How naturalistic is this sound change?

[collect_gluesticks]

I'm thinking of a (diachronic) sound change rule where aje -> ai̯ and ajo -> au̯ when unstressed, and before a word boundary. But I'm not sure how naturalistic it is.
My conlang's phonology permits the diphthongs ai and au, but not others like oe or ou.
I have seen examples of e, o -> i, u in natural languages, but not with these particular conditions.

If it's somewhat naturalistic, I'd like to include it, because applying this change to words in the protolang produces interesting effects. In particular, it would introduce an irregularity in how one of the prepositions inflects, which I like. But I am wary doing something contrived in order to justify it.

[VaptuantaDoi]

Yes that is perfectly naturalistic. It would probably also affect aji and aju if they occur.