Smaller, smarter, safer: How to build agentic AI on the right foundation

Thakrar, speaking at a CIO event, argued that organizations should favor architecture and human-in-the-loop systems over simply deploying ever-larger AI models; he recommended using smaller models where appropriate, letting surrounding systems decide more than the model, and verifying outputs at the point errors are locked in, with a focus on rich context layers to reduce reliance on frontier models.