Healthcare App
Product UX
Case studiesCase template · SayMo
SayMoDictation · macOS app
Services
Tools
Technology stack
Dictation · macOS app
SayMo
SayMo turns speech into text in the Mac app you are already using. Built in-house for our daily work, it also brings AI-agent permission requests to one place, so responding does not mean searching through open chats.
The problem
Dictating a task and answering an agent are small actions repeated across many windows. We wanted both to fit into the work already on screen, with a way to recover a dictation when recognition or insertion did not go as expected.
Design decisions
We designed a compact control at the screen’s edge and shortcuts that start recording without opening another window. Permission requests surface with enough context to answer the right chat.
We made recovery part of the normal workflow: recognized text is saved before insertion, and unfinished recordings remain available for another attempt. Speech recognition runs locally on the Mac.
Our contribution
We defined the capture, insertion and recovery workflows, designed the interface and built the native macOS app. Our daily use guides development. This case shows the current development version.
01 / Agent requests
Bring the waiting request to the person.
A permission request can stop work in one of many open chats. We bring the project, chat and exact command into a panel at the top of the screen, giving each request enough context to make a decision.
The person can allow the request, decline it or give a different instruction. The reply belongs to that request, so answering does not depend on which work window is in front.
02 / Dictation
Start speaking from the task already open.
A keyboard shortcut or side mouse button starts recording while the current app keeps focus. SayMo transcribes on the Mac and inserts the result at the active cursor.
The recording control stays compact, and the trigger and recording mode can be chosen in one settings view. This keeps repeated dictation close to the way the person already works.
03 / Recovery
Save the words before inserting them.
A failed insertion should leave a usable dictation. We save recognized text to history before inserting it, with the time and app to help locate it later.
If recognition fails, the recording is kept for another attempt. History provides a place to search and recover the text. The archive is encrypted on disk, with a history lock enabled by default.
The result
A native tool shaped by daily use.
We designed and built SayMo around our own work across apps and AI chats. The app page lets you explore browser demonstrations of the controls and agent requests. Microphone input and settings in those demonstrations are simulated.
