You can talk to Nisa instead of typing. Record a voice message in Slack, WhatsApp or iMessage, and she listens.
What she does with it
She transcribes it and treats it as what you said — then answers the request in the thread, the same way she would have if you had typed it.
What she does not do with it
She does not take facts from a transcription straight into your Brain.
Machine transcription mishears, and on quiet or hesitant speech it occasionally invents a whole sentence that was never spoken. A wrong fact in the Brain is worse than no fact, because later nobody remembers where it came from.
So when something you said in a voice message would become durable — a decision, a number, a name — she quotes the sentence back to you first and asks. Say yes and it goes in.
What it costs
A voice message is charged as its own action, once per message — not per second of audio. It costs more than a typed turn, because she has to listen to it as well as answer it.
See Credits for how actions are charged.
What does not work yet
Video. A file with pictures in it is not transcribed, even though it has a soundtrack — otherwise she would be claiming to have watched something she only heard.
