AI patient recall systems: how practices are recovering lost revenue from missed recalls
Recall is the most reliable recurring revenue a medical practice has — and the least consistently pursued. Most practices run recall once a month at best, and skip 30-50% of due patients. The math says practices are leaving six figures on the table annually. Here is how AI recall systems change that.
Why recall is the practice growth lever no one runs well
Patient recall — calling patients due for annual exams, six-month follow-ups, chronic disease monitoring — is the most predictable revenue line in any medical practice. The patient exists. The need exists. The visit is medically indicated. The reimbursement is already negotiated. All the practice has to do is reach them.
In practice, recall is almost always under-performed:
- Optometry: 60-70% of revenue depends on annual exam recall. Most practices skip 30-40% of due patients.
- Dermatology: Annual skin-screening recall is 23% of practice revenue. Most practices run it manually once a month.
- OB/GYN: Annual wellness exam recall drives the relationship and the screening reimbursement.
- Primary care: Diabetic, hypertensive, and hyperlipidemic patients need quarterly follow-up. Most practices send postcards.
- Dental: 6-month recall is THE recurring revenue line. Most practices run it inconsistently.
The reason it does not get run: it is repetitive, slow, low-status work that front-desk staff get pulled off of to handle inbound calls. Recall is "we will get to it" — and a year later, the patient has gone to the practice that did reach them.
The math of skipped recall
Run the numbers on a 4-provider primary care practice with 4,000 active patients:
| Recall category | Patients due/year | Avg revenue/visit | Annual revenue if 100% recalled |
|---|---|---|---|
| Annual wellness exam | 4,000 | $185 | $740,000 |
| Diabetes follow-up (quarterly) | 600 × 4 = 2,400 visits | $140 | $336,000 |
| Hypertension follow-up (semi-annual) | 900 × 2 = 1,800 visits | $110 | $198,000 |
| Hyperlipidemia follow-up (semi-annual) | 700 × 2 = 1,400 visits | $110 | $154,000 |
| Medicare wellness (annual) | 900 | $170 | $153,000 |
Total possible recall revenue: ~$1,581,000. At industry-typical 60% recall hit rate (manual outreach, skipped due to inbound load), the practice captures roughly $948,000. The skipped 40%: $633,000 of margin walking out the door.
How AI recall systems work in practice
1. Pull the recall queue weekly from the EHR
AI pulls a list of patients due for their next visit per your recall protocol (12-month, 6-month, 3-month, custom). Patient demographics, last visit date, chronic conditions, prior outreach history all feed the prioritization.
2. Multilingual outbound calls + SMS
The AI calls patients during appropriate hours (TCPA-compliant — 8am-9pm local) in their preferred language. Voice or SMS or both, configurable per patient.
3. Books the visit on the call
The patient does not have to call back. The AI checks provider availability, books the slot, confirms by SMS. The conversion rate from "I will call back later" to actual booking is ~12%. The conversion rate from "let me book you right now" is ~62%.
4. Handles the gnarly objections
"I have not been feeling well, I should see the doctor sooner" — books urgent slot. "I am moving to Arizona" — flags chart for transfer. "I do not want to come in" — captures reason, routes to outreach manager, deactivates if appropriate. Each call is handled to a clean conclusion.
5. Reports the queue every week
You get the recall numbers without asking: how many called, how many booked, how many declined, how many no-answer (queued for next week), conversion rate trending.
What good recall protocol looks like
The recall protocol is the difference between mediocre and excellent results:
Tiered outreach
Highest-priority patients (uncontrolled diabetic, hypertensive with last A1C > 9, etc.) get prioritized in the queue. The AI calls them first each week. Lower-priority recall (routine annual wellness) gets called later in the week.
Multilingual native
In US markets with significant Spanish, Mandarin, Vietnamese, or other language populations, recall in the patient preferred language converts ~3x better. Generic English-only recall under-performs on the patient mix that most needs the visit.
Right-channel routing
Some patients answer the phone. Some only respond to SMS. Some require a portal message. The AI learns the channel that works per patient over time, escalating to a call only when SMS does not get response.
Booking flexibility
The biggest conversion lift comes from booking flexibility — multiple slot options, multiple providers, multiple locations. Patients book when convenient. "Next available with Dr. Smith is Thursday at 3pm" converts worse than "would Tuesday morning at 9, Wednesday at 4pm, or Friday at 11 work for you?"
What it does not replace
Recall AI is operational, not clinical. It does not:
- Decide which patients need recall (your clinical protocol does)
- Adjust medications or treatment plans
- Provide medical advice beyond your standardized recall script
- Replace the relationship the provider has with the patient (it scales the access, not the relationship)
The role is to execute the outreach work that your clinical team has already decided needs to happen — consistently, at scale, in every language your patients speak, without burning out your front-desk staff.
See Claire run your recall queue.
30-minute demo. Real recall workflow on your real patient list. Real numbers on what gets recovered.