Calling policy and settings

Voice → Policies

A call is allowed only when all of these hold:

• Voice is on for the organisation, with a configured provider.
• The number is dialable. A call to a patient or guarantor needs the patient on file, so consent can be checked.
• Consent to be called is on file and not withdrawn; there is no do-not-call, no do-not-call request from an earlier call, and no cease-and-desist, attorney representation or bankruptcy notice on file.
• It is between 8 AM and 9 PM in the recipient's local time (every time zone the recipient's state spans), narrowed by the state's own contact rules for patients and guarantors, including weekends and federal holidays.
• Frequency limits are not reached: your organisation's attempts per number per day, state limits, and for debt calls the account limits and, unless an administrator has switched it off, the Reg F limit of seven attempts per debt in seven days with no call within seven days of a conversation about that debt.
• The use case's autonomy level allows this kind of call.
• No unresolved high-risk signal from an earlier call blocks it.

Each decision is stored with the policy version it was made under, so you can see which rules applied to any call.

The person sees the reason before placing the call, in these words:

Recording is controlled by policy and is off by default. When it is on, the decision for each call is made in this order:

1. Recording is off for the organisation: not recorded.
2. The patient withdrew recording consent: not recorded.
3. The stricter of the recipient's state and the practice's state applies. All-party states (CA, CT, DE, FL, IL, MD, MA, MI, MT, NV, NH, PA, WA) need a disclosure. A state that is not known is treated as all-party, or as no recording, depending on the setting.
4. Under Record only with consent on file, a patient or guarantor without active recording consent is not recorded.
5. Otherwise the call is recorded, with a disclosure. The current rules never record without one.

The recording decision never blocks the call itself.

Each use case has an autonomy level; the default is level 1, "Person calls, AI drafts notes". Level 2 allows a person to call with live assistance. Levels 3 and above describe AI-agent calls, which are locked on this platform.

Voice → Voice agents

AI-agent calling is not available on this platform. It is locked in the software itself, not by a setting, so turning on Allow AI-agent calls in Settings changes nothing until a release enables it after an owner decision. Today, every call is made by a person; the AI drafts the note afterwards and a person reviews it.

The Voice agents page only configures and checks the scope an AI session would have for each use case: its purpose, and which actions it may attempt (such as verifying a status, getting a reference number or a rep's name, or requesting a call-back). Transfer to a person is always in scope. For each use case the page shows exactly why an AI session would be refused now.

• An AI agent could only ever call payers, provider offices, facilities and referral sources - never a patient or guarantor.
• Clinical decisions, changing treatment, accepting financial liability, promising payment, interpreting coverage, changing patient information and disclosing unrelated health information are refused whatever a scope says.

Only administrators change agent scope.

Voice → Settings

Administrators change these. Until they do, safe defaults apply: calling off, simulated provider, recording off. Every change is audited. The Readiness panel shows what this deployment's configuration actually supports.