Based on an article from SiliconANGLE, the evolution of artificial intelligence is moving from generative AI to agentic AI, particularly in enterprise applications. Generative AI, which operates on a request-response model, has not fully met expectations due to its limitations in delivering consistent and actionable business value. Agentic AI, however, aims to build on generative AI by creating autonomous agents that perform tasks with defined goals and supervision, thus enhancing efficiency and ROI.

Agentic AI in consumer settings poses challenges due to the open-ended nature of tasks, whereas in enterprises, tasks are well-defined and mapped, making them more suitable for agentic AI. The technology involves multiple agents collaborating to achieve complex objectives, such as in Amazon’s supply chain operations where various agents handle forecasting, distribution, and order fulfillment. This orchestration of agents can significantly optimize operations by coordinating interdependent tasks.

The shift to agentic AI requires an evolution of the current AI stack, moving from large language models (LLMs) to large action models (LAMs) that not only retrieve data but also orchestrate workflows. This transition entails integrating legacy systems and harmonizing data to create a digital representation of the business, facilitating near-real-time action and decision-making. Companies like OpenAI, UiPath, Celonis, and Palantir are leading this space by developing tools and platforms that enable the deployment of agentic AI.

Despite its potential, agentic AI must overcome challenges such as ensuring data accuracy, maintaining accountability, and addressing privacy concerns. The development of a robust semantic layer to represent business entities and activities is crucial for the effective functioning of agentic AI.