Upload 3–5 sample emails or marketing assets that represent how you sound. The agent reads these and calibrates its drafts. You can have multiple voice configs (founder, AE, marketing) and assign different ones to different agents.
What good samples look like
- Real, sent messages — not aspirational drafts. The agent learns the voice you actually use.
- Variety — 3–5 samples covering different scenarios (cold outreach, follow-up, post-meeting).
- Recent — within the last 6 months. Voice drifts over time; old samples teach an old voice.
- From the right person — the agent learns the specific person's voice you upload. Match samples to the persona the agent will represent.
Upload the samples
Settings → Brand Voice → New Voice Config → upload samples. Name the config (e.g. "Founder outbound voice" or "AE follow-up voice"). The agent processes the samples in under 2 minutes.
Test the voice
Test → Generate test draft. The agent writes a sample message in the configured voice. Review for tone, sentence structure, vocabulary, and signoff style. If it sounds wrong, add or replace samples and re-process.
Assign voices to agents
Agents → [Agent] → Settings → Brand Voice. Pick the voice config to use. Different agents can use different voices — e.g. the AI SDR uses the SDR persona's voice, the AI Marketing Agent uses the marketing team's voice.
