A multi-site primary care group we talked to last quarter was running an 18% no-show rate across four locations. Twelve providers, average reimbursement around $185 per visit, roughly 22 patient slots per provider per day. The arithmetic is brutal: about $200K a year in lost revenue, before counting the staff time spent calling no-shows the next day to rebook them.
They had reminder texts turned on in Athena. One text, 24 hours out. That is the modal setup in US ambulatory care, and it leaves a lot of money on the floor.
This post is about what the data actually says on no-shows, what reminder programs move, and where AI voice scheduling earns its keep beyond just being a louder reminder.
The 10-30% range hides specialty-driven extremes
National datasets on outpatient no-shows have been remarkably stable for two decades. Across US ambulatory care the average sits in the 18-23% range depending on year and source, but the cross-specialty spread is what matters operationally.
- Behavioral health and psychiatry: 20-50%, with community mental health centers regularly reporting north of 35%.
- Primary care and internal medicine: 10-20%.
- Pediatrics: 12-25%, higher in Medicaid populations.
- Dental (general): 5-15%.
- Specialty surgical clinics (ortho, ENT, urology pre-op): 1-5%, because patients are usually motivated by acute pain or scheduled procedures.
- Dermatology and aesthetic: 8-12%.
The variation is not random. Specialties where patients carry shame or stigma (mental health, addiction medicine, weight-loss clinics) see the highest silent no-shows. Specialties where the patient sought the appointment to relieve immediate pain see the lowest. Operators who benchmark themselves against a national average without segmenting by specialty are usually either feeling good about a number that should worry them, or beating themselves up over one that is structurally normal.
A no-show is worth more than the lost copay
The cost-per-no-show conversation usually starts at the reimbursement number and stops there. It should not.
For a primary care visit reimbursed around $150-$200, the direct lost revenue is one piece. Add to it:
- The fixed cost of the slot (provider time, room, MA support) which does not go away when the patient does not show.
- The opportunity cost of a patient who could have filled the slot off the waitlist if anyone had known 12 hours in advance.
- The downstream care a no-show patient does not receive (lab orders not placed, refills not authorized), which often shows up as an urgent-care or ED visit later and costs the system more.
- Staff time the next day chasing the rebook, typically 8-12 minutes per attempt.
Specialty clinics with higher per-visit revenue see numbers above $300 per no-show easily. Surgical consults and specialty infusion can run over $500. Annualized for a single full-time provider, no-shows cost most clinics $50K-$150K a year in lost revenue, and that excludes the downstream cost-of-care effects.
The number worth knowing for your own clinic is not the industry average. It is: (no-show rate) x (slots per day) x (working days) x (average reimbursement). Run that for each provider. The variance between providers in the same clinic is often larger than the variance between clinics.
Reminder programs cut no-shows 20-40%, depending on how many touches
The literature on appointment reminders is dense and mostly converges on the same conclusion: reminders work, multi-touch is better than single-touch, and the marginal touch decays fast.
Single 24-hour text reminders cut no-shows by roughly 15-25% relative to no reminder at all. Most clinics are here, on Epic MyChart, Athena, or a bolt-on like Solv or Phreesia.
Multi-touch programs (3-day text, 24-hour text, 2-hour text-or-call) cut deeper, into the 30-40% reduction range. The 2-hour reminder is the underused one: late enough that the patient has decided whether they will go, early enough to get them to confirm or release the slot.
Voice reminders specifically beat text reminders for older patient populations and for behavioral health, where the conversation matters more than the prompt. AI voice gets you the staffing economics of texting with the response rates of human calls.
Beyond about three reminder touches, additional reminders annoy patients faster than they reduce no-shows. The diminishing returns curve is real.
Reminders are the floor, not the ceiling
If reminders were the only thing AI voice did, the ROI math would still pencil for most clinics. But reminders are the lowest-leverage part of an AI scheduling stack. Four features matter more.
Rescheduling on the call. When a patient says "actually I cannot make it," a text reminder bounces them to a portal that 30-50% of patients abandon. AI voice keeps them on the line and rebooks them in real time, usually inside 90 seconds. The slot gets released back to the schedule, where waitlist logic can fill it.
Waitlist filling. Most EHRs have a waitlist field that nobody uses because the workflow to call through it is human-expensive. AI voice can call the next three waitlist patients in priority order the moment a cancellation hits, and book the first one who confirms. Behavioral health clinics with chronic capacity constraints get the biggest lift here.
Cancellation routing. Same-day cancellations are not no-shows, they are recoverable revenue if the gap-to-book is short. The KPI to track is "minutes from cancellation to next-patient-booked." Pre-AI, this is hours or never. With voice scheduling it can be under 10 minutes during business hours.
Insurance verification at booking. A meaningful percentage of "no-shows" are actually patients who were told at check-in that their insurance lapsed or their copay was higher than they expected, and who walked out. Catching eligibility problems at the booking call (or the day-before reminder) reroutes the patient before the wasted slot, not after.
The "make-it-easy-to-cancel" effect is real
There is a counterintuitive finding from behavioral health research that gets ignored because it sounds like it should hurt the business. Giving the patient a frictionless way to cancel reduces silent no-shows more than it increases cancellations.
Mechanism: a patient who knows they cannot make the appointment but has no easy way to cancel will simply not show up. They feel embarrassed, they avoid the call, they ghost. If the AI voice reminder ends with "would you like to keep, reschedule, or cancel?" and the patient picks reschedule or cancel, the slot is recoverable. If the only options are silence or a portal login, the slot is lost.
Behavioral health and addiction medicine see the largest swings here, because shame is doing more of the work than logistics. Dental and primary care see smaller but still meaningful effects.
This is the part of AI voice scheduling that operators usually undervalue when they buy. They optimize for "did the reminder reach the patient" instead of "did we recover the slot." The metric to watch is recovered-slot rate, not contact rate.
Specialty matters more than vendor choice
The right scheduling stack depends on which no-show pattern dominates your specialty.
- Behavioral health: rescheduling and waitlist matter most. Reminders alone underperform because the underlying cause is avoidance, not forgetfulness. Voice beats text.
- Primary care: multi-touch reminders plus same-day cancellation routing. The lift is moderate but compounds across many slots.
- Dental: text reminders work disproportionately well. Most dental no-shows are forgetfulness or schedule conflicts, both of which a 24-hour text catches. AI voice earns its keep on hygiene recall, not reminders.
- Specialty surgery: focus on insurance verification and pre-op compliance, not reminder cadence. No-show rates are already low; the loss per no-show is high.
- Pediatrics: parent-friendly multi-channel (text plus voice for after-school confirmation). Medicaid populations benefit most from voice.
A clinic operator running behavioral health and dermatology under the same roof should not buy one scheduling solution and apply it identically. The metrics that matter, the reminder cadences, and the waitlist policies are different by line of service.
Five vendor-evaluation questions for AI scheduling
If you are evaluating an AI voice scheduling vendor, these are the questions that separate real platforms from reminder-bot wrappers.
- Does the agent reschedule on the call, or does it bounce the patient to a portal? Reschedule-on-call is table stakes for ROI. Bouncing to a portal recovers maybe a third of what real-time rebook does.
- How does it handle waitlist filling, and does it integrate with our EHR's waitlist field? Epic, Athena, eClinicalWorks, NextGen each store waitlists differently. A vendor that says "we can do waitlist" without naming your EHR is bluffing.
- Does it run real-time eligibility checks at booking? If the answer is "we can flag inactive insurance based on your data," that is not real-time. Real-time means a 270/271 transaction at the moment of booking.
- What is the recovered-slot rate on cancellations, not the contact rate on reminders? Vendors love to quote contact rates because they are high. The number that drives revenue is what percentage of cancelled slots get refilled inside the same business day.
- How does it handle BAA, PHI in call transcripts, and recording-consent state laws? Healthcare adds compliance overhead that consumer voice agents do not handle. Get the BAA in writing before pilot.
The contrarian closer
Most clinics will get more lift from fixing their reminder cadence than from buying an AI scheduling platform. A 3-touch reminder program built on the EHR you already pay for catches the easy 30-40% reduction without any new vendor. AI voice scheduling earns its keep on the harder problems: reschedule-on-call, waitlist filling, and the "make-it-easy-to-cancel" behavioral lift in specialties where shame drives no-shows. If your clinic is still on a single 24-hour text and you are evaluating AI voice as the first move, you are paying for sophistication you have not earned yet. Fix the reminder program first. Then buy the platform that solves the residual problem, not the average problem.
If you want to compare notes on what scheduling looks like in your specialty, see why most AI voice agents fail in 30 days, the 5-minute rule and after-hours leads, HIPAA and BAA requirements for AI voice, and patient intake compliance.
Ready to look at your numbers? Book a call, start a free trial, or check the pricing page.