Platform staff voice and spoken responses¶
Platform staff voice lets authorized Zerohive staff prepare an assistant request by speaking, review the resulting text, and request an audio version of an eligible assistant response. Speech is another way to interact with the staff assistant; staff remain responsible for the instruction and any decision it leads to.
Who this serves¶
This capability belongs to the team operating Zerohive. It can support staff who prefer dictation or listening, but it does not grant additional authority. Tenant administrators, managers, employees and external participants do not gain staff access through voice. The organization's separate tenant speech experience has its own access and commercial rules.
Managers and executives should distinguish easier input from automated approval. Recognizing a recording, accepting its text, approving a proposed action and completing the action are separate outcomes.
Current availability¶
The backend implements recording, language detection, transcript correction, exact submission, server voice preferences, spoken responses, cost accounting and audio cleanup. The full staff voice interface and deployed environment have not been verified.
The inspected Platform Final Screen shows a staff action review, but no recording, transcript-review or playback controls. The guide therefore explains the supported workflow and recovery rules without inventing button names or claiming a completed screen journey. Staff voice design work remains open.
What happens to a spoken request¶
A recording is converted to text. The staff member reviews and, where necessary, corrects that text before sending it to the assistant. A proposal still requires a deliberate decision tied to its current details. Listening to a proposal does not approve it.
Recognition supports short recordings up to two minutes. The available languages and voices depend on the server's currently qualified assets. An uncertain or unsupported recording requires correction or a fresh text request; the system must not silently turn it into an instruction.
Costs and information lifecycle¶
Staff recognition and voice generation use platform internal cost accounting. They do not spend the affected tenant's speech credits. Recognition and generation are separate measured jobs, and assistant processing can have its own cost.
A failure to deliver text or audio does not establish zero usage. Unknown usage stays unresolved until matching evidence is available.
Audio has a ten-minute validity window from creation of its recording or spoken-response record. Physical cleanup happens separately and can be delayed. Saved transcripts, submitted instructions and accounting records have a different lifecycle; audio expiry does not erase them.