Agentic AI vs AI Agents

This article avoids the clinical definitions of Agentic AI and AI Agents. The terms have been over used in marketing and has created confusion on whether Agentic AI is the same as the AI Agents (of 2024).

Agentic AI is a distinct new class of technology. It is powered by thinking Large Language Models (LLM) at the core of the capability (not as an add on). Even if LLM-driven technology may not have sufficiently impressed you, the trajectory of General Intelligence AI is growing at a rapid rate, and backed by hundreds of billions of dollars of investment. It’s not “if”? but “when”.

Agentic AI is positioned to be autonomous-first guided by Reasoning and Thinking AI-engines, defined by a scope of activity which is essentially a job description. This top-down approach assumes that AI-engines have the intelligence to determine a course-of-action. This is what software like OpenClaw, Hermes Agent (and others) attempt to deliver.

AI Agents on the other hand are ground-up technology with lower intelligence. They are programmed with rules, and configured. They tend to be deterministic and limited by guardrails (by design). Power Platform and Robotic Processing Automation (RPA) software have thrived in this space.

For now, Agentic AI is driven by rapid innovation. It struggles with guardrails and safe enterprise deployment patterns. However, there is no doubt that it presents the future. Designs with technologies like AWS Bedrock and Azure AI Studio present possible secured deployment environments for Agentic AI. Other isolated, secure deployment patterns are also possible.

Rapid innovation is transforming Agentic AI for the enterprise. The question is should organisations start allocating budgets for this change? and when? I propose that strategies, roadmaps and 2-3 experimental projects are needed today. Enterprises can then learn, and can better navigate this change.

See our other articles on Vibe Coding.