First, I don't know any particular reason that IOTA would be well suited for artificial intelligence. The more I've looked into it, the more clear it seems to me that IOTA isn't actually very well suited for the things its founders claim to have built it for. They claim something they call "infinite scaling," but it seems that all they mean is that there aren't any particular limits on transaction rates. By allowing an unbounded rate of transactions, they're just pushing their scaling issues somewhere else. They haven't even "taken the training wheels off" yet, and the Tangle is already much too large for a small device to handle. They claim "no fees," but it's likely that full-node operators will need to finance their operations in some way, and a simple way to do that will be to charge transaction fees. In the long run, I suspect that the economics of IOTA will look very similar to those of any other distributed ledger system that is secured by proof-of-work.
Second, I don't know any particular reason that artificial intelligence would find a currency (as we think of it) useful. I suppose you're imagining some kind of economy between various artificially-intelligent computational entities? Ask yourself this: what would an AI use a currency for? To buy things from other AIs? To pay humans to do grunt work for it? If the former, then "currency" would probably look more like "universal barter," a system of what you might think of as "smart contracts" that allow for arbitrary trade of goods, services, and favors. If the latter, then perhaps an AI would want to have a currency to interact with humans. In this case, you should ask if it would be useful for an AI to use a system whose security relies on Proof-of-Work -- or if a more sophisticated and less costly security system might make more sense.
