Text, web, or in person
A resident can ask a question over SMS, from a web page, or by walking up to a tablet at a partner site. No install, no account, no friction for a basic question.
What Concierge does
Four capabilities that show up in every Concierge deployment.
English and Spanish at launch
New languages get prioritized by what partner populations actually speak, not by what the underlying model happens to handle well.
Grounded in real Phoenix data
Concierge answers from the live Resource Nodes catalog and partner integrations. It is not a generic chatbot that reads Wikipedia and guesses.
Says when it does not know
When the model is unsure, it stops trying to answer and hands over a real phone number to a real person. We would rather hear about that escalation than read a wrong answer.
Boundaries
Concierge has guardrails. Hard ones.
A person handles anything safety-critical
Clinical, legal, and safety questions are always routed to a named human responder. Concierge drafts a suggestion, never the decision.
No identifying questions by default
A resident can ask "where is the nearest cooling center" without giving a name or phone number. Concierge collects what is needed, when it is needed, and not before.
Audit log on every conversation
Every exchange is logged for partner review, both the model output and the operator decisions that followed. Partner orgs get quarterly reviews.
Bad answers get flagged in one tap
Operators and residents can mark an answer as wrong, off-context, or harmful. Flags feed directly into the next round of grounding data and prompt updates.
Need a concierge that knows your network?
Bring Concierge to your residents.
Pilots ship grounded in your partner directory and resident-facing services on day one.