This is good stuff, I’m working on something where this will be useful thinking for a GUI app.

However… I’m very sceptical about VUI in this regard. For short interactions, fine, but getting people’s names spelt correctly? Having people check the address details they’ve given? That’s best done with a screen, or at least confirmed using one.

Voice input is transient; screen input isn’t. In the context of many forms, where you need the user to check the details, a screen is both more accurate and a faster method.