Enterprise AI Breakthroughs Are Coming Fast

Real Large Action Models, or LAMs, are still in their infancy. Early examples like Google’s RT-2 or consumer devices such as Rabbit R1 exist, but consistent production-grade outcomes remain elusive. LAMs require enormous amounts of diverse training data covering environmental states, actions, and feedback loops. Gathering this data at scale is both expensive and technically challenging.

Physical action carries real-world risk. Unlike text, errors in the physical world can cause harm, slowing enterprise adoption. Despite these hurdles, LAMs represent the next logical step beyond LLMs. The rapid adoption of LLMs ensures that research into LAMs will accelerate in the coming years. Today, however, agentic systems dominate. They may lack physical capabilities, but their versatility and adaptability already outperform traditional rules-based automation when paired with proper orchestration and safeguards. Enterprises can already deploy agentic AI to manage complex workflows, schedule tasks, and integrate across multiple tools, delivering measurable operational ...