The Front Desk Is Understaffed, and the Phone Is Where It Shows
Health systems have spent the last three years pouring their budget into patient portals, text reminders, and online scheduling widgets. Those tools sit on top of a front desk that remains structurally short-staffed. MGMA's May 2026 member poll of 303 practice leaders found staff turnover has "stabilized but not solved" itself. Front-desk, medical assistant, and nursing support roles see the most turnover and MGMA links that churn to slower check-ins and rising call abandonment. Source: MGMA, "Stabilized but not solved: Staff turnover in medical practices looking no better, no worse in 2026," May 2026
Multiple digital scheduling tools won't fix that. They route around it. A patient with a straightforward, low-stakes visit will happily tap through an app. A patient with a torn rotator cuff with a referral she doesn't understand wants to talk to someone immediately. Phone volume concentrates exactly where the stakes are highest, and that's exactly where a short-staffed front desk struggles most.
We wrote earlier this year about who actually owns the front door of healthcare now that the first contact happens across search, text, and referral queues. Read: Who Owns the Front Door in Healthcare? Voice remains the piece of the front door that health systems have been least able to staff their way out of.
Meet Luna: The Voice Agent Embedded in FastTrackCare
Luna is the AI voice agent embedded in FastTrackCare, Steer Health's omnichannel patient demand capture engine. She isn't a separate product with her own contract like your typical chatbot bolted onto your phone tree. She's the front desk's phone line, running on the same platform that captures web, text, and referral demand, booking it straight into the EHR.
Explore Luna, the Front Desk agent that handles new patient booking, rescheduling, symptom triage, insurance questions, referral intake, reminders, care gap outreach, reactivation, and balance recovery, all in a natural conversation, logged against one shared patient record.
Across live deployments, the numbers hold up:
- 247+ concurrent calls per instance, so a flu-season surge or a marketing campaign spike never produces a busy signal
- Under 400 milliseconds of response latency, consistent with a human beat that patients can't tell the difference
- 88% of calls booked without a staff member touching the call
- 65% of those bookings happen after normal business hours, when the front desk isn't staffed at all
- A 42% reduction in inbound calls reaching human staff, freeing that staff for the calls that require their judgment
- An average handle time of roughly three minutes per call
- Live today across 45+ hospitals and 472 locations
Luna does the job a phone tree was never built to do: finishing the call instead of routing it somewhere to die.
What Changes When Luna Answers
| Legacy phone tree / voicemail | Luna | |
|---|---|---|
| After 5pm and weekends | Voicemail, or nothing | Answers live, books live |
| High call volume | Busy signal, hold music | 247+ concurrent calls, no queue |
| Booking | Callback requested | Books appointments directly into the EHR schedule |
| Insurance questions | "Someone will call you back" | Answered in the same call |
| Referral intake | Faxed, queued, worked manually | Captured and routed in the call |
| Escalation | N/A | Hands off to staff the moment judgment is needed |
The distinction that matters here isn't "AI versus human." It's completion versus deflection. A phone tree that routes a caller to a menu of options considers the call "handled". But, hasn't delivered any outcomes. Luna's 88% completion rate is a measure of finished work, not call volume processed.
Voice Fits Inside a Bigger Shift, Not Ahead of It
Health systems aren't experimenting with agentic AI anymore. They're scaling it. In a Deloitte survey of 100 US health care technology executives from February 2026, 61% said they're already building or implementing agentic AI, or have secured budget to do so, and 85% plan to increase that investment over the next two to three years. Ninety-eight percent expect at least 10% in cost savings, and 37% expect savings above 20%. Source: Deloitte, "Many health care leaders are leaning into agentic AI as adoption hurdles ease," February 11, 2026
Voice is one of the easiest places to start, for a plain reason: a phone call has a clear beginning and end, as well as a clear success condition (booked, rescheduled, answered, escalated). It's also the front door channel patients are most experienced with, so there's no adoption curve to climb. They just call, and someone or something answers.
None of this replaces the EHR. FastTrackCare, with Luna embedded, books directly into the existing EHR schedule across 14+ EHR integrations via FHIR R4 and HL7. The EHR stays the system of record. FastTrackCare is the system of action on top of it.
The Part Most AI Rollouts Skip: Proving It Worked
Adoption is outpacing proof. In a separate Deloitte survey of 64 US healthcare CFOs published in August 2026, 75% of organizations Deloitte classifies as AI "scalers" plan to increase generative AI investment within 12 months, yet only 18% of those said scalers report having mature financial attribution capabilities. Deloitte's own framing for the gap: "Momentum isn't a metric." Source: Deloitte, "Momentum isn't a metric: Many healthcare CFOs report they need clearer proof of AI value," August 12, 2026
That gap is exactly why "calls answered" is the wrong headline metric for a voice AI rollout, while "visits booked, attended, and billed" is the right one. We've written before about how patient engagement metrics can rise while the outcomes underneath them stay flat. Read: The Outcome Gap Behind Your Patient Engagement Metrics A voice agent that books a lot of calls but can't trace them to a completed, collected visit is optimizing the same vanity layer. Luna's bookings are traceable end to end because they run on a platform that handles both the visit and the claim. Luna's not a point tool sitting next to those two processes.
The Referral Call Nobody Follows Up On
The referral is a category of call that deserves its own mention. A referral that arrives by fax or portal message that must wait for the front desk to notify patient is essentially a phone problem with a different hat. We've covered where that referral volume leaks out before a patient is ever contacted. Read: Orthopedic Referral Management: How to Stop Leakage Before It Starts
Luna's referral intake capability means the outbound call that used to sit in a queue for days happens the same day the referral lands, with insurance already checked before the patient picks up.
The Only Question Left
The staffing math on the front desk isn't going to change soon, and the phone isn't going anywhere. Patients with the highest-stakes and highest-intent questions are still going to call, and they're going to call at the worst possible times for a short-staffed office. The system that answers first keeps the patient. The one that doesn't loses her to whoever does.
Book a demo to see Luna answer a live call inside FastTrackCare.