While it's true this is mostly justification for further investigation, correctly categorizing 12/12 people into 2 categories actually has a p-val of .000244141 = (1/(2^12)), which would easily allow you to reject the null hypothesis of random categorization. The stronger the effect, the fewer samples you need.

We consider n=12 generally underpowered only because many real-world effects are way weaker than the ability this woman demonstrated.