Explainable AI (XAI) is crucial for fostering human trust in deep neural network (DNN) predictions. Traditional XAI methods often rely on individual pixel-based explanations, which highlight significant pixels but may not align with human intuition. In contrast, concept-based explanations are more interpretable but typically depend on human-annotated concepts, limiting scalability. In this paper, we present Any Segment Explanations (ASE), a black-box method that generates explanations by evaluating model prediction through concept insertion and deletion. ASE extracts human-understandable concepts from any image without manual annotations. Experiments demonstrate that ASE is 38 times faster than the state-of-the-art EAC method, achieving superior Area Under the Curve (AUC) scores and providing more intuitive explanations. The source code is available on the link.