On longer text boxes in a log form there is a Dictate button in the box, on a computer as well as on a phone. The incident form on your phone has one next to the note box too. Speak the note, and the text appears in the box for you to check and edit before you save.

Dictating

Select Dictate. Your phone asks for the microphone the first time.
Speak. A timer runs while you record. Recording stops on its own after two minutes.
Select Stop. The text comes back with Transcribing… shown while it works.
Read it and fix it. The text lands in the note box. It is a draft like any other — you are the author of whatever you save.

Voice is an accelerator, never a gate. Every failure leaves the box usable and tells you to type instead: "Microphone access was refused — please type your note.", "Couldn't transcribe — please type your note.", "Transcription took too long — please type your note."

If dictation is not configured for your system you will see "Voice transcription isn't enabled yet — please type your note."

Where your voice goes

Spoken notes about a young person are personal data, so the audio is held to the same rule as everything else in ARC.

Your recording is uploaded to ARC, then to Amazon Transcribe in London (eu-west-2), transcribed there, and returned as text. Both the storage holding the audio while the job runs and the transcription itself are in the same UK region.

The recording is deleted

The audio file exists only while the job is running. As soon as the transcript has been fetched, ARC deletes it. It is not attached to the log, not kept as evidence, and not stored anywhere you can go back to — that is a deliberate choice to keep the amount of personal data as small as possible.

The transcript itself is not stored on ARC's side either. It goes into your note box, and from there it is only whatever you choose to save.

Why not the phone's own dictation

Browsers offer a built-in speech feature, and ARC deliberately does not use it. That route sends the audio to Google or Apple, which would take a spoken record about a young person outside the UK. ARC uses its own transcription so the audio stays where the rest of your data is.

What is recorded about the call

One line in ARC's AI audit log: the home, the young person if one was selected, who dictated, where in the app it came from, and how many words came back. Not the audio, and not the transcript.