AI receptionists: 4 of 28 Australian platforms list one, and the no-show claims are still unproven
The AI receptionist is the second AI worker arriving at Australian practices. It is real, it is booking into the major platforms, and its evidence base is the thinnest in the AI stack.
Four of the 28 Australian practice management platforms in our comparison listed at least one AI receptionist when we counted on 30 October 2026: BP Premier, Cliniko, Pracsuite and Shexie. Thirteen listings, roughly 12 distinct products. Cliniko's directory alone carries four, including the Able and BookedSolid receptionists; Shexie lists Facere, which handles phone and email.
Distribution is real
Healthengine sells Helen, an AI receptionist built for general practice: it answers around the clock, books directly in Best Practice or Pracsoft, speaks 30-plus languages, and hands calls to reception with the caller's context on screen. Cliniva appears in Best Practice's own partner network, answering calls and booking into Bp Premier.
The distribution runs through the same integration directories that decided which scribes a practice can buy. The platform a practice already runs is again the gate.
No published study we found isolates a conversational AI receptionist's effect on no-shows. The 60% claims are marketing-tier.
What is measured, and what is marketed
The solid ground predates conversational AI. Penn Medicine's randomised trial across more than 244,000 patients added an automated voice call to standard text reminders: no-shows fell from 11.3% to 9.6%. A Cochrane review of eight randomised trials found text reminders lifted attendance from 67.8% to 78.6%. Automated contact works.
The peer-reviewed literature has not reached the front desk itself. An umbrella review in the International Journal of Medical Informatics (March 2026) covered 44 reviews of conversational agents in healthcare: most are text chatbots, scheduling is understudied, and the one domain where every study reported positive outcomes was addiction support. A JAMIA systematic review found even the effect of predictive overbooking on no-shows uncertain. We found no published study, here or anywhere, that isolates a conversational AI receptionist's effect on no-shows. The 60% vendor claims are marketing-tier.
Healthengine publishes its own results: 63% of routine calls fully handled, 130 minutes of call time saved per day, 4.5 out of 5 patient satisfaction, all from one GP clinic. The same page carries the disclaimer: outcomes vary, and no outcome should be expected for other clinics. Keep that caveat in the demo room.
The fair summary: an AI receptionist measurably answers calls and writes bookings into the practice management system. Its effect on no-shows at your practice is unknown until measured at your practice.
What patients think
Patient attitudes are measured, and they point two ways. A Salesforce consumer survey (June 2026) found patients three times more likely to trust an AI agent embedded in their provider's system than the same technology as a public chatbot, and 89% call an escalate-to-human option essential. A poll of 6,000 people the next month found 89% would rather speak to a real person, 29% would hang up on an automated system, and 47% do not trust AI to book accurately. Weigh the commissioning: the first ran on a software vendor's panel, the second was paid for by a human answering-service company. Both are US surveys; no Australian patient survey was found.
The condition survives the conflicts: patients accept the AI receptionist that is transparent about what it is, embedded in the practice's own systems, and one button away from a human.
Recording calls: the privacy rules
A scribe drafts a note a doctor signs. A receptionist talks to unverified callers on the practice's behalf, and the risks are different.
Confirming to a caller that someone is a patient is a privacy breach, and an AI receptionist cannot verify a caller over the phone. The products handle this by design: Helen handles admin calls only, never gives clinical advice, and transfers anything needing judgment to a human. That admin-only line is the one that matters when you evaluate a product.
Most of these products record and transcribe calls. Recording laws vary by state, and in some jurisdictions recording without consent is a criminal offence. Vendors and insurers converge on one answer: disclose the recording on every call, in the first sentence. The Office of the Australian Information Commissioner treats information entered into an AI tool, and its output, as collected health information, and an overseas vendor triggers APP 8 cross-border obligations.
Overseas, the rules are hardening: California requires verbal AI disclaimers in healthcare patient communications. That is not Australian law, but disclosing the assistant anyway costs nothing.
Four checks before one answers your phone
The disclosure script. Every call opens with the recording and AI-assistant disclosure. Listen to the vendor's demo call and check the first sentence.
The verification line. What exactly will the product confirm about a patient to an unverified caller? The only safe answer is nothing clinical and nothing confirming.
Where the recordings go. Storage location, retention, and whether the vendor is overseas. An overseas vendor means APP 8 obligations for the practice.
The booking depth. Does it write into your practice management system, or does it take messages? A product that only takes messages does less, and often costs the same.
The same rule as the scribe
The receptionist is real, the distribution is in the integration directories, and the evidence for the headline claims is not. Buy on the measured facts: calls answered, bookings written to the record. Discount the marketed miracles. The comparison table records the full integration list for all 28 platforms. Read it before the demo.
For the documentation side of the AI workforce, read AI scribes and AI practice management software, and the practice management software comparison behind them.
If you are weighing an AI receptionist against your platform's integration list, get in touch and we will walk it with you.
About the data
Platform counts come from the Airteam pmsProvider dataset, a human review of each vendor's website, counted across 28 platforms on 30 October 2026. Trial evidence: Penn Medicine (2026) and the Cochrane review of reminder trials. Reviews: the International Journal of Medical Informatics umbrella review (March 2026) and the JAMIA rapid systematic review (2023). Vendor results: Healthengine's published Helen figures, single clinic. Patient surveys: Salesforce (June 2026) and OnePoll (July 2026), both US-based with disclosed commissioning.