Could the agency have this task? Could this work through complex and dynamic branch points, make autonomous decisions and act on them? This requires a logic of threading together on thousands of decisions. I have spent more than 25 years working with automatic learning and automation technology, and agency AI is clearly a difficult problem to solve.
A potential change of game for and against fraud
The more complicated a system, the more vulnerable it is to attack. The agentic AI worries me about this front because fraudsters can use technology to exploit the weaknesses of security.
The verification of documents, for example, may seem simple, but it involves several steps, including image capture and data collection, behind the scenes. This creates a large area for fraudsters to probe with an agentic AI, and they can do it much faster with this technology.