Agentic AI: The “New” SOA and Why Your Security Strategy Needs a Reboot

AI agents aren’t that different from past innovations. Credit: Finland Leskinen

The proliferation of AI buzzwords can feel overwhelming, with “Agentic AI” being the latest to promise revolutionary changes. From my experience, deeply involved in the security side of AI, machine learning, and data analytics for over a decade, Agentic AI, while incredibly powerful, is fundamentally a re-imagination of Services-Oriented Architecture (SOA), a concept that emerged more than 20 years ago.

When SOA first appeared, it aimed to break applications into independent modules with simple APIs, promising a sense of operational nirvana. However, that vision largely fell short in practice because vendors were reluctant to expose individual functions, and at scale, SOA often proved more expensive than anticipated.

The Rebirth of SOA: Agentic AI’s Powerful Evolution

Agentic AI advances the SOA concept by integrating repeatable AI-driven functions with emergent agent frameworks, thereby significantly enhancing portability and simplifying orchestration. Previously, the hypothetical operational costs of unused “agents” presented a barrier, particularly for smaller, innovative companies and startups. However, platforms like Amazon’s AWS Bedrock Agentcore and Google Cloud Run are now eliminating these obstacles by offering a utility…