I really enjoy the current Voice experience, especially the natural conversational flow. However, one major UX problem remains: as a user, I do not have a clear mental model of what is currently available.
Three things should be visible directly in the Voice interface:
1. Remaining Voice allowance
Show how much Voice usage remains before a limit or model fallback is reached.
This does not necessarily need to be a distracting timer. The existing Voice sphere itself could communicate it elegantly:
- gradually decrease in size as the allowance is consumed, or
- transition from green toward yellow and finally red as the limit approaches.
The important point is that the user should never be surprised by the session suddenly changing or ending.
2. Active Voice model and fallback
Clearly show which Voice model is currently active and when an automatic fallback occurs.
I immediately noticed a substantial change in conversational character when the system switched models, but the interface gave me no useful explanation of what had happened.
A small label such as “Live” / “Live Mini” and a brief notification when switching would solve much of this confusion.
3. Available capabilities in the current context
Voice can apparently have different capabilities depending on device, mode and product context. For example, web access may be available while actions involving Work, email or other tools may not be.
The user should not have to discover these boundaries through trial and error.
Add a simple capability indicator showing what the assistant can currently do, for example:
Available now: Web · Files · Email · Calendar · Work actions
with unavailable capabilities visibly disabled.
This is not primarily a request for fewer restrictions. Restrictions are understandable. The problem is invisible restrictions.
Users can adapt very well to boundaries when the boundaries are clear. What creates frustration is discovering them only after asking the assistant to do something.
In short: make the system state legible.
The Voice interface already feels remarkably natural. Giving users a clear view of remaining usage, the active model, and currently available capabilities would make it feel much more coherent and trustworthy.