Agentization of Digital Assets for the Agentic Web: Concepts, Techniques, and Benchmark

A research team from institutions including Shanghai Jiao Tong University has published a foundational paper outlining a new paradigm called the 'Agentic Web.' This concept redefines the internet by enabling autonomous, goal-driven interactions between AI agents. The core innovation is 'agentization'—the process of transforming static digital assets (like interactive web elements, forms, or APIs) into functional, communicative agents. The paper formalizes this A2A (Asset-to-Agent) Agentization process, breaking it down into critical technical stages and identifying the key hurdles to automation.

To solve the automation problem, the team developed an 'Agentization Agent,' an AI system designed to perform the conversion of digital assets into agents. To rigorously evaluate such systems, they also introduced A2A-Agentization Bench, the first benchmark explicitly designed to measure the quality of agentization. It assesses two crucial metrics: fidelity (how accurately the agent replicates the asset's function) and interoperability (how well it collaborates with other agents). Their experiments show their approach successfully activates asset capabilities and enables functional multi-agent collaboration, paving the way for a more intelligent and automated web ecosystem.