Access to powerful AI is no longer the hard part. As frontier and open-source models get cheaper and more interchangeable, the real question for businesses is what they own that others cannot copy.

That question anchored the TechSparks 2026 session ‘Beyond the model: The moat for durable AI.’ Divakaran Ullampuzha Mana, Director of Solutions Architecture, India, MongoDB; Vijeth Pandit, CPO, Pine Labs Payments; and Karan Kumar, VP, Product, Spinny, spoke with Shivani Muthanna, Senior Director, Strategic Content & Video, YourStory, about what companies must own to build AI that lasts.

Mana began by noting that open-source models are catching up with frontier models in capability and performance, and access costs keep falling. What separates one business from another, he said, is how seamlessly it can access its own enterprise data, which sits across systems in structured and unstructured formats, and turn it into insight or action.

Pandit pointed to distribution and accumulated experience. Pine Labs works with almost three million merchants and hundreds of banks in India and globally. He noted that “models in isolation will only help in inferring”, and that the interactions, data, and history a company builds are what make a product relevant.

Kumar offered an analogy. “I look at models as electricity,” he said, adding that the winner is decided by what a company plugs in and what the layers on top are built to solve. At Spinny, those layers rest on verified definitions of internal metrics, the reasoning behind past decisions, and how the team runs and evaluates experiments.

Source link