Consent & contact eligibility
Consent is one input among seven to a single expiring decision about one proposed contact, and on its own it never authorizes a call.
Consent is an input to contact eligibility, not its synonym. An active grant cannot override a bankruptcy restriction, cease instruction, inconvenient-time policy, wrong-party risk, or resolved debt. A team that reads consented as may call ends up phoning a protected borrower at seven in the morning about a debt they have already paid, and every one of those facts was available at the time.
POST /v1/servicing/calls/outbound, the scheduled callback’s start-call command, email-thread replies — and in voice dial authorization, so the decision is made by whatever is about to act. The standalone planning contract is described at the end of this page.- Destination verification and deliverability
- Consent grants, texting permission, and revocation
- Customer local time and convenient-time facts
- Contact history and debt-scoped budgets
- Cease, representation, bankruptcy, dispute, fraud restrictions
- Current debt, case, payment, and promise state
- Applicable tenant policy and rule versions
Seven inputs, one expiring decision
Every term can change between planning and execution, which is why the decision expires rather than sitting on the borrower.
The worker’s clock is irrelevant
Policy resolves the customer’s applicable local time from approved facts, with explicit behavior for conflicting or unknown zones and daylight-saving transitions. Missing a required zone fails closed or routes to a narrower safe channel. A dialler running in one time zone reaches a borrower three zones away before breakfast, while every clock in the system reads mid-morning.
A revocation reaches queued work too
An SMS STOP, voice request, portal change, or other verified opt-out closes the relevant consent interval and may create a broader restriction review. Cached eligibility is invalidated immediately and queued work rechecks before execution — the calls already sitting in a dialler are precisely the ones a borrower has just asked you not to make.
Passing one consent passes only that one
Recording consent, call permission, and artificial/prerecorded-voice policy are separate controls. Agreeing to be called says nothing about being recorded, and neither says anything about hearing a synthetic voice. The configured journey records capture method, evidence, terms version, effective time, and applicable destination/purpose, so which permission was actually given stays readable long after the call.
Decision is not send
Planning and execution stay separate on purpose. The interaction endpoint owns execution and re-evaluates volatile facts immediately before provider invocation, so a caller cannot attach an earlier decision ID and force a send once it has gone stale.