Key takeaways
- On outpatient lines, most calls are routine and predictable: booking, moving or cancelling an appointment, asking about an exam, getting ready for a visit.
- An AI voice agent can handle these requests from start to finish in the hospital's scheduling system, as long as it applies each department's rules.
- Your patient access team keeps urgent calls, transferred immediately, complex requests, and patients who want to talk to a person.
- At the Adolphe de Rothschild Foundation Hospital in Paris, the agent resolves more than three-quarters of calls fully autonomously, across more than 25,000 calls handled.
- Roll it out department by department, with your schedulers, and track answered calls, the share resolved without staff and the transfers.
- The vendor handles PHI, so it is a business associate: sign a BAA before go-live and run it through your security review.
It's 8:30 a.m. on the outpatient scheduling line. The phone hasn't stopped ringing, three patients are waiting at the registration desk, and the call queue keeps growing. At the Adolphe de Rothschild Foundation Hospital, a university hospital in Paris, the front desk received between 100 and 150 calls a day before Vocca, and some patients stayed on hold for up to 40 minutes.
If you run patient access or outpatient operations, you have seen a version of that morning. This guide is for patient access leaders, central scheduling managers and directors of outpatient services: what an AI voice agent can take on, what should stay with your team, and how to roll it out without disrupting departments.
Why outpatient phone lines overflow
Outpatient lines overflow because a large number of simple requests arrive at the same time, and the people answering them are also checking patients in. Front desk staff pick up the phone while they register the patient in front of them, prepare charts and deal with schedule changes from providers. Every call interrupts something else, and the patient at the desk waits too.
Calls bunch up in a few hours, first thing in the morning and right after lunch. Most of them are repetitive: an appointment, a reschedule, a cancellation, prep for an exam, a document to bring. And each department has its own rules: which provider, which visit type, how long the slot, how far out.
An outsourced call center absorbs volume, but it rarely knows those rules well. It often takes a message that your team then has to act on. We compare the two approaches, with a cost calculator, in AI receptionist vs. answering service.
Which calls an AI voice agent can handle on its own
An AI voice agent can handle the calls that follow a known rule: scheduling, information requests and administrative steps. Connected to the hospital's scheduling system, it can, during the call:
- book, move or cancel an appointment in the right department's schedule, with the right provider;
- answer questions about an exam: preparation, documents to bring, hours, directions;
- help patients prepare for a visit and complete administrative steps such as pre-registration;
- call patients to confirm an appointment or remind them how to prepare.
The condition is that the agent knows each department's rules. In ophthalmology, for example, a surgery request, a consultation and a pediatric visit are not scheduled the same way. At the Rothschild Foundation, the assistant is trained on the medical topics specific to each specialty, which lets it handle that level of complexity.
What should stay with your patient access team
Urgent calls, complex situations and patients who want to talk to a person stay with your team. An AI voice agent gives no diagnosis and performs no medical triage. Its job is to recognize a call it shouldn't handle, and hand it to the right person:
- urgent calls are transferred immediately, under the escalation rules you define;
- complex requests that can wait get a callback from your team, with a summary of the call;
- a patient who would rather speak to someone is put through, under the hospital's rules.
A medical secretary at the Rothschild Foundation describes it this way: "the assistant correctly identifies urgent situations and transfers them to me immediately. I can focus on the patients who are right in front of me."
One number for every department
The patient dials one number, and the agent applies the rules of the department concerned. For a hospital with many specialties, that is the biggest change: patients no longer hunt for the right line, and every department keeps the way it works. It is what central scheduling sets out to do, without a phone tree.
In practice, the agent identifies the request, routes it to the right department, then applies that department's rules: providers, visit types, slot lengths, held slots, prep instructions. Your team finds the completed requests and the callbacks to make on one platform, instead of switching between tools. Health systems with several hospitals can run the same agent across all their sites, as we describe on our hospitals page and for multi-site groups. Our guide to patient calls in multi-location groups covers the single-number setup.

Rolling it out department by department
Start with a few departments, alongside your schedulers, then expand. The Rothschild Foundation launched the agent in ophthalmology, imaging and dental services before extending it to every department of the hospital. The rollout was done step by step, working day to day with the medical teams and the front desk, the people most affected by the project.
An approach that works:
- Pick one or two pilot departments with high call volume and well-defined requests.
- Write the rules with the schedulers: what the agent handles, what it transfers, what it leaves for a callback.
- Measure for a few weeks before expanding.
- Extend department by department, reusing the rules you have validated.
What to measure
A handful of indicators is enough, as long as every department tracks them the same way:
- the share of calls answered, and hold time;
- the share of requests resolved without staff;
- the number of transfers to your team, and why;
- the callbacks still open at the end of the day;
- patient satisfaction, rated at the end of the call;
- cost, compared with your previous setup.
At the Rothschild Foundation, more than 2,000 reviews collected at the end of calls during the first rollout phase give an average rating of 4.3 out of 5. On the financial side, the hospital reports that the cost of its outsourced call center has been cut by a factor of five, with a wider range of operating hours.
HIPAA and security review: what the hospital should plan for
An AI voice agent that handles patient calls processes PHI, so the vendor is a business associate of the hospital and must sign a BAA before the first call. Bring it into your HIPAA risk analysis like any other system that touches PHI. Most hospitals also run a security or vendor risk review before go-live: plan for it early, because it often sets the timeline.
The questions that matter most:
- Where call data is hosted, how long recordings and transcripts are kept, and which subcontractors process the audio.
- Who can access PHI: access should follow the minimum necessary standard, be limited by role, and be logged so you can see who opened a recording or a transcript.
- Whether call data is used to train AI models.
- How patients are told they are speaking with an AI voice agent. There is no single federal rule for this, so make it hospital policy and check state requirements with your counsel.
Our HIPAA checklist for AI voice agents lists the 15 questions to send every vendor.
A hospital example: the Adolphe de Rothschild Foundation Hospital in Paris
At the Adolphe de Rothschild Foundation Hospital, a university hospital in Paris, the agent resolves more than three-quarters of patient calls on its own. The hospital treats head and neck conditions: ophthalmology, neurology and ENT. Its leadership made the phone the first link in digitizing the patient journey, "not to automate it, but to make the patient's path through the hospital more human."
According to the press release published on April 23, 2026:
- the agent has handled more than 25,000 calls, with a target of 400,000 calls per year at full scale;
- it supports more than 300 physicians, across every department;
- it resolves more than three-quarters of calls fully autonomously: appointment booking, information on examinations, consultation preparation and administrative matters;
- for the remaining quarter, it immediately transfers urgent calls and calls back complex, non-urgent requests.
The front desk team sums it up: "Since Vocca arrived, we are much more relaxed and much more available for patients." Patients can now reach the hospital 24/7. Read the full Rothschild Foundation case study and the announcement of the rollout.
How Vocca fits into a hospital
Vocca answers patient calls 24/7 on a single number and books, moves or cancels appointments directly in the hospital's scheduling system, following each department's rules. Urgent calls are transferred immediately, complex requests are called back, and your team finds everything on one platform. The systems supported at your facility are confirmed with your IT team before launch. See how the agent works on our product page.
For U.S. customers, data is hosted in the United States, encrypted in transit and at rest, and access to patient data is limited to what is needed to handle the conversation. Our infrastructure partners are SOC 2 and ISO 27001 certified, and we sign a BAA as a HIPAA business associate. Our security page has the details, and our team can walk your IT or compliance lead through how Vocca handles patient data.
Is your hospital struggling to keep up with call volume? Explore Vocca for hospitals, or book a demo to watch the agent handle a call from start to finish.
FAQ
Can AI handle a hospital's patient phone calls?
Yes, for requests that follow a known rule. An AI voice agent connected to the hospital's scheduling system books, moves and cancels appointments, answers questions about exams and helps with administrative steps. Urgent calls and complex requests stay with your patient access team, and the agent transfers them.
What share of calls can an AI voice agent handle on its own?
At the Adolphe de Rothschild Foundation Hospital in Paris, more than three-quarters. The agent resolves appointment booking, information on examinations, consultation preparation and administrative matters fully autonomously. The remaining quarter is transferred or called back.
Does an AI voice agent replace patient access staff?
No. It takes the repetitive calls so your schedulers and front desk staff can focus on the patients in front of them and on situations that need a person. As the front desk team at the Rothschild Foundation puts it: "It doesn't replace people, it supports us."
How does an AI voice agent handle urgent calls?
It transfers them immediately, under your escalation rules. The agent gives no diagnosis and performs no medical triage: it recognizes that a call isn't one it should handle and passes it to the right person, under the rules the hospital sets.
What does a hospital need before deploying an AI voice agent under HIPAA?
A signed BAA before any PHI is processed. The vendor is a business associate of the hospital. Include it in your HIPAA risk analysis, run it through your security or vendor risk review, and check where call data is hosted, who can access it and how a breach is reported.







