Tuning Mode
Tuning mode is a safety net for onboarding. When enabled, all emails are held for human review regardless of guardrail results — even emails that pass every guardrail. This lets you observe how your guardrails perform on real traffic before trusting them to auto-forward.
How it works
| Tuning mode | Green emails | Orange/Red emails |
|---|---|---|
| On (default) | Held for review | Held for review |
| Off | Auto-forwarded instantly | Held for review |
Tuning mode is on by default for new organizations. Toggle it from Settings > Tuning Mode.
When to turn it off
Turn off tuning mode when:
- You've sent enough test emails to understand how each guardrail performs
- You've adjusted thresholds and settings to match your needs
- You're confident that emails marked "green" by all guardrails are safe to send automatically
There's no rush. Keep tuning mode on for as long as you need. The only cost is the time it takes to manually review and approve each email.
Recommended workflow
-
Week 1 — Keep tuning mode on. Send emails through your AI agent and review every one in the dashboard. Note which guardrails are too sensitive (false positives) or not sensitive enough (false negatives).
-
Adjust — Tweak guardrail settings based on what you've observed. Whitelist expected PII types, customize tone and content policy definitions, adjust thresholds.
-
Week 2 — Continue with tuning mode on. Verify that your adjustments have reduced false positives without missing real issues.
-
Go live — Turn off tuning mode. Green emails now auto-forward. Orange and red emails are still held for review.
-
Monitor — Check the audit log periodically to verify that auto-forwarded emails are meeting your standards. You can always turn tuning mode back on if you need to.