AI Receptionist for Optometrists: Eye Test Bookings and Recalls

An AI receptionist can answer calls and messages, book and move eye tests from your diary, send recall reminders, take contact-lens reorders and capture details for staff. It should not give clinical advice. Anything involving symptoms, pain, injury or sudden changes in vision goes straight to a person, using a list your optometrists write.

That split is the whole design. The software handles the predictable calls, which in most practices are the majority: "Can I book an eye test?", "I need to move Thursday", "My lenses have arrived, can I collect?". The unpredictable ones, where a caller might be describing something serious, are routed to staff in seconds. Making the split reliable is mostly work done before launch: an appointment table, a recall rule table and a short trigger list your optometrists have signed.

Follow me on Instagram@sagnikteaches

Which calls an optometry practice can hand over

Start by sorting a week of real calls into three piles. Ask the front desk to tally them on a sheet; it takes a few minutes a day and gives you actual numbers instead of guesses.

Connect on LinkedInSagnik Bhattacharya

Say a three-chair independent practice tallies 140 calls in a week and finds roughly 85 green, 40 amber and 15 red. Those are illustrative numbers, but the shape is typical: a clear majority of routine work, a middle band of calls that need a human but not instantly, and a small tail where the stakes are high. The goal is to take the green pile off the desk, capture the amber pile neatly, and never let the red pile wait.

Subscribe on YouTube@sagnikteaches

Define your appointment types before you touch any software

An AI receptionist is only as good as the booking rules behind it. Vague rules produce wrong bookings, and a wrong booking in an optometry practice means a patient in the wrong chair with the wrong amount of time. Sit down with your optometrists and write a one-page table of every appointment type the AI may offer.

A filled-in appointment table

The durations below are illustrative; use your own. Yours will depend on your equipment, your clinicians and how you run the test.

Two points about this table. First, it includes a "callback only" row on purpose. Giving the AI a safe place to put everything that does not fit is what stops it forcing a square call into a round slot. Second, each row says who the appointment suits, in plain words the AI can ask about. It never asks "do you have a problem with your eyes?" and interprets the answer; it asks the caller to choose, and anything that sounds like a problem lands in the callback row.

Recall intervals: let your optometrists set them, then automate

Recalls are where an AI assistant earns the most, because a missed recall is a missed test and eventually a lapsed patient. The interval itself is a clinical decision. Professional bodies publish general guidance: for example, one professional optometry association's public page suggests adults aged 18 to 64 have an eye exam at least every two years, and at least annually where a patient is at risk or as their optometrist recommends. The same page suggests a yearly exam for school-age children and for adults aged 65 and over. Treat that kind of guidance as a starting point for a conversation with your own optometrists, not a rule for the software.

What the AI needs from you is a simple rule table keyed to the interval the optometrist actually set at the last visit, stored in your practice system.

A recall rule table

The message timings above are illustrative. The principle is that routine recalls run automatically, while anything linked to a clinical follow-up, such as a six-month review or a lens-supply condition, escalates to a human sooner. If an overdue patient has a condition your optometrist wanted watched, a reminder text is not an adequate reminder.

A sample recall exchange

Here is how a recall message and reply might run, as an illustration. Use "[first name]" style personalisation drawn from your system, never anything the AI invents.

Hello [first name], this is the automated assistant at [practice name]. Your next eye test is due in March. I can offer Tuesday 14th at 10:30 or Thursday 16th at 17:00. Reply 1 or 2 to book, or reply CALL and a member of the team will ring you.

If the patient replies "I've been getting headaches and my vision seems blurry", the assistant must not book and must not reassure. It replies that it is passing the message to the team now, and it triggers the red or amber route you defined. The wording of that reply, and the trigger phrases, belong in your written script, not in the AI's discretion.

The hand-offs: contact lenses and urgent symptoms

This section is the one to write most carefully, with your optometrists in the room. A hand-off needs three things: the trigger, the destination and the speed.

Urgent-symptom triggers

Your clinicians write the list. It will probably cover sudden changes in vision, an injury to the eye, a chemical splash, severe pain and new flashes or floaters, plus any wording the caller uses such as "emergency" or "urgent". Keep it short and clear. A long list produces false alarms that train staff to ignore transfers, and a vague list misses real ones. Add every phrase your front desk has heard on real calls.

The AI's behaviour at a trigger should be fixed and scripted: it states that it is connecting the caller to a team member now, transfers the call to a named person or number, and if nobody answers it tells the caller what to do next as your clinicians instruct, for example to contact emergency services or the nearest urgent eye service. Those instructions are your practice's, not the software's. After hours, the script should give the same instruction and log the call for the first person in the next morning.

Contact-lens problems

A national regulator's contact lens guidance lists symptoms that can signal a problem, including discomfort, excess tearing or discharge, unusual sensitivity to light, itching or a gritty feeling, unusual redness, blurred vision, swelling and pain, and warns that irritation can point to a more serious condition. A receptionist, human or AI, is not equipped to judge which is which. The safe rule: any contact-lens wearer who mentions discomfort, redness, pain or a change in vision gets routed to a person quickly, and the AI does not say "that's probably normal".

Routine lens tasks stay with the AI: reorders, collection reminders, booking an aftercare check, and reminding wearers to bring their lenses and case. The aftercare reminders tutorial goes deeper on that side of the workflow.

A hand-off record

Every transfer or callback should create a short record staff can read in seconds. A good template:

Caller name and number:
Existing patient (yes/no, as told by caller):
Reason for call, in the caller's own words:
Which trigger or pile it matched (green / amber / red):
What the assistant said to the caller:
Transferred live (yes/no, time):
Callback promised by (time):

The "caller's own words" line is the important one. Staff need the exact phrase, not an AI summary that may have tidied away the detail that mattered.

Booking mechanics: diaries, double-bookings and timing

Most failures here are not clever AI errors. They are ordinary scheduling bugs. Check these before launch.

• One source of truth. The assistant must read and write the same diary your staff use, in real time. A copy that syncs every few minutes will produce double bookings.
• Room and clinician constraints. If only one optometrist does lens fittings on Tuesdays, the booking rule must know that. Test by asking the assistant for a lens fitting on a Monday and watching it decline.
• Buffers. Add a few minutes between tests if your team needs them, for cleaning and notes. Put the buffer in the appointment type, not in staff memory.
• Cancellations. Decide the notice period and what the assistant says if a patient cancels inside it. Do not let it apply a fee or a penalty; it records the cancellation and flags it to staff.
• Waiting list. If a slot opens, the assistant can offer it to the next name on a list. Make sure patients opt in to this.

For a wider view of what goes wrong when an automated receptionist books badly, and how to fix it, see what to do when an AI receptionist gets a booking wrong.

Insurance, funding and price questions

Callers often ask whether their vision cover pays for the test or for a lens fitting. These answers vary by plan and change, so the assistant should not state them. A safe approach is to have it say that cover depends on the individual plan, capture the insurer or scheme name and the member number if the patient volunteers it, and book the test while noting that the team will confirm cover before the visit. The insurance verification workflow shows how a front desk can check cover step by step.

The same logic applies to prices. The assistant can quote a published test fee if you load it, and nothing else. For lens and frame pricing it books a dispensing appointment or a callback.

Four trial weeks at a two-optometrist practice

Take an independent optometry practice with two optometrists and a part-time receptionist who is often stuck dispensing while the phone rings. The owner decides to trial an AI receptionist for out-of-hours calls and for overflow when the desk is busy. All figures are illustrative.

1. Week 1: the rules. The owner and one optometrist spend an afternoon writing the appointment table, the red-trigger list and the recall rules. They tally a week of calls: 140 calls, of which 20 went unanswered at lunchtime or after 17:30.
2. Week 2: shadow mode. The assistant listens and drafts what it would have said, without speaking to callers. The receptionist compares its drafts with her own replies each evening and notes every difference. Two gaps show up straight away: it did not recognise "I got something in my eye" as a trigger, and it offered a 30-minute slot to someone who wanted a first lens fitting. Both rules are corrected. A shadow-mode pilot exists for exactly this reason.
3. Week 3: out-of-hours only. The assistant answers after 17:30 and during lunch. It books routine tests, takes messages and transfers any trigger call to the on-call number. Of 22 calls, it books 9 routine tests, takes 8 messages for the morning and transfers 1 call.
4. Week 4: review. The owner reads every transcript from the week. One caller had asked for "an eye test and my lenses are bothering me"; the assistant had booked a routine test. The rule is rewritten so that any mention of discomfort overrides the booking and routes to a person.

The outcome is not a miracle: it is fewer unanswered calls, a tidy list of morning callbacks and a small set of rule changes. That is a sensible result. The tutorial on measuring return over 90 days explains which numbers to track, such as missed calls, bookings made by the assistant and bookings it got wrong.

Testing before launch: ten calls you should make yourself

Call the assistant from a personal phone and play awkward callers. Write down what happens each time and fix the script until every call behaves as intended.

1. A plain request for a routine eye test next week.
2. A caller who changes their mind halfway through.
3. A request for a time you do not offer, such as a Sunday.
4. "I need a new contact-lens fitting" from someone who is not sure what it involves.
5. A caller who says their eye is sore after wearing lenses all day.
6. A parent booking for a four-year-old.
7. A caller who mumbles a name and number that must be repeated back.
8. A caller who simply says "I want to speak to someone".
9. A caller who says there has been a sudden change in vision.
10. A caller who asks whether their eye problem is serious.

Pass marks: calls 5, 8, 9 and 10 must reach a person, call 3 must be declined politely with alternatives, and call 7 must repeat details back before booking. Run the set again after every script change, and after any software update.

Privacy, disclosure and call recording

Calls to an optometry practice often include health information, which is why it matters where audio and transcripts are stored and who can read them. Before you connect any AI receptionist, check the vendor's data terms: where recordings and transcripts are kept, how long, whether they are used to train models, and whether the vendor will sign the data-processing or business-associate terms your rules require. Ask your compliance adviser which rules apply to you. The privacy-terms checklist turns that into a list of questions.

Tell callers they are speaking to an automated assistant, and say if the call is recorded. Check local rules on both, because they differ.

Keeping it honest over time

• Read a sample of transcripts weekly for the first two months, then monthly. Focus on calls that ended in a transfer or a callback.
• Track three numbers: calls answered, bookings made without staff touching them, and bookings staff had to correct.
• Log every near miss. A caller who mentioned a symptom and was booked as routine is a near miss even if nothing bad followed. Fix the rule the same day.
• Re-check recall rules whenever an optometrist changes how they set intervals.
• Keep a person reachable on every call. If the assistant cannot reach one, the caller hears what to do next, as your clinicians scripted.

If you would rather work through your appointment rules and hand-offs with someone, that is what my AI implementation consultation covers.

Questions optometry practices ask about an AI receptionist

Can an AI receptionist tell a caller if their symptoms are serious?

No, and it should never try. It can recognise trigger words such as sudden loss of vision or an eye injury and pass the call to a person immediately, but the judgement belongs to your clinicians. Write the symptom list with them, keep it short, and test it with deliberately awkward calls before you go live.

Should the AI choose the type of eye test for a caller?

Only by asking set questions your optometrists wrote, such as whether this is a routine check, a contact lens check or a new problem. It should not interpret symptoms. If the answers do not fit a defined appointment type, it offers a callback from staff rather than guessing.

How long does setup take for a small practice?

Most of the time goes into writing the rules, not into the software. Expect a day or two to define appointment types, durations, recall intervals and hand-off lists, then a few weeks of listening to real calls and tightening the script before you widen its role.

Will patients mind talking to an AI?

Many accept it if it is honest, quick and has an easy route to a person. Say at the start that it is an automated assistant, offer to transfer, and never let a caller get stuck in a loop. Check your local rules on disclosing AI to callers and on recording calls.

Does it replace the front desk?

It takes the repetitive calls and after-hours enquiries so staff can focus on patients in the building. Complex conversations, upset patients, insurance queries and anything clinical stay with people. Measure missed calls and booking accuracy to decide how much more to hand over.

Further reads

• Optometry Referral Letter Template: Draft With AI, Then Check — Draft referral letters with AI, then check them.
• ChatGPT for Optometrists: Recall Letters, Reviews and Front-Desk Replies — Recall letters and review replies drafted with ChatGPT.
• What Is an AI Receptionist and How Does It Handle Bookings? — How an AI receptionist handles bookings, explained plainly.
• AI in an Optometry Practice: A Week-by-Week First-Month Plan — A week-by-week plan for a practice's first month with AI.
• Explaining Lens Options and Eye Test Results With AI Help — Explaining lens options and test results in plain words.
• AI Receptionist or Answering Service for a Booking-Based Business? — Compare an AI receptionist with a human answering service.
• How Independent Opticians Can Use AI for Recalls and Bookings — Recalls and bookings in the wider plan for an independent practice.
• AI Chatbot for Opticians: Appointments, Repairs and Lens Questions — Which optician enquiries a chatbot can handle from an approved fact sheet, which must go to staff, and how to set up the hand-off for eye symptoms.
• GDPR in an Optical Practice: Which AI Tools May Touch Records — A scored 12-point checklist for vetting any AI tool before it touches eye-test records, with four tools worked through and the paperwork behind each.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: a professional optometry association's recommended eye examination frequency tables and a medical-device regulator's contact lens risks page (both checked 4 November 2026); recall intervals and urgent-symptom lists must come from your own optometrists.