question

So, I think there are a couple things here. First of all, as I discovered recently Alexa has a really hard time doing anything even semi-complex from a resting state. Once it's in your skill, it's a lot easier for it to pick an utterance that maps to an intent with a high level of confidence, but making the jump into your skill seems to be a big leap. I had a sort of meltdown about this the other day after working for a couple months on a skill and realizing right before submission to certification that we couldn't use it the way we wanted. You can read that thread here (jjaquinta goes into a bit of detail about the natural language parsing reasons for why this happens): https://forums.developer.amazon.com/forums/thread.jspa?threadID=10102&tstart=0 Beyond that, though, I think the problem with that specific utterance could be that you're having the utterance match two different things. Every alexa skill has, by default, some triggers for the LaunchRequest. Things like "Open skillname" and "launch skillname". With your sentence, you're also adding an intent on to there, "and give me random dog name". The problem is that Alexa will parse this sentence and see that it matches two different things pretty well - it matches the launch request, and it matches the "random dog name" intent. You didn't say exactly what is happening in your case, but I'm guessing it's just going into the default "launch" state, which means it's probably matching the launch request with a higher confidence. That said, I think what you want is this: (From a resting state, no active skill) "Alexa, ask dog name to give me a random dog name" "[b]ask/tell[/b] [i]skillname[/i] [b]to[/b] do something" is the pattern for launching straight into an intent. Like I said above, though, the voice matching on requests of that form is much flakier, to the extent that my group just gave up on it entirely.