HaloNineHealth

HaloNine Health operates a telehealth nurse line for a regional hospital network, handling roughly 1,800 patient calls a day across symptom triage, medication questions, and post-discharge follow-up. Call volume had grown steadily for two years, but nurse staffing had not kept pace, and queue times during peak hours were stretching past acceptable limits.

Nurses were manually deciding call order based on whatever context they could gather in the first minute of a call, with no consistent way to flag a subtly urgent case (a post-surgical patient describing symptoms that sound minor but warrant escalation) ahead of a routine medication refill request sitting in the same queue.

HaloNine needed a way to surface urgency signals earlier and reduce the documentation burden per call, but any AI-drafted guidance had to go through a licensed nurse before reaching a patient, given the clinical and liability stakes involved.

Aivora AI worked with HaloNine's clinical informatics team for four weeks to define the triage criteria the copilot would score against, pulling from HaloNine's existing nurse triage protocols rather than inventing new clinical logic. The system ingests the call intake summary in real time and produces an urgency score plus a draft response, both visible to the nurse before any action is taken.

The architecture separates the scoring model (a smaller, low-latency classifier tuned on anonymized historical call data) from the drafting model (an LLM that composes a suggested nurse response using approved clinical language templates). Every draft carries a visible confidence indicator and a one-click path for the nurse to edit before sending, keeping the clinician firmly in control of every patient-facing message.

The team ran a two-month pilot on the overnight shift, where volume-to-staff ratio was highest, before expanding to all shifts over the following ten weeks. HIPAA-compliant data handling and full audit logging were built into the pipeline from the first prototype.