Four artefacts people collapse into one
- Raw audio: the sound in the air, handled by the device as you speak.
- Transcription: the text the recogniser produced from that sound.
- Edited text: what the person changed, fixed or deleted before saving.
- Saved note: the record that exists in the system afterwards.
Why 'we don't keep the audio' is a feature, not a gap
Stored audio is a liability with a storage bill attached. It contains background conversation, other customers, family members, and everything said before someone remembered the microphone was on. It has to be retained, secured, searched and eventually deleted under whatever policy you write.
A dictated note has none of that surface. The person speaks, reads what appeared, fixes it, and saves a sentence. What ends up in the record is what they chose to put there — which is also what makes them accountable for it.
When you actually need a recording instead
There are cases where the sound itself is the point: a verbal authorisation you expect to be disputed, a compliance requirement, an incident. A dictation tool does not serve those cases and should not be stretched to cover them.
For those, the usual answer is a written confirmation the customer acknowledges — an approval in the portal, a signed change order, an email trail — because that is evidence a reader can evaluate without pressing play.
What this means for how you speak a note
- Say the numbers out loud — nobody can go back to the audio to check them.
- Speak in short, complete statements rather than trailing thoughts.
- Name the customer's words as theirs: 'customer said', not 'they want'.
- Read the text before you save it. The text is the record, permanently.