How AI Voice Agents Cut Missed Calls at Australian Pharmacies

Most pharmacy owners have a rough sense that the phone gets away from them. Very few know the number. The hardest part is that missed calls leave no trace: a dispensed script is recorded, a call that rang out is not.

We know the scale of that gap because we measured it. In a 30 day audit of one independent Australian pharmacy, 3,988 calls came in and 74.9 percent went unanswered.

This guide covers what those missed calls actually are, how AI voice agents work, and how they compare to voicemail and answering services. It also covers what to look for in one, what the first month realistically looks like, and when an AI voice agent is the wrong answer.

How AI voice agents cut missed calls, in short

An AI voice agent answers your phone line automatically and holds a natural spoken conversation. It resolves routine calls using your live pharmacy data, and escalates clinical or complex calls to a person.

For pharmacies whose phone is under pressure during trading hours, that turns a call that would have rung out into a completed task. In our beta, 57 percent of the calls the agent handled produced an actionable ticket for staff.

It is not a fit for every pharmacy. If you already answer most of your calls, it will not pay for itself. The section below on when not to buy one is there for a reason.

Why pharmacy phones are under so much pressure

Community pharmacy is one of the highest traffic corners of Australian healthcare. There are more than 6,000 community pharmacies in the country, and Australians visit them more than 443 million times a year.

That foot traffic sits on top of a dispensing workload of 335.1 million PBS and RPBS prescriptions in 2024-25. Every one of those scripts can generate a phone call: a status check, a refill, a question about timing or interactions.

The mix is also changing in a way that adds phone traffic rather than script traffic. The Pharmacy Guild reports that prescription volumes remain flat after the shift to 60 day dispensing, while consultation services grow quickly. More than 80,500 UTI consultations were delivered in 2025, and pharmacists now administer 29 percent of all flu vaccines. Consultations get booked, rescheduled and confirmed, and most of that happens by phone.

So the phone is not a side channel. It is a primary one, and it rings hardest at exactly the moments staff are busiest.

What we found when we actually counted

Most writing about pharmacy missed calls recycles the same handful of borrowed statistics about small business phone answering, none of which come from pharmacies or from Australia. We got tired of citing them, so we ran our own audit.

Across 30 days at one independent Australian community pharmacy:

Measure	Result
Total inbound calls	3,988
Answered	1,001 (25.1%)
Unanswered	2,987 (74.9%)
Unanswered calls per day	99.6
Average wait before a caller gave up	19.2 seconds
Calls arriving during opening hours	99.1%
Busiest hours	11am (702 calls), 1pm (699 calls)

Two findings changed how we think about this product category.

First, this is not an after hours problem. 99.1 percent of calls arrived while the pharmacy was open. After hours coverage, which is how most of this category is sold, addresses under 1 percent of the missed volume. The other 99 percent is a capacity problem during trading hours.

Second, answer rate gets worse as volume rises, not better. Of the weekdays measured, the busiest had the worst answer rate, at 19.9 percent against 213.8 calls a day. Capacity does not stretch under load, it breaks.

The full dataset, methodology and limitations are in the Australian Pharmacy Phone Report. It is one site, and we say so plainly there.

The hidden cost: the phone competes with the bench

There is a second reason to care about this that has nothing to do with lost sales.

A 2025 Monash led scoping review of 51 studies found pharmacy staff are interrupted as often as every two to six minutes during dispensing. Phone calls are listed first among the most common external interruptions.

That puts a pharmacy in a bind with no good side. Answer the phone and you interrupt a person doing accuracy critical work. Leave it and three quarters of callers do not get through.

An AI voice agent is one of the few options that improves both sides at once, because the call gets answered without a human leaving the bench. The full evidence on interruptions and dispensing accuracy is in our guide to reducing phone interruptions for pharmacy staff.

How AI voice agents work

An AI voice agent is software that answers the phone and holds a natural spoken conversation. It listens, understands the caller’s intent, looks up the information it needs, and responds, all in real time.

Under the hood, three things happen in sequence. The agent converts speech to text. It decides what the caller wants and what to say back. It converts its reply to speech. Modern agents do this fast enough to feel like a normal call rather than a menu.

That speed matters more than it sounds. In our audit the average abandoned caller gave up after 19.2 seconds, and only 16.1 percent of abandons happened inside the first 10 seconds. Most callers waited through several rings first. An agent that picks up in under five seconds is answering while 83.9 percent of those callers were still on the line.

The important detail for pharmacies is integration. An agent is only as useful as the systems it can see. One that cannot check dispense data or opening hours can only guess.

What the calls actually are

This is the part most vendor material skips, usually because they do not have the data.

Across our beta, 57 percent of answered calls produced an actionable ticket. That beta was small and deliberately limited: 195 calls across 46.8 hours of active operation, which is a fraction of the volume in the baseline audit. Here is what those 112 tickets were:

Request type	Tickets	Share
Delivery requests	45	40.2%
Pickup confirmations	31	27.7%
Prescription inquiries	27	24.1%
Other	9	8.0%

n = 112 tickets from 195 answered calls, one Australian pharmacy, 2025.

Almost none of that is clinical. It is logistics: where is my medication, can you bring it to me, is it ready yet. That delivery share was large enough that we wrote it up separately, in what call data says about pharmacy delivery demand. That is precisely the category of work that does not need a pharmacist, and precisely the category that gets lost when nobody can reach the phone.

It also reframes what the agent is doing. It is not deflecting calls away from staff. It is converting calls into structured work that staff can action later, on their own schedule, instead of being interrupted mid dispense.

Staff completed 84.8 percent of the tickets generated. We report that number because it measures whether the handoff actually works, which matters more than whether the AI sounds convincing. An agent that generates work nobody completes has just moved the problem.

AI voice agent versus the alternatives

Most pharmacies already have some way of handling overflow calls. Here is how the common options compare.

Capability	Voicemail	Answering service	AI voice agent
Answers every call	No	Sometimes	Yes
Available 24/7	Yes, but silent	Limited hours, higher cost	Yes
Sees live pharmacy data	No	No	Yes, when integrated
Handles routine queries end to end	No	Partially	Yes
Escalates complex calls to staff	No	Yes	Yes
Cost to scale	Low but ineffective	High, per call or per seat	Flat and predictable
Works during the in hours rush	No	Partially	Yes

Voicemail is cheap but loses the many callers who will not leave a message. Answering services put a human on the line but cannot see a caller’s script status or your live stock, so they take a message and hand you back the work.

An integrated AI voice agent is the only option that answers the call and resolves it with real information. We compare the human option in detail in AI receptionist versus answering service, and look at the wider market in our roundup of the best AI voice agents in Australia.

What to look for in a pharmacy AI voice agent

Not every AI voice agent suits a pharmacy. A regulated, trust sensitive setting raises the bar.

Direct integration with your dispense software

This is the single biggest differentiator, and the easiest to check.

Ask what the agent can actually read at the moment a caller asks. Can it see whether a script is dispensed and waiting? Can it see today’s real opening hours, including public holidays? Can it see whether an item is in stock? Can it write back, so a delivery request becomes a record your team sees rather than an email someone has to re transcribe?

An agent without that access is reading from a script and guessing, which is exactly the experience callers dislike about automated systems. If a vendor cannot name your dispense system, they are not integrated with it.

Sensible escalation design

Routine questions should be handled end to end, but clinical or complex calls need a clear path to a person.

Good escalation is specific. It should trigger on clinical intent, on caller frustration, on repeated misunderstanding, and on explicit requests for a human. During trading hours it should warm transfer to a pharmacist. Outside them it should take a structured message with a stated callback commitment, rather than pretending it can help.

An agent that tries to answer everything will eventually answer something it should not.

Australian hosting and Privacy Act fit

Health related calls carry sensitive information, and pharmacies are health service providers. That means the small business turnover exemption does not apply, because the OAIC lists health service providers as an exception to it. All 13 Australian Privacy Principles apply regardless of pharmacy size.

Ask where call audio and transcripts are stored, how long they are retained, whether they are used to train models, and whether data leaves Australia. Get the answers in writing. We cover the detail in AI, patient data, privacy and compliance for pharmacies.

A natural voice, and a graceful failure mode

Callers decide within seconds whether they trust the voice on the line. A clear, natural voice keeps them on the call instead of hanging up to try again later.

Test the failure mode as hard as the success case. Call it with a strong accent, with background noise, with a mumbled brand name, with a question it cannot possibly answer. What it does when it does not understand matters more than what it does when everything goes smoothly.

A generic, multi industry chatbot bolted onto a phone line fails the first two tests. It moves the problem rather than solving it.

What the first month actually looks like

Based on how our beta ran, rather than on how these things are usually pitched.

Week 1: connect and listen. The agent is connected to the phone line and the dispense system, and configured with your hours, services and escalation rules. Most pharmacies start it in backup mode, where staff get first go at the phone and the agent only picks up after several rings.

Week 2: tune on real calls. You will find gaps. A local suburb name it mishears, a service you offer that nobody told it about, an escalation rule that fires too often or not often enough. This is normal and it is the week that determines whether the deployment works.

Week 3: widen the window. Once escalation behaves, most pharmacies extend the agent into the peak hours that were losing the most calls. On our data that means 11am and 1pm.

Week 4: measure against your baseline. Compare answer rate, abandonment and ticket completion against the audit you ran before you started. If you did not run one, you cannot tell whether this worked, which is why we published the method for running your own.

Expect the honest version of this: the first fortnight involves fixing things, and the value shows up in month two.

When an AI voice agent is the wrong answer

We would rather say this than have you find out after signing something.

If you already answer most of your calls. Run the audit first. If your answer rate is above 80 percent, your phone is not your bottleneck and this will not pay for itself.

If your call volume is genuinely low. A pharmacy taking 20 calls a day can usually solve this with a roster change or a second handset. Fixed cost software is the wrong shape for that problem.

If your dispense system cannot be integrated. Without live data the agent falls back to a scripted responder, which is a worse product at a higher price than voicemail. Confirm integration before you buy, not after.

If your calls are mostly clinical. A compounding pharmacy or one with a heavy consulting load may find most calls escalate anyway, at which point the agent is adding a step rather than removing one.

If nobody will action the tickets. The agent creates structured work. If there is no capacity to complete it, you have converted invisible missed calls into a visible backlog, which is more honest but not obviously better.

If you have not told your regulars. Elderly and long standing customers react badly to being surprised by an automated voice. Pharmacies that announce the change, explain why, and keep a clear path to a human get a very different reception from those that just switch it on.

How Emily fits

Krepko built Emily as an AI voice agent for Australian pharmacies specifically, rather than a general purpose receptionist pointed at a pharmacy.

Emily answers calls during trading hours and outside them, so calls do not ring out during the 11am and 1pm peaks that our audit found were the worst. She connects to the pharmacy’s dispense software, phone line, email and SMS, which means she answers with current information rather than generic responses. She runs on a private, Australia based server. Clinical and complex calls go to your team rather than being attempted.

She can run in backup mode behind your staff, in busy mode during peak periods, or across all calls, and you can turn her off entirely. Most pharmacies start in backup mode and widen from there.

We are candid about the limits. Our published results come from one site and a limited beta window, which is why we publish the methodology alongside the numbers. If missed calls are quietly costing you scripts and regulars, that is the gap Emily is built to close.

Start with the call data before you talk to any vendor, including us. If you want to talk after that, we are at krepko.com.au.