How Autopilot Decides Who to Contact (Consent, Refusals, Opt-Outs)

Autopilot's selection is conservative by construction, and every “no” is recorded with its reason. If Autopilot isn't touching a contact you expected it to, one of the rules below is why — and that's the system working.

It never asserts consent it can't prove

Autopilot works a contact only when it can point at real evidence of a consent basis: an inbound message from that contact, closed business together, or a first-party sign-up on one of your own capture pages (a form, a lead magnet, an open-house sign-in). A raw cold list has none of those, so it doesn't get worked — by design.

Nothing real to say means no message

Autopilot refuses to write a contact it knows nothing specific about. A touch has to be grounded in something true and sayable — how the contact reached you, what they typed into your calculator, who referred them, a note the agent recorded. Behavioral signals (property views, score changes) can justify the timing of a touch, but are never stated to the contact and can never carry a message on their own. A message with nothing real to say is a generic drip, and generic drips train your leads to ignore your agents' real mail.

Your agents' momentum is never interrupted

A contact the assigned agent has touched in the last 7 days is off-limits, and no score or signal overrides that. Autopilot exists to work the leads nobody is working.

Opted out means out

Unsubscribes, suppressions, STOP replies, and revoked consent are checked before a message is even composed, and again independently at send time. On the SMS bridge, an opt-out also stands the texting platform down on that contact. An opt-out on one channel is never treated as permission on another.

The rest of the guardrails

• Under contract: a contact in a transaction is a human's job — Autopilot stands down.
• Mid-conversation: a contact waiting on a human reply, or being worked by another AI assistant, is left alone.
• Sending windows: touches respect the recipient's local hours.
• Budgets: account-wide and per-playbook daily caps are hard ceilings, reserved before anything sends.
• Compliance and disclosure: every touch — draft or send, email or SMS — passes the Fair-Housing compliance scan and carries the AI-assistance disclosure. There is no setting that turns either off.
• No invented facts: Autopilot never quotes scores, probabilities, equity estimates, or market figures it can't source, and it never claims to have read messages it can't see.

You aim it — by tag if you want

On the Autopilot page, Who Autopilot may work chooses between your whole eligible book and only contacts carrying the ace-autopilot tag in Follow Up Boss — add the tag to a contact, a Smart List selection, or from an action plan, and Autopilot picks them up on its next pass. The ace-autopilot-exclude tag keeps Autopilot off a contact in either mode and always wins. A tag is a scope control, never a consent grant: a tagged contact still has to clear every rule on this page.

Never loops

A playbook that completes without a response closes for good on that contact — Autopilot does not restart a finished sequence on the same person. The contact simply returns to your normal nurture visibility.