AI for Optometrists Reads the Retina. The Prescription Walks Out

Mrs Patel's eye test at Priya's practice went well. New prescription, a small change in her left eye, nothing to worry about at the back of it. She spent ten minutes at the frame wall, tried on four pairs, and asked for a quote on the second ones.

"I'll have a think," she said, and folded the printed prescription into her handbag.

Nobody called her. Nobody was ever going to. Priya only found out six weeks later, when Mrs Patel came back to have a pair of online glasses adjusted. They were made from the prescription Priya's optometrist wrote.

Priya is a composite of the owners I talk to: one independent store in a regional Victorian town, two optometrists, three dispensers, a front desk that also does the ordering. That month a vendor showed her what AI for optometrists usually means: software that reads retinal photos and flags early signs of disease.

It would not have called Mrs Patel.

The eye test is the part of your practice that already works. The money leaks in the ten minutes after it.

Where AI for optometrists is being pointed

Retinal screening, OCT analysis and AI note taking are the pitch right now, and I will be fair to them. A clinical tool that helps your optometrists catch something earlier deserves a serious look. Whether it is good enough is their call, not mine.

But clinical AI makes the test better, and the test is rarely where an independent practice loses money. If you bulk bill your eye tests, you already know the test is not where the margin lives. The margin lives at the dispensing desk, and that is exactly where the patient slips away.

Why the prescription walks out

A sale in optometry is a relay, and the eye test is only the first leg:

booking → test → prescription → frame wall → quote → follow up → order → fit and collect → recall

The leg in bold is the one nobody runs. The optometrist moves to the next patient. The dispenser is with someone else. The quote lives on a printed sheet in a handbag and a half note in your practice software.

Meanwhile the chain store at the shopping centre and a dozen online retailers are one search away, and the prescription works at all of them.

At Priya's, nobody owns the "thinking about it" pile. The dispensers are measured on what they sell on the day. The front desk is measured on answering the phone. A patient who left without ordering is nobody's job, so the follow up happens when someone has a quiet afternoon, which is never.

The same happens at the other end. A tray of finished jobs sits uncollected at the back, each one a patient who has not had a fit check or been asked about a second pair.

A patient who says "I'll think about it" is not a lost sale. They are an unfinished one.

Run the numbers on your own walkouts

Treat this as arithmetic you fill in yourself.

Say your optometrists see three hundred patients a month. Say a third of them leave with a changed prescription and no order. That is a hundred walkouts. Say a quarter of those would have ordered if someone had called within a week with their quote and a question. That is twenty-five orders. Say your average spectacle job is worth four hundred dollars. That is ten thousand dollars a month walking out in handbags.

Every number there is invented and yours to replace. Pull last month's test list, mark everyone with a new prescription, and check who ordered from you. If your number is small, good: you have just saved yourself a project.

What AI for optometrists should do first

Fair question: why AI, and not a reminder in Optomate, Sunix or whatever you run? A reminder tells someone to call. It does not tell them who matters most, what the patient tried on, or what to say. A model can read the notes, the quote and the history in seconds and draft the next move. A reminder rule cannot.

So the system watches every test from the chair to collection, and keeps one status per patient:

• It spots a new prescription with no order attached after a few days.
• It reads the dispenser's notes and the quote, and drafts a short, personal follow up for a person to send.
• It flags finished jobs that have sat uncollected for longer than you choose.
• It catches patients whose recall date has passed with no booking.

Every morning the front desk gets one list, warmest first:

• Patel, A: new prescription, quoted on two frames, no order after five days
• Nguyen, T: progressive lenses ready, uncollected for twelve days
• Brooks, M: recall overdue, last test two and a half years ago

Nothing goes out on its own. A dispenser reads the draft, changes what needs changing, and decides whether to call or text.

Measure two things each month: the share of new prescriptions that become an order with you, and the jobs still waiting on collection.

Automate the chasing, never the clinical call

No system should tell a patient anything about their eyes. Results, referrals and clinical advice belong to your optometrists, and their registration is on every one of them. The follow up is about frames, lenses and appointments, never a diagnosis. I build the AI. You keep the business, and the clinical judgement stays in the test room.

The software tells you which patient left with a prescription. Your dispenser is the one who makes them glad you called.

What AI for optometrists costs

Work down this list and stop at the first rung that works.

• The features already in your practice software. Recall reminders, uncollected job reports and quote tracking are often there, switched off or ignored. Turn them on first.
• A simple routine. One person, twenty minutes a day, working the walkout list by hand. If that recovers orders, you have proved the value before spending a cent.
• A custom build that reads your test notes, quotes and orders, ranks the follow ups and drafts the messages. A few thousand up to the low teens of thousands of dollars, depending on how cleanly your software lets data in and out, plus a modest monthly running cost.

Against the worked example above, even the top of that range is a month or two of recovered orders. Against your real numbers, it might not be worth it, and that is a fine answer too.

I broke down where the running costs hide in what AI actually costs a small business. If you also run allied health services, the same pattern of booked care that never gets finished shows up in AI for physios.

The Australian layer

Three cautions. First, this is health information. Prescriptions, test notes and recall dates are sensitive under the Privacy Act. Ask any vendor whether the data is stored in Australia, whether it trains someone else's model, and whether you can delete it on request. Get the answers in writing.

Second, a follow up text about frames is a commercial message. Under the Spam Act it needs consent, a clear sender and a way to opt out. Check how your practice records consent before anything is sent.

Third, patients can buy glasses wherever they like. The follow up is a service, not pressure. "Your quote is ready if you want it, and Sam is in on Thursday" earns trust. A nagging message loses the patient for good.

Common questions about AI for optometrists

What should an independent optometry practice automate first?

The walkout follow up. It runs on patients you have already tested, and one number tells you inside a quarter whether it worked: the share of new prescriptions that become orders with you.

Will it replace my dispensers?

No. Knowing which frames suit a face and how to talk someone through progressive lenses is the job. Digging through last week's notes to work out who to call is not.

Go back to Mrs Patel, sitting in your chair while a dispenser adjusts glasses you did not sell. The test was good. She liked the second frames. Nobody asked her again.

If you want to know how many of your prescriptions are walking out right now, that is what a first call is for. Book a strategy call and bring last month's test list and your orders. We will find the walkouts before we talk about building anything.