organizations will use small, task-specific models three times more often than general-purpose LLMs.
India’s strength lies in its diversity, healthcare, agriculture, finance, supply chains, and governance, each offering opportunities for highly targeted, use-case-driven AI models that deliver far greater precision and reliability than one-size-fits-all systems. In the long term, LLMs risk becoming monolithic and generalist. The shift is already underway toward composite AI systems, multi-component architectures that integrate domain knowledge and operate reliably in high-stakes environments.
AI Models for India
Global trends indicate that domain-specific AI models, knowledge-driven systems, and neuro-symbolic architectures will form the foundation of next-generation, robust and trustworthy AI, particularly for enterprise and India-centric applications. By 2027, more than 50% of enterprise-grade generative AI models are expected to be domain-specific, built on curated datasets aligned with industry workflows and regulatory requirements. These models prioritise depth over generalisation, making them significantly more effective in solving real-world problems.
The next layer involves the explicit embedding of structured knowledge. This mitigates hallucinations, enhances explainability, and ensures that AI systems remain aligned with domain rules, safety constraints, and human values, especially in high-stakes environments such as healthcare and crisis management.
Building on this, neurosymbolic AI frameworks integrate data-driven neural networks with symbolic reasoning. Neural components handle statistical learning, while symbolic systems enable transparent, rule-based reasoning and the encoding of domain expertise.
This hybrid approach provides stronger guarantees around accuracy, consistency, and compliance, requirements that are essential for enterprise adoption. Most current neural network-based LLMs process patterns without understanding meaning and lack built-in mechanisms for reasoning or value alignment. This makes them less reliable where reliability matters most.
India has an opportunity to go further, towards a Value-Inspired AI (VAI) architecture that embeds human values, social norms, and cultural context directly into decision-making systems.
From Concept to Application
The real test of AI lies in its application. In semiconductor manufacturing facilities emerging in Gujarat, and pharmaceutical production units in Hyderabad, quality control is mission-critical.
In such environments, neurosymbolic AI can integrate domain frameworks such as Failure Mode and Effects Analysis (FMEA) with data-driven models to enable real-time defect prediction, diagnosis, and prevention. Root cause analysis, powered by multimodal data, from sensors to imaging, allows not just detection, but systemic correction.
Similarly, in pharmaceuticals, domain-specific enterprise AI models can support production monitoring, regulatory compliance, and clinical processes. By combining sensor data with encoded domain rules, these systems enable proactive anomaly detection and improve reliability in critical operations.
These are not abstract possibilities. They represent a shift from reactive systems to intelligent, context-aware decision-making frameworks.
A Strategic Opportunity for India
To realise this vision, India must invest in building its data and knowledge infrastructure, capturing domain expertise, structuring it, and embedding it into AI systems. This is not just a technological shift, but an economic one. It creates opportunities for building new capabilities, generating knowledge assets, and developing globally relevant AI solutions.
India can do big things with small models. Instead of following Big AI practices in the West and China, India can focus on small AI.
By focusing on precision, efficiency, and contextual intelligence, India has the opportunity to move from being a consumer of AI to becoming a creator of globally relevant, high-impact solutions and deliver them as products with these new generations of AI models at their core.
In a world focused on scale, India’s advantage may well lie in building what truly matters.
