Politeness is the product, and most of it is built in: every agent the Builder creates carries a baseline of behaviour that cannot be accidentally removed. This page explains those defaults and what the operator still needs to manage.
This page describes product behaviour, not legal advice. Rules on AI disclosure, call recording and outbound calling vary by where and how an organisation operates, so operators should check their own obligations.
Callers always know it is an AI
An agent's first words disclose that it is an AI, woven into the greeting. Agents never impersonate a person, borrow a real staff member's name, or clone a voice. The full set of guarantees is in What every agent promises your callers.
A human is available on request
Every team the Builder designs includes a defined path to a human. See Teams, transfers & subagents. When the agent does not know something, it says so and offers an alternative. A caller in distress is directed to the right help; an abusive caller gets one warning, then a polite end to the call.
Recording, with clear notice
Calls are recorded by default. Recordings are stored encrypted and any call can be reviewed afterwards. You can turn recording off per agent in the agent editor's Operations section, or by asking the Builder to switch it off. The default is on, so assume a call is recorded unless you have deliberately turned it off.
Because recording is on by default, the notice matters more, not less. The agent's generated instructions include a recording notice so callers hear about it before the conversation continues; keeping that notice is the operator's responsibility.
Outbound calls people expect
polite.ai is inbound-first by design. There is no dialler and no scheduler: an outbound call happens only when the operator's own system requests one through the API, and every destination passes an outbound filter before it is dialled.
Outbound works best where the recipient expects the call and the agent arrives with context: reminders for tomorrow's appointments, promised callbacks or order-ready notifications. Cold calling people with an AI is not what the product is built for. One clear limitation: agents cannot reliably detect answering machines, so the agent's instructions should say what to do if the call sounds like voicemail.