Interesting question. My thought is that the 'compression mode' model of language - that it doesn't actually communicate very much, but relies on the recipient having a similar enough understanding of the world to the sender to decode it - is relevant here. I'm not sure, but it seems at least plausible to me that English and other similar languages are compressed in such a way that while an AI could decode them, it wouldn't be very efficient and wouldn't necessarily be something that we would want.

ETA: If this is the case, conversational Lojban probably has the same problem, but Lojban appears to be extensible in ways that English is not, so it may do a better job of rising to the challenge by way of something like a specialized grammar.