iOS app and voice commands

Built for owners on the move, and works without signal. Voice is optional; touch does everything just as fast.

Design principles

Three things, not thirty. Today shows at most three priorities, in HQ’s rank order, each with a status pill and one obvious action.
Calm by default. Dark, spacious, few colors. Motion only to confirm an action. Notifications are batched and respect quiet hours.
Minimal steps. Acknowledge in one tap. Log progress in two taps or one sentence.
Voice optional and equal. A persistent, unobtrusive mic button. Every voice result is shown on screen and is editable.
Offline-trustworthy. Pending changes show a quiet “will sync” mark, never an error wall.
Explain the why. Each priority shows HQ’s one-line reason it matters to the unit.
Accessible. Dynamic Type, VoiceOver labels, large touch targets and checked contrast.

Signing in

Owners sign in once with Continue with Google. The app keeps you signed in, so you aren’t asked to log in every time you open it. Passkeys, Face ID unlock and Sign in with Apple are Planned.

The five voice commands

Tap the mic (push-to-talk) and speak naturally. SignaUnit understands five kinds of request:

Intent	Example	What it does
Brief	“What matters today?”	Reads and shows your top priorities, due dates and anything at risk
Log	“Log a 92 percent review-response rate for the north unit.”	Records a check-in on the right priority and unit
Note	“Note: seasonal promotion launched Tuesday.”	Adds a note to a unit or priority
Acknowledge	“Acknowledge the new customer-greeting standard.”	Acknowledges the current version of a priority
Status	“How am I doing on repeat customers?”	Reads back status and trend

There are no other voice actions. SignaUnit won’t take free-form instructions, and it can’t do anything by voice that you couldn’t do by touch.

Confirm before every write

Anything that changes data, meaning logs, notes and acknowledgements, is shown on a confirmation card before it’s saved:

Log 92% for North unit on Review replies? Confirm / Edit

You confirm with a tap or by saying “yes”. Read-only requests, the brief and status, don’t need confirmation. Once confirmed, the change goes through exactly the same path as a touch update.

How speech is processed

On-device speech recognition. Your iPhone turns speech into text on the device, using Apple’s on-device recognition. Raw audio never leaves the phone.
On-device understanding first. A small on-device grammar handles the five intents, matching unit and priority names against what’s cached on your phone. This also works offline.
Server help only when needed. If the on-device step isn’t sure and you’re online, the transcript text (not audio) is sent to SignaUnit, where a language model maps it onto one of the five actions. The model can only return one of those actions in a fixed format.
On-device speech output. Spoken replies, such as the brief, use the iPhone’s built-in voices.

Privacy

No audio is stored or uploaded. Only text is ever sent, and only when the server step is used.
Transcripts for interpretation are kept for 30 days for debugging and then deleted, unless your brand sets a different retention period.
Transcripts behind notes become the note’s text, which you can edit.
Requests to AI providers carry IDs rather than names wherever possible. Providers must offer zero-retention, no-training terms, and your brand’s operations content is never used to train a public model.
The mic indicator and permission prompts follow Apple’s rules. Voice can be turned off per person and for the whole brand.
Voice is for owners and HQ users, about outcomes. It is never used to direct franchisee staff.

Offline voice

Because speech recognition is on the device, voice works without signal. The five intents are handled against cached data. If a request can’t be understood offline, it’s saved as a draft note with the transcript and a “will interpret when online” badge, so nothing is lost.

Full touch parity

Every voice action has a touch equivalent:

Brief → the Today screen.
Log → open the priority, enter a value, save.
Note → the note button on a unit or priority.
Acknowledge → the Acknowledge button.
Status → the priority detail view.

Owners who prefer a computer get every function on the web except voice.

Planned additions

Siri and Shortcuts, so “What matters today in SignaUnit” works from the lock screen or the Action button. Planned
Server speech recognition as an opt-in for long notes in noisy places, enabled per brand. Planned
iPad layout. Planned
Android, if brands ask for it. Planned