This is the question we respect most, because it's the right one to ask. Not "is my data safe" — that's table stakes. The real concern is: what if the AI writes something off-tone, or misreads a prospect's post and responds in a way that's technically fine but socially awkward?

By default, Reachly drafts every message into an approval queue. Nothing sends until you review it. The AI does the research and writing — you keep the judgment call. That single design decision is what makes the rest of this manageable.

"The agent doesn't have opinions about your prospects. It reads, drafts, and waits. The judgment call stays with you until you decide it doesn't need to."

What the AI can do

1. Read a prospect's profile, posts, and mutual connections to draft a personalized opener

2. Write follow-up messages that reference the original outreach

3. Adjust tone based on the settings you configure per campaign

4. Flag prospects where it found very little to personalize against

What it cannot do

It cannot send anything without passing through your approval queue, unless you've explicitly enabled auto-send. It cannot message anyone outside the list you've defined. And it cannot learn from or apply context across customers — your campaign data and tone settings stay isolated to your account.

Quick tip: review your first 50 or so drafts carefully and edit anything that's slightly off. The AI calibrates to your corrections over time, and your review time drops significantly after the first few campaigns.

If something does go out wrong

If you've enabled auto-send and a message goes out that you wouldn't have sent yourself, you can pause the campaign immediately from your dashboard and edit the remaining queue before it continues. Trust is something the system earns gradually — most customers keep manual approval on for the first two to three weeks, then loosen it as confidence builds.