How Tone Inference Works
What Handlet analyses during onboarding, what is excluded, and how you stay in control.
During onboarding, Handlet can estimate a starting tone from recent outbound messages.
What is analysed
- recent outbound messages only
- sampled text within the tone inference window
- redacted message text after direct identifiers are removed
What is excluded
- attachments
- raw long-term training datasets
- personality labels
- demographic traits
- behavioural scores
- segmentation outputs
What you can do next
After onboarding, you can open Settings -> Channels & Tone to:
- review tone calibration
- accept or reject AI recalibration suggestions
- adjust tone manually
- delete stored tone inference data
Metadata shown to users
When an AI-generated suggestion is available, Handlet shows:
- structured reason codes
- confidence when available
If required reason codes are missing, the suggestion cannot be confirmed.