A 71-year-old British woman said she tried five times to book a doctor’s appointment after suffering a stroke and could not get past her medical practice’s AI receptionist, according to an Inc. opinion piece by Tomas Gorny. The practice later decommissioned the tool.
Gorny, co-founder and CEO of business-communications company Nextiva, argues that the case should not be treated as proof that AI receptionists are inherently a bad idea. For small businesses, the account raises a more specific question: what happens when an automated customer-facing system cannot understand a person who needs help?
That question applies beyond healthcare. Any small business using an automated phone, chat or booking system needs a clear route for customers who cannot complete a request through the usual path. Without that route, a faster answering system can become a new bottleneck.
The failed booking attempts show the access risk
The patient account described in the BBC reporting and Gorny’s Inc. article involved repeated failed attempts to book an appointment after a stroke. The patient said the AI receptionist could not understand her, and the GP practice later stopped using the system. The available reporting does not establish whether the tool was defective in all circumstances, how often similar failures occurred, or whether the same issue affected other patients.
Even with those limits, the incident illustrates the access problem created when automation becomes the first or only gatekeeper. A customer may not know the exact phrase needed to reach a person. A caller may have an accent, speech impairment, limited digital literacy or an urgent request that does not fit a scripted workflow. In those cases, the system’s failure is not merely a poor user experience. It can block access to the business itself.
The same principle applies to nonmedical businesses. A customer trying to report a delivery problem, cancel a service, dispute a charge or request accommodation may need a human route precisely because their issue does not fit the automated path.
AI can reduce phone bottlenecks, but only with fallback routes
Gorny’s broader argument is that medical practices face real call-intake pressure. He describes phones as a persistent bottleneck, with staff capturing requests, clarifying details, documenting information, routing calls and resolving routine issues by hand. He also cites industry estimates on missed appointments and the cost of inefficient scheduling, though the article does not present an independent performance study showing that AI receptionists reliably solve those problems at scale.
That makes the strongest conclusion narrower than the vendor-friendly framing. AI receptionists may help with repetitive tasks such as routing calls, scheduling, eligibility checks and intake forms. But speed is only useful if the system also identifies when it is failing and hands the customer to a person before the interaction collapses.
The article also links automation to privacy and compliance considerations, including HIPAA-compliant answering services. Because those references come from a vendor and vendor-linked resources, they should be read as industry commentary rather than independent proof that any given AI receptionist meets healthcare privacy obligations.
Removing human reception can have real human consequences
The human impact of replacing reception staff with AI is not limited to job displacement. Front-desk workers often recognize repeat callers, notice distress, interpret incomplete information and know when a request needs special handling. Those judgment calls are especially important for older customers, people recovering from illness, people with speech differences and customers who do not use digital systems comfortably.
When businesses replace that human layer with automation, they may reduce wait times for routine requests while making the service less accessible for people whose needs are less routine. That can shift the burden from the business to the customer, who must keep calling, find another route or abandon the request entirely.
For small businesses, there is also a trust cost. Customers often judge a business by how it responds when something goes wrong. An AI receptionist that works for simple calls but traps vulnerable or frustrated customers in a loop can damage the relationship the tool was meant to protect.
Human review remains essential in customer-facing automation
A safe automation strategy should define what the AI system may do, what it must not do, and when a person must take over. The failure described in the patient account points to several control questions: how many failed attempts trigger escalation, whether a caller can request a human at any time, who reviews unresolved interactions, and how quickly the business can identify repeated failures.
For healthcare providers, those questions sit alongside privacy and accessibility obligations. For other small businesses, they are still relevant because customer-service failures can create reputational, operational and insurance concerns. Businesses weighing exposure from automated service breakdowns may also want to review where standard business insurance policies leave coverage gaps and how human review fits into AI-managed processes.
Practical controls before automating the front line
The source article does not present a tested checklist, but the failure scenario supports several practical controls for small businesses considering AI reception, booking or chat systems:
- Create a clear human fallback. Customers should be able to reach a person through a simple phrase, key press or visible option.
- Set failure thresholds. Repeated failed attempts, abandoned calls or unresolved requests should trigger human review.
- Test edge cases. Include accents, speech impairments, urgent requests, incomplete information and customers unfamiliar with automated systems.
- Track more than answer speed. Measure completed resolutions, failed transfers, abandoned interactions and repeat contacts, not only call volume handled.
- Review privacy and compliance. Healthcare, financial and legal businesses should verify data-handling rules before routing sensitive information through an automated system.
- Assign ownership. A named employee or manager should review difficult interactions and decide whether the tool remains appropriate.
- Document customer complaints. Complaints about access failures should be treated as operational evidence, not isolated frustration.
A note on the source
Gorny’s article is an opinion piece written by the co-founder and CEO of a business-communications company that sells contact-center and AI-answering technology. Its pro-automation claims should be read with that commercial context in mind. The verifiable news hook is the patient account, the BBC reporting and the practice’s decision to decommission the AI tool.
The measured takeaway is not that AI receptionists are always unsafe. It is that an automated front line needs a human escalation path, especially when customers have urgent, sensitive or hard-to-classify requests. Automation may reduce routine workload, but it should not remove the business’s responsibility to make sure people can still reach help.
