Next phase of enterprise AI will be domain-specific, not just bigger models: Articul8 AI CEO Subramaniyan

Bengaluru: Enterprise AI is moving towards domain-specific systems that combine deep domain knowledge with reasoning, governance and execution. The next AI advantage will not come from having the biggest model; it will come from having AI that understands the domain deeply enough to reason and solve problems that matter, Arun K Subramaniyan, Founder and CEO of Articul8 AI, an enterprise generative AI company focused on building domain-specific AI..Swarms of agents, not just assistants: The new way developers build software.On what is preventing enterprises from moving AI into production, he said the biggest misconception is that the primary problem is model capability. "It isn't. The hard problem is what happens when multiple models are needed, and decisions need to be taken in real-time," he said. "A system that is 90% accurate can make for an impressive pilot. In production, the other 10% matters enormously. In engineering, manufacturing, energy or financial environments, a wrong answer can have catastrophic operational, financial, safety or regulatory consequences. Another subtle but big issue is that AI can fail silently. A model can sound equally confident when it is right and when it is wrong," he said.According to him, one mistake enterprises make is assuming the largest frontier model should solve every problem. That is both expensive and often technically suboptimal.Different models are good at different things. A frontier model may be excellent at general reasoning. A smaller model may be faster and substantially cheaper for another task. And a domain-specific model can outperform much larger general-purpose models when deep industry knowledge matters, he said.Can AI become autonomous? "AI can already execute increasingly complex workflows autonomously. For an enterprise, autonomy is not the main objective. Outcomes are. The important question is how much autonomy an AI system should have for a particular decision. Some actions can be executed automatically. Others should require validation. And some decisions should always remain with a human expert. An enterprise AI system should know what it can do, what it cannot do, when it needs additional evidence and when it must escalate a decision to a person. That becomes especially important in engineering, manufacturing, energy and other environments where the consequences of an incorrect action can be significant," he explained.The company works with more than 50 customers globally. Going forward, its focus is on scaling the adoption of domain-specific AI and moving more enterprise AI from experimentation into production. "We see significant opportunities across the industries we serve and across markets including India, Japan, Korea and the US," he said.India is an important part of the company's global strategy and will continue to be a research, development and commercial hub for Articul8. The company has grown its presence in India by more than 2x in the past 12 months, and expects to grow another 2x to 3x as it expands this year."What makes India particularly important to us is the combination of deep technical talent and a rapidly growing market for enterprise AI. We see India as a market where we can build technology, develop domain-specific AI and serve local and global customers. We are already working with customers and partners such as Aditya Birla in India," he said.