The paradox begins when an AI system is tuned for maximum recall, flooding users with every marginally relevant signal. Each ping forces a mental switch, and the brain’s “decision‑making budget”—the limited attention you can allocate before fatigue sets in—gets depleted far faster than if only the top‑ranked alerts arrived.

As attention wanes, users start deferring choices, opening the same alert twice, or asking a colleague for confirmation, which adds hidden latency. In a mid‑size fintech firm, a newly deployed risk‑monitoring model began pushing three to four alerts per hour to each analyst; within a week, the average time to close a flagged transaction stretched from minutes to nearly an hour, because analysts were juggling the noise instead of acting on the few truly urgent cases.

The root cause is not the AI’s intelligence but the system’s bias toward quantity over quality, which crowds out the mental space needed for rapid judgement. The remedy is to treat alerts as a scarce resource, curating them through a simple relevance filter that respects the human attention curve.