We acknowledge that both humans and "medical grade AI" are flawed, but they're flawed in very different ways and until we can understand how and why an AI model fails, it should be supplemental.