Diagnosing a sick child is exactly the kind of decision AI must never make alone.

Every presentation of pediatric abdominal pain is clinically unique, involving different symptom combinations, timelines, and physical signs. There is no repeatable structure an agent can reliably apply to reach a safe diagnosis.

Success criteria are life-or-death and cannot be ambiguous — a wrong triage decision (send home vs. go to ER) has irreversible consequences. The task demands crisp, correct answers in a domain where ambiguity is dangerous.

An AI agent has no access to the child's physical presentation, vital signs, palpation findings, medical history, or real-time clinical context. All the data that actually drives diagnosis is inaccessible to any agent.

A missed or incorrect diagnosis could result in a child's death or serious permanent harm. This is among the highest possible error costs — completely irreversible and catastrophic.

This requires licensed clinical judgment, physical examination skills, and the ethical accountability of a medical professional. No AI agent can substitute for a physician or emergency clinician in this context.