How Ember decides what to send — and what it refuses to do

The core idea

Ember is a deterministic, relationship-first agent. It does not run a large language model loosely over your inbox and hope for the best. It reads the actual state of each relationship — past emails, meetings, notes — and reasons, per contact, about whether there is a genuine reason to reach out and what that message should say.

How a decision is made

1. It loads the real history for a contact from your inbox and CRM.
2. It assesses the current state of the relationship and whether action is warranted.
3. If — and only if — there is a real reason, it drafts a message in your voice.
4. A separate quality check reviews the draft before it ever reaches you.
5. You approve. Ember sends from your real inbox.

What Ember does NOT do

Documenting our limits is part of earning your trust. Ember deliberately will not:

• Send anything without your explicit approval.
• Run people through fixed sequences or cadences.
• Blast a list, rotate inboxes, or play volume games.
• Invent facts or claim it sent something it did not.
• Reach out with no genuine reason just to "stay top of mind."

Why timing is a judgment call, not a rule

Good follow-up timing comes from context — a meeting that happened, a timeline the contact stated, a commitment you made — not from a calendar of fixed intervals. Ember anchors timing to real signals and gives you the dates so you stay in control.

Where personalization comes from

Every draft is grounded in the relationship itself, not a template with merged variables. That is what makes Ember's outreach feel personal: because it is. If Ember does not have enough context to write something genuinely specific, it tells you rather than guessing.

The principle behind all of it

Relationships close deals. Ember exists to let you treat every relationship like it is your only one — at the scale of your entire pipeline — without sacrificing the personal touch that makes outreach work.