Letting AI draft your replies without losing sleep

Approvals, limits and plain labels: the guardrails that make AI in outreach safe to turn on.

AI can write a good reply in a second. It can also promise a discount you don't offer, or answer a lead who asked to be left alone. The answer isn't to avoid AI; it's to put guardrails around it.

Guardrails we built in

• Approval first. AI drafts land in an approval queue. A person reads and sends them.
• The thread is data, not instructions. AI reads the email thread as untrusted content, so a lead can't talk it into doing something else.
• Rate limits. AI requests are limited per workspace, so a runaway loop can't run up a bill.
• Honest status. If AI isn't configured, the app says so instead of showing a fake draft.

Guardrails you should add

• Write down what the AI must never offer: prices, discounts, delivery dates.
• Review a sample of approved drafts each week for tone.
• Keep unsubscribe and complaint handling out of AI's hands entirely.

What comes next

Automatic sending of AI replies will stay off by default, and when we add it, it will start with replies you've approved many times before, like confirming a meeting time.

Try it

Warmup, verification and the team inbox are free. Start free; you only pay $1 per 1,000 emails that send.