AI in Optometry (2026): What It Actually Does Today — Scribes, Receptionists, Coding and Claims, With Honest Costs
Quick answer. In 2026, AI in an optometry practice does five concrete jobs: it answers the phone and books appointments, sends reminders and intake forms, verifies vision and medical benefits before the visit, drafts exam documentation with ICD-10 and CPT suggestions for the doctor to approve, and runs recalls and contact-lens refills. It does not diagnose, and it does not bill without a human reviewing the claim.
Last reviewed: August 29, 2026 · Jelo Editorial Team
Jelo costs $300/month per location or $3,000/year per location, with unlimited doctors and staff at each paid location. Two locations cost $600/month or $6,000/year. Annual billing is paid upfront. Optional billing and revenue cycle management (RCM) with Jelo AI is live at 10% of insurance reimbursements received, separate from the platform subscription. AI Receptionist and AI Scribe remain in development and are separately priced. Review pricing and applicable fees, or review ai availability.
Five jobs AI does in an optometry practice today
AI in optometry today is operational, not clinical. The five jobs it reliably performs are phone reception, scheduling and reminders, insurance benefits verification, exam documentation and coding, and patient recall. Each replaces repetitive front-desk or charting work, and each requires the practice's EHR to be the system of record for the AI to be useful rather than a message-taker.
1. Phones and reception
2. Scheduling and reminders
AI scheduling in optometry means the agent knows provider availability, exam-length rules, and appointment types, then confirms by text, sends the intake form, and reminds the patient before the visit. The measurable result is fewer no-shows and fewer empty slots; the no-show guide walks through the reminder cadence that works for eye exams specifically.
3. Benefits verification
4. Exam documentation and coding
5. Recalls and contact-lens refills
AI recall contacts patients before their vision benefits expire, offers the contact-lens reorder before the box runs out, and books the annual exam. Because a third of eye-exam patients do not come back on their own, recall is the highest-leverage revenue job in the practice; the recall playbook covers the sequence.
AI scribe comparison table
When comparing AI tools, verify the current product, data inputs, review controls, availability and commercial terms with each vendor. Jelo’s AI Biller is live at 10% of insurance reimbursements received as a separate RCM add-on. AI Receptionist and AI Scribe remain in development; review the scribe roadmap.
Vendor features and pricing change often; verify with each vendor. The practical question for an OD is not which scribe writes the prettiest paragraph, it is whether the note lands in your exam template with the codes attached, or whether someone still has to paste it in and code it afterward.
AI receptionist comparison: generic vs. EHR-native
An AI medical receptionist that "books into the schedule" needs four things a calendar link does not provide: live access to the provider schedule and appointment-type rules, the ability to capture and verify insurance on the call, a lookup into the patient's chart for reorders and recall status, and a defined escalation path to a human. Generic receptionists (Goodcall, Smith.ai) and phone platforms (Weave) deliver call answering and message capture well; they cannot see the chart or the plan.
When comparing AI tools, verify the current product, data inputs, review controls, availability and commercial terms with each vendor. Jelo’s AI Biller is live at 10% of insurance reimbursements received as a separate RCM add-on. AI Receptionist and AI Scribe remain in development; review the scribe roadmap.
If your EHR is one of the legacy systems and you are not switching, a generic receptionist is still better than voicemail. If you are choosing an EHR, choose one where the receptionist is native. The full breakdown, including what it escalates and a live-call demo, is on the AI receptionist page.
What it costs vs. a front-desk hire
A worked example for a one-doctor practice, with illustrative assumptions you should replace with your own call log:
| Assumption | Value |
|---|---|
| Inbound calls per day | 40 |
| Calls that go unanswered (busy desk, lunch, after hours) | 20% → 8 calls |
| Missed callers who would have booked | 30% → 2.4 visits |
| Average exam + materials value | $180 |
| Recovered per day | ≈ $430 |
| Recovered per month (21 working days) | ≈ $9,000 |
| AI receptionist add-on per month (from) | $290 |
Cut every assumption in half and the receptionist still returns more than fifteen times its cost. The ROI calculator on the receptionist page lets you run your own numbers. For the exam side, the value of an AI scribe is time rather than recovered revenue: a doctor who spends 6 to 10 minutes per encounter on the assessment, plan, and codes gets most of that back, which across 15 to 20 exams a day is one to two more patients or an earlier end to the day.
HIPAA, BAAs, and where the data goes
AI in optometry is HIPAA compliant when the vendor signs a Business Associate Agreement, encrypts protected health information in transit and at rest, keeps role-based access and audit logs, and does not use PHI to train external models. Any tool that hears a phone call, reads a chart, or drafts a note touches PHI, so a BAA is not optional. Ask each vendor four questions before you sign.
- Will you sign a BAA? If the answer is anything other than yes, stop. A consumer chatbot subscription is not a covered arrangement.
- Is PHI used to train your models or a third party's? The right answer is no, in writing. Confirm the applicable data-use terms with each vendor in writing.
- Who can see the transcript or draft, and is every AI call and every doctor decision logged? Ask the vendor to demonstrate logging, access controls and clinician review in the specific product available to your practice. Optional billing and revenue cycle management (RCM) with Jelo AI is live at 10% of insurance reimbursements received, separate from the platform subscription. AI Receptionist and AI Scribe remain in development and are separately priced.
- Where is the data stored and for how long? Ask for the region, the retention policy, and how a patient's data is deleted on request.
Call recording adds a state-law layer on top of HIPAA: several states require all-party consent, so the greeting should disclose recording where required and the practice should confirm the rule with its state board or counsel. Jelo's HIPAA checklist for optical shops covers the rest of the practice, including texting and POS.
A note on optometry telehealth
Optometry telehealth in 2026 is mostly follow-up: post-op co-management check-ins, red-eye triage, dry-eye and medication reviews, and contact-lens compliance visits that do not require a slit lamp. AI's role here is the same as in the office: the receptionist books the video slot, the scribe drafts the note and the code from what the doctor documents, and the billing engine routes it to the right payer with the right modifier. Refraction and dilated exams still happen in the chair.
How to evaluate optometry AI in a 30-minute demo
The fastest way to separate an optometry-ready AI tool from a general-medicine product with an eye-care landing page is to run the demo on your own workflow instead of the vendor's script. Ask for these six things, in this order, and watch what the software actually does rather than what the slide says.
- Call the number. For an AI receptionist, ask the vendor to ring your practice line, or to let you call theirs. Book a new-patient exam, then reschedule it, then ask "do you take EyeMed?" If the agent takes a message for any of the three, it is not booking into a schedule.
- Open the booking in the EHR. The appointment should appear in the real schedule with the right provider, the right exam length, and the insurance captured. If someone has to re-key it, you have bought a message service.
- Enter a real exam. For a scribe, document a diabetic annual with a dilated fundus exam and a refraction. The draft should propose the ICD-10 for the diabetes and any retinopathy, put the refraction on its own line as 92015, and suggest 92014 rather than a generic E/M, with a citation to the finding that supports it.
- Reject a suggestion. The tool should let you decline one code without wiping the rest, and the rejection should show up in an audit log.
- Ask for the BAA and the training policy in writing. Before the demo ends, not after the contract.
- Ask what happens when it fails. A missed call should become a callback task; an uncertain code should be flagged, not silently filed. The right answer to "what does it do when it isn't sure?" is "it asks a person."
Book a 40-minute review of the live Jelo platform and AI Biller RCM workflow. Ask separately about the roadmap for AI Receptionist and AI Scribe. Use synthetic examples for the demonstration; do not bring identifiable patient records.
What AI does not do yet
The honest list matters more than the feature list, because it is what an OD is really asking when they type "optometry ai" into an assistant.
- It does not diagnose. A scribe drafts an assessment from findings the doctor entered; it does not read the OCT for you or decide the patient has glaucoma. The doctor reviews and signs everything.
- Separate clinician review of the record from the work performed by a billing service. Code suggestions are reviewed before they hit the claim. At Jelo, AI Biller is a live RCM add-on at 10% of insurance reimbursements received, separate from the platform subscription. Jelo AI handles claim preparation, submission, denial work, appeals and payment follow-up as an optional end-to-end billing service. The platform includes core claims, ERA and denial-worklist tools.
- It does not replace the technician. Pretesting, imaging, and the patient in the chair are still human work. AI removes typing and phone tag, not clinical staff.
- It does not fix a bad schedule or a bad fee schedule. If the practice under-codes 92014 as 92012, or does not know what VSP actually pays for a refraction, AI will make the same mistake faster. Start with the eye-code vs. E/M decision and VSP vs. EyeMed reimbursement.
Frequently asked questions
What EMR do optometrists use?
Is AI in optometry HIPAA compliant?
Can AI answer phone calls for an optometry office?
What is an AI medical scribe?
Does AI replace the optometric technician?
No. The technician still pretests, images, and prepares the patient. AI removes the typing, the coding lookups, and the phone interruptions around that work.