← All posts

Approval by default, not by force

In May we published a post called You approve every message, and that's the feature. It ended: "On purpose. Forever." Forever lasted about two months. This is the honest account of what we changed, and why we think the principle survived it.

Why we changed it

Two things wore the absolute version down. The first: once a voice profile is tuned, the best drafts need nothing. You read them, agree with every word, and click approve. For those messages the click isn't judgement any more, it's friction, and friction is how a review queue turns into a backlog that nobody reads properly.

The second was the opposite case: a message you've already written and tested yourself, going to a list of people you already know. Routing that through an AI draft and a review step adds nothing but a chance to change words you chose on purpose.

Exception one: auto-send for perfect drafts

Every draft OnlyLeads writes is scored before you see it. Points come off for running long, for generic sales phrases ("hope this finds you well", "circling back", "quick call"), for fake familiarity, for a pushy ask, and for not mentioning the reason you're reaching out at all.

Auto-send takes only drafts that lose no points at all: a score of 100 with no warnings. Everything else still waits for you. It's:

  • Off by default. Nothing changes unless someone turns it on.
  • Per LinkedIn account. You switch it on for one sending account at a time, on the accounts page, and only that account's owner or a workspace admin can do it.
  • Inside the normal limits. Auto-sent messages use the same daily caps and the same randomised pacing as everything else. They don't jump the queue.
  • Still stopped by a reply. If the person writes back before the message goes out, it doesn't go out.

Exception two: verbatim steps

A campaign message step can now hold your exact words instead of AI guidance. Write it once with {first_name} where the name goes, and it's sent as written. No drafting, no rewriting, and no review queue, because there's nothing for you to review: you wrote it.

We're explicit about that last part on the homepage too. A verbatim step is a decision you make once, for a list you've chosen, and the campaign still stops for anyone who replies.

What didn't change

The thing we were actually defending in May was this: an AI never decides on its own what gets said in your name. That's still true. Auto-send doesn't lower the bar, it only removes the click for drafts that clear the highest bar we have. Verbatim doesn't involve the AI at all. And for every account where nobody has changed a setting, every message still waits for a human.

Approval is the default because it's the right default. Making it impossible to change was us being stubborn on your behalf.

We'd rather tell you we changed our mind than leave a promise on the blog that the product no longer keeps.

Outreach that respects both sides of the conversation

Deterministic pipeline, LLM drafting where it earns its keep, a human in charge of what gets sent. No agent loops. No credit burn.

Get started →

More from the blog

Does a profile view before you connect actually help? Our numbers · 4 October 2026 Where positive replies go to die, and the pipeline we built to stop it · 4 October 2026