Key takeaways
- “Books appointments” covers four levels: a request passed on, a time suggested from a schedule view, an agent booking with a login, and booking through a tested software connection.
- Ask whose login is used, which calendars and visit types the service may touch, and what stops two people taking the same slot.
- Every booking should arrive tagged with its source and a clear status, so staff can tell a confirmed visit from a request[2].
- Run scripted test calls, including a slot race and a rule-breaking request, before any real call is forwarded.
Four things “we book appointments” can mean
Four answering services can make the same promise to book appointments and deliver four different things. One writes down the caller's preferred day and emails it to you. One looks at a copy of your schedule, offers a time and leaves the confirming to your staff. One logs into your practice management system and books the visit itself. One books through a software connection that checks open slots at the moment of the call. All four get called scheduling on a sales call. They put very different things on your schedule.
| Level | What the caller hears | What reaches your schedule | Main risk |
|---|---|---|---|
| 1. Request taken | “The office will call you to confirm a time.” | A message with preferred days and times | Phone tag; the patient books somewhere else |
| 2. Time suggested from a schedule view | “Tuesday at 10 looks open; the office will confirm.” | A tentative time for staff to enter | The slot is gone by the time staff try to enter it |
| 3. Agent books with a login | “You're booked for Tuesday at 10.” | A booked visit entered by an outside person | Wrong visit type or provider; shared logins with no audit trail |
| 4. Booked through a connection | “You're booked for Tuesday at 10.” | A booked visit written by software | Your rules not mirrored in the connection; sync gaps |
Scheduling standards draw the same line. In FHIR, the data standard many EHRs use to exchange appointments, an appointment is “proposed” or “pending” until every participant has accepted it, and “booked” only once it is confirmed to go ahead at the date and time specified[2]. Ask each service which of those it creates. A request your staff must still confirm is useful, but it is not a booking, and a patient who heard “you're booked” will turn up expecting one.
Example script for levels 1 and 2, so callers know exactly what they have: “I've noted Tuesday morning as your first choice and Wednesday afternoon as your second. This is a request, not a booking yet. The office confirms requests by 10 am on weekdays and will call from this number. If you don't hear from us by then, please call back and mention your request.”
That last sentence matters. It gives the patient a deadline and a way back in if the request falls through the cracks, instead of a silent wait that ends with them booking elsewhere.
Why practices want this off the desk
Scheduling is the phone work practices most want help with. Asked which phone tasks take staff the most time, practice leaders put eligibility and prior authorization first at 45% and scheduling second at 31%, well ahead of intake at 9% and refills at 6%[1]. Among practices returning to AI and automation to improve access, scheduling is the first target, named by 31%[3].
Source: Medical Group Management Association (MGMA), 2026[1]
Show the data
| Phone task | Share of leaders |
|---|---|
| Eligibility and prior auth | 45% |
| Scheduling | 31% |
| Intake | 9% |
| Prescription refills | 6% |
Online booking has not taken the load off the phone. Only 11% of medical group leaders said most of their patients self-schedule, and 73% said a quarter or fewer use digital scheduling tools at all[4]. Most bookings still start with a call, and after hours that call goes somewhere: to voicemail, to a message-only service, or to a service that books. Online booking vs phone booking looks at which patients choose which.
Schedule access: whose login, which calendars
Levels 3 and 4 need access to your system, and the details of that access decide how safe the setup is. Settle each of these in writing before go-live.
- Named logins. Every agent who books gets their own user in your practice management system, so each appointment shows who made it. A shared “answering service” login hides every mistake.
- Limited rights. Booking and viewing the appointment book, not clinical notes, billing screens or more demographics than booking needs.
- Named calendars. List the providers, locations and visit types the service may book. Everything else stays a request.
- A business associate agreement. A service that sees your schedule handles patient information, which HIPAA allows only with satisfactory assurances in a business associate agreement[5].
- Removal. Who switches off access when an agent leaves the service, and how quickly.
Software connections are harder for small practices than the brochures suggest. Only 32% of solo physicians used an EHR from one of the five leading developers in 2024, against 90% in practices of more than 50 physicians[6]. A service that connects to the big systems may have nothing tested for yours. Ask for a reference practice on your exact system and version, and read what EHR write-back involves before relying on a level 4 promise.
The booking rules a service must follow
A service books only as well as the rules you hand it. Write them on one page, in the words patients actually use, and expect to revise the page after the first month of real calls.
- Visit types and lengths, with the plain phrases callers use for each: “a physical”, “a sick visit”, “my blood pressure follow-up”. Setting appointment types and visit lengths helps here.
- New or established patient, and which providers are accepting new patients this month.
- Plans accepted, plans not accepted, and the exact wording for a caller whose plan you do not take.
- Visits that need a referral, prior authorization or outside records before anything can be booked.
- Same-day slots held back for sick visits, and who is allowed to release them.
- Age limits, provider preferences and visits that need prep instructions read to the caller.
- What the service must never do: book over blocked time, double-book outside your template, or promise a named provider to a patient who has never seen them.
Self-pay patients need a rule of their own. When an uninsured or self-pay patient schedules a service at least three business days ahead, the provider owes them a good faith estimate within one business day, or within three business days if the visit is ten or more business days out[7]. Decide who sends it when the service did the booking, or the clock runs out overnight. See a good faith estimate workflow.
How double bookings happen, and how to stop them
Double bookings are a timing problem. Your front desk, your online booking tool and the answering service can all be looking at the same open 10 am slot. Whoever saves first gets it, and everyone else creates a clash unless the system checks availability again at the moment of saving.
So ask how the service sees availability. A live login or a live connection sees the real schedule. A screenshot, a printed template or a nightly export is out of date the moment a patient cancels. FHIR gives slots a status of free, busy, busy-unavailable or busy-tentative, the last for time that has been tentatively scheduled[8]. A service that holds a slot while it talks to a caller should mark it tentative and release it if the call ends without a booking. Ask whether it does.
- What happens if two people book the same slot in the same minute?
- How often is availability refreshed, and is it checked again before saving?
- Can the service see blocked time, provider leave and held same-day slots?
- Who fixes a clash, and who tells the patient?
When double-booking is a deliberate part of your template, the service needs those rules too; double-booking patients covers when it works.
How bookings should reach your staff
- Caller
- System
- Your team
01Caller
New patient calls, 6:30 pm
Asks for a first visit with any provider who takes their plan.
02System
Service checks the rules
Plan accepted, provider taking new patients, visit type matched.
03System
Time offered as pending
The caller hears that the office will confirm by 10 the next morning.
04Your team
Front desk confirms
Staff enter the visit, check for clashes and attach any forms.
05Your team
Patient told it is booked
Confirmation goes out from the practice's own system with date, time and address.
Example. At levels 3 and 4 the service books in the third step, and staff review a booking report instead of confirming each one.
Whatever the level, a booking should arrive in a form staff can act on in seconds:
- A source tag on every appointment the service creates, so you can count them and audit them.
- A status: booked, or a request awaiting confirmation. Never a vague “appointment set”.
- A list of everything booked or requested overnight, read by a named person before 9 am.
- The call recording or notes behind each booking, so a disputed one can be checked.
- A correction route: if a booking breaks a rule, who calls the patient, and who tells the service so the rule gets fixed.
Test calls before you sign
A clean booking
A new patient with an accepted plan asks for a routine visit. Check visit type, provider, length and the confirmation wording.
A plan you don't take
The caller should hear your approved out-of-network wording and not be booked.
A visit that needs a referral
The service should follow your referral rule or take a request, not book on the spot.
A slot race
Book a slot at the front desk while a tester books the same slot through the service, and see who wins and what the loser is told.
A cancel and a reschedule
Check that the freed slot reopens and the old visit is cancelled, not left sitting on the schedule.
A caller with symptoms
Any booking path must still give your emergency instruction and hand clinical questions to a person.
Score each call against the recording, not the service's own report of it. Common scheduling errors is a good list of what to listen for.
Where an SI Employee fits today
On a medical line, an SI Employee does not book visits. It answers the questions that come before a booking, whether you are taking new patients, which plans you accept and what to bring, from the list your practice signs off, and then puts the caller through to your scheduling line. Booking into a practice management or EHR system is planned, and it depends on a tested connection and a signed business associate agreement, which AssistCall does not offer today. Until then the appointment itself stays with your staff. What connects today, and how, is on the integrations page.
The physical therapy scheduling example models that split for a three-clinic group: office questions answered in full, every scheduling call sent to a coordinator, and the coordinators' freed time spent filling cancellations. Whichever level of booking you buy, price it against what unanswered scheduling calls cost you now.
| Input | Example |
|---|---|
| Scheduling calls a week | 300 |
| Share unanswered or sent to voicemail | 8% |
| Share from new patients | 16.8% |
| Share offered a visit when answered | 72.4% |
| Value of a new patient's first visits | $160 |
| Unanswered calls a month | 104 |
| From new customers | 17 |
| Bookings that would have followed | 13 |
| Monthly value at risk | $2,022 |
How it is worked out: Calls a week × 4.33 weeks × unanswered share × new-customer share × booking rate × average value.
Assumptions
- New patients made up 16.8% of US physician office visits in 2019.[9]
- In a Texas audit of primary care offices, 72.4% of simulated new-patient callers were offered an appointment.[10]
- Call volume, the 8% unanswered share and the $160 value are invented for this example; take them from your phone log and billing data.
Example. Replace the inputs with your own figures; the result is value at risk each month, not a forecast.
Frequently asked questions
Can an answering service book directly into my EHR?
Some can, either through an agent login or a tested software connection. Ask for a reference practice on your exact system and version, and remember the service will handle patient information, which needs a business associate agreement[5].
What is the difference between an appointment request and a booking?
A request still needs someone at the practice to accept it; a booking is confirmed for the date and time given. The FHIR standard marks the first as proposed or pending and the second as booked[2]. Make sure callers hear which one they have.
How do I stop an answering service double-booking?
Insist on live availability, a re-check at the moment of saving and a tentative hold while the agent is on the call[8], then prove it with a slot-race test before go-live.
Should after-hours callers be booked into tomorrow's same-day slots?
Only if your rules say so. Many practices hold same-day slots for sick visits sorted in the morning and let the service book routine visits further out.
Who sends the confirmation when the service books?
Ideally your own system, so the patient sees the practice's name, address and any prep instructions. If the service sends it, approve the wording and check it matches what is on your schedule.
References
Medical Group Management Association (MGMA) (2026). Phones are still a bottleneck costing medical practices time they can't afford (opens in a new tab) ↑
HL7 International (current). AppointmentStatus value set (FHIR R4) (opens in a new tab) ↑
Medical Group Management Association (MGMA) (2026). Automatic for the people: AI moves for medical practices to boost the front office and access (opens in a new tab) ↑
Medical Group Management Association (MGMA) (2024). Putting the power of scheduling into patients' hands (opens in a new tab) ↑
Legal Information Institute, Cornell Law School (Electronic Code of Federal Regulations text) (current). 45 CFR § 164.502 - Uses and disclosures of protected health information: General rules (opens in a new tab) ↑
Office of the National Coordinator for Health IT (ASTP/ONC), HHS (2024). Office-based Physician Electronic Health Record Adoption, 2008-2024 (ONC Data Brief) (opens in a new tab) ↑
Legal Information Institute, Cornell Law School (Electronic Code of Federal Regulations text) (current). 45 CFR § 149.610 - Requirements for provision of good faith estimates of expected charges for uninsured (or self-pay) individuals (opens in a new tab) ↑
HL7 International (current). SlotStatus value set (FHIR R4) (opens in a new tab) ↑
CDC National Center for Health Statistics (2019). National Ambulatory Medical Care Survey: 2019 National Summary Tables (opens in a new tab) ↑
JAMA Network Open (via PubMed Central) (2020). Association of Simulated Patient Race/Ethnicity With Scheduling of Primary Care Appointments (JAMA Network Open) (opens in a new tab) ↑