Голосовой помощник

Planned workflow, not available software. The design goal is to remove command grammar from the interaction.

1. State the goal. In the intended workflow, you would say what you want to achieve in ordinary language, including the context you already have.
2. Let the assistant restate it. The assistant would describe the task as it understood it, so a misunderstanding is visible before anything happens.
3. Answer a clarifying question. If the request is ambiguous, the assistant should ask one useful question rather than guess.
4. Keep your own wording. You should not need to learn a fixed phrase list. The design target is intent, not incantation.

This is a design description. No voice assistant interface is available to use.

Text + clip The animated clip is available above; the text steps remain below for reference.

Planned workflow, not available software. Confirmation is the safety rail of a voice interface.

1. See the proposed action. Before acting, the assistant would show or say what it is about to do, including the parts it inferred.
2. Correct it. You would be able to change one detail without starting over. Correction should be cheaper than repeating the whole request.
3. Confirm or cancel. A state-changing action waits for an explicit confirmation. Silence is not consent.
4. Review the result. After the action, the assistant would report what happened in the same terms it used for the plan.

This is a design description. No confirmation flow exists in a released product.

Text + clip The animated clip is available above; the text steps remain below for reference.

Planned workflow and design constraints, not available software. Trust in a voice product is built before the first command.

1. See when it listens. Listening should have a clear, visible state. If the microphone is on, the interface should say so.
2. Stop it immediately. You should be able to stop listening without hunting through menus or saying a magic word.
3. Know what is kept. Recording, transcription and retention policies must be published before release, in language a normal person can read.
4. Choose what is shared. Any sharing or training use of voice data should be opt-in, not buried in a default.
5. Hold us to it. Until those policies exist, treat this page as a design intent, not a guarantee about shipped software.

No microphone, app or privacy policy for this project exists yet. This page describes constraints, not shipped behaviour.

Text + clip The animated clip is available above; the text steps remain below for reference.