Patent landscape study on agentic AI systems and related technologies

Abstract

Over the years, AI has progressed from task execution to autonomous systems creating a foundation for next–generation digital automation and intelligence. Early AI models such as Generative AI were designed to perform task based functions such as classification, search optimization, or rule–based decision making. The next stage AI Agents introduced systems that can automatically execute tasks and interact with application programming interfaces (API) on behalf of a user. Agentic AI represents the latest and most transformative systems and has also introduced new orchestration and architectural paradigms. Agentic AI systems are built using multiple technology layers that enable the step to perceive, plan and conduct multi–step reasoning, decision–making, and coordinated action at scale. The architecture comprises of five simple layers such as user interaction layer, the Agentic AI orchestration layer, AI Agents layer, application programming interface (API) and data integration layer and the final output layer that delivers the results from all the layers. Annual global AI patent filing is estimated to exceed USD 50 billion in value creation [8]. These technological advances and product development in AI are occurring in parallel with an increase in global AI patenting. Generative AI, AI Agents and Agentic AI Technologies appear to overlap as Generative AI provides reasoning/content generation, while AI Agents add tools/actions and Agentic AI adds autonomy, planning and orchestration. Most of the prior publications focus on patent landscape on AI or Generative AI. This paper presents a robust search strategy to capture and segregate Agentic AI–related patents from the broader foliage of AI–related inventions and presents highlights of patent filing trends on Agentic AI and related technologies.