What is voice mode?
Voice mode is the chat analyst as a live spoken conversation. You ask a question out loud, the analyst turns it into the same tool calls it would make from a typed question, receives computed figures, and speaks the answer, conclusion first. The conversation is continuous, so a follow-up does not need the context restated.
It is the same analyst, not a lighter one. The tools are the analytics engine's tools, the freshness is the freshness of each source, and the answer's transcript carries the same trust footer as a typed answer, so what was said can be checked afterwards.
It is switched on per user from the analyst's own panel, and it can be switched off just as easily for a user who prefers to type. Nothing about the store's data, the tools or the permissions changes with the interface.
Barge-in is what makes it usable
A spoken answer about a category's margin can be long. Barge-in lets you interrupt it the moment you have what you need, or the moment you realise you asked the wrong question, and the analyst stops and takes the new direction. Without it, voice becomes a lecture; with it, voice is a conversation.
Barge-in also handles the shop floor. An owner interrupted by a staff member can pause, return, and pick the thread back up, because the conversation's context is held on the engine's side rather than in the owner's memory.
The rules do not change when the interface changes
The second law holds: the voice analyst cannot produce a number without a backend tool. If you ask for a figure the engine cannot compute, it says which input is missing rather than estimating. The third law holds: if the answer implies an action, the action is prepared for a Review & Send screen, and nothing is sent by saying yes into a microphone.
Honest data holds too. If the returns source is stale, the spoken answer says so, exactly as the footer would, before it gives the figure that depends on it.
Where voice earns its place
On the shop floor, asking which lines in this store are sitting still while looking at the shelf. At a supplier, asking the sell-through and cover on last season's order before agreeing this season's. In the car between two locations, asking what moved on the watchlist. In each case the owner has a question and no screen, and the answer is a computed figure, not a guess.
It is not built to replace the Dashboard. A long investigation belongs in the chat workspace with projects and pinned answers. Voice is for the question that arrives when the hands are busy.
- On the shop floor: which lines in this store are sitting still, asked while looking at the shelf.
- At a supplier: the sell-through and days of cover on last season's order before agreeing this season's.
- Between locations: what moved on the watchlist since the morning, with the history read aloud.
- In a stockroom count: the cost and holding cost of a bin of units, from the dated cost table.
- After a manager's call: a draft reply prepared for Review & Send with the figures already inside it.
What does voice mode keep?
Every spoken conversation is transcribed and kept with its trust footer, so a figure quoted on the shop floor can be found and checked at a desk. A follow-up action prepared in voice, such as a draft email to the store manager, waits in Review & Send with the transcript attached as its origin.
This is the same record-keeping the rest of the engine practises: an answer is a claim, a claim has sources, and the sources are stored with the claim.
How does voice fit the roles on ftes.ai?
The Digital FTEs on ftes.ai report in the workspace, in writing, with evidence. Voice does not change what they produce; it changes how a manager can interrogate it when away from the screen. Asking the analyst what the Buyer FTE proposed this morning, and why, is the same question typed or spoken.
Start with the question you most often carry around unanswered. If the engine can compute it, voice can answer it, with the same footer you would get at a desk.