An EY report claimed that agentic AI could handle 80% of routine treasury tasks. ^[1] The author states that every AI agent built today eventually encounters a wall caused by memory limitations rather than a lack of reasoning power or coding ability. ^[2] The EY report noted that companies must first establish reliable data systems. ^[3] The same EY report said this would let treasury professionals focus on strategic work. ^[4] According to the report, 60-70% of treasury bandwidth at mid-levels is currently spent on low-value activities including collecting data from different systems, manual validation, and preparing reports. ^[5] The Inside Ecommerce digital leaders series initiated a fresh perspectives session featuring industry experts from Boden, Commerce, and The Drapers. ^[6] The article advises developers to explicitly write down where the AI agent is permitted to access information and where it must stop. ^[7] The author recommends documenting how to verify that the AI agent has successfully completed its assigned tasks. ^[8] The author states that one of the first steps in creating an AI agent is to decide on a specific name for the model. ^[9]