When humans are asked to give an explanation of a decsion, they do not try to give a neuron-by-neuron account of the firing pattern within their brain, but instead provide a high level description of their reasoning that is sufficient for the purpose. For example, when asked why you bought a new shirt, you might say, "I was running out of summer shirts and I liked the pattern".
In explainable AI there can be a temptation to give an exact account of how it works, but this is akin to the neuron-by-neuron account of the human decisoon. Instead, XAI should focus on giving sufficient reason for its outputs. What is sufficient depends on context and this might need to be negotiated. In the human case, if asked you could probably elborate on the reason for the shirt selection, for example if challenged on the time of year you might answer, "yes I know it is October, but I am about to go on a holiday to Australia." Similarly XAI can be designed so that it can interactively offer deeper explanations on demand.