Responsible AI
How Angelica⟡ stays on script, protects members and hands off to people.
Angelica⟡ talks with members on your behalf, so she follows your rules, not her own. These are the guardrails every duty runs with.
She says she is an AI
Angelica⟡ introduces herself as an automated assistant for your organization at the start of a conversation, or says so whenever a member asks. Your team approves the wording, including any disclosure your state requires.1
She stays on an approved script
Angelica⟡ works from a script and approved answers built from your own SOPs. She does not make up policy, and she does not give clinical advice. When a question is outside the script, she follows your escalation rules.
She checks identity first
Before she discusses anything personal, Angelica⟡ verifies who she is talking to with the checks you choose. Callers she cannot verify are tagged, and she shares nothing personal with them.
A person is always one step away
A member can ask for a person at any time. Angelica⟡ makes a warm transfer to your staff during the hours you set, or books a callback. You choose which topics always go to a person.
Every call is reviewed
Every call is reviewed after it ends, and every transcript and recording is available to your team. Before a duty goes live, your team places test calls and signs off. New duties roll out to a small share of members first.
Links carry no health information
Links that Angelica⟡ sends by text use a secure token. They never include health information in the URL.
Footnotes
-
For example, California's AB 2905 (2024) sets disclosure rules for prerecorded messages that use an artificial voice. ↩