When will AI be smart enough to pick the right agents for a task by itself?
A Facebook group member asks when AI systems will be able to take a one-sentence task and figure out on their own which agents, systems and apps to use — without twenty clarifying questions. The emerging answer: the blocker is trust, not intelligence. The practical pattern today is having the agent propose a plan (which tools, in which order) for one human approval, then run; full autonomy makes sense only where the cost of a wrong guess is near zero.