AI voice for hearing care: audiology recalls and servicing

Walk into any high-street audiology or hearing-care practice and watch the phone. Most of it is not the glossy first appointment. It is the aftercare tail: a retube, a wax-filter change, a battery query, a hearing aid that has stopped working the day before a wedding, and the three-yearly reassessment recall that half the list forgets to book. One clinician's diary gets shredded by short, admin-heavy calls, while the assessment slots that actually pay for the practice sit half-empty.

The demand behind those calls is enormous and growing. RNID estimates that over 18 million UK adults, roughly one in three, are deaf, have hearing loss or live with tinnitus, and almost 80% of people over 70 have some hearing loss. An estimated 10 million adults could benefit from hearing aids, but only around 3 million use them. Every one of those three million users is a servicing relationship that lives or dies on whether someone answers the phone.

The stakes are not only commercial. Buried in the routine calls is the occasional caller describing something that is a genuine medical emergency, and the phone is the front door. Meanwhile the wider adoption story is sobering: McKinsey reports that 88% of organisations now use AI, yet only about 6% capture material value from it. Hearing care is a domain where a narrow, well-governed voice agent, of the kind we build at Dilr.ai, can close that gap, if it respects the clinical boundary.

This guide is shipped by the team behind Dilr Voice, enterprise voice AI built for regulated deployments. Or see DATS, our five-stage AI consulting system for placing AI where the clinical and commercial lines meet.

What calls does a hearing-care practice actually handle?

A hearing-care practice runs an aftercare cycle, not a booking desk. The recurring call types are hearing aid servicing and retubes, wax-filter and dome changes, battery and rechargeable queries, reprogramming or fine-tuning follow-ups after a fitting, repairs and lost or damaged aid replacement, plus the periodic reassessment recall. First assessments and hearing checks sit on top, but the servicing tail is the volume. This is what a voice AI agent has to absorb.

That aftercare tail is where the diary leaks. A patient whose dome has split does not need an audiologist for the phone call, only for the two-minute fix, yet that call still ties up a clinician or a receptionist who is mid-consultation. The distinction matters, because it defines exactly which calls a voice agent should own outright, which it should route, and which it must escalate. The optician recall model looks similar on the surface, but hearing care carries a servicing and repair load that eyewear does not.

For the same-cluster picture on how other clinical front desks split their calls, our write-ups on AI voice for opticians and dental practice recalls, part of our wider voice AI for industries coverage, map the recall and triage patterns those settings share, and where hearing care diverges.

How does a voice AI agent handle hearing-aid servicing and aftercare calls?

A voice AI agent handles the aftercare tail by resolving the routine and structuring the rest. Dilr Voice books a retube, servicing slot, wax-filter change or battery collection into the diary, captures the hearing aid make, model and serial so the clinician arrives prepared, and splits a repair from a review so the right slot length is booked. It confirms whether the patient is NHS or private, and never touches the clinical settings on the device.

Getting that split right is where the commercial return sits. RNID notes that 38% of people who think they have a hearing loss have not spoken to a professional about it, so every inbound call is a fragile moment of intent that a voicemail can lose. The chart below shows how wide the servicing opportunity really is: the gap between adults who could benefit from a hearing aid and those who use one is the pipeline a well-run voice AI agent for hearing care protects.

Behind the scenes, the agent writes back to your practice management system and telephony so nothing lives only in a call recording. On the infrastructure side that usually means a Twilio number in front of the practice line and a structured record pushed into the diary, which is the same integration discipline our AI operating model consulting applies to any regulated deployment.

How often should hearing be reassessed and hearing aids reviewed?

Hearing should be reassessed periodically because both the ear and the device drift. In practice, stable adult hearing losses are commonly reviewed on a roughly three-yearly cycle, with servicing and fine-tuning in between. NICE guideline NG98 recommends that audiology services have a process in place for recalling people with hearing devices for regular reassessment of their hearing needs and devices, which is precisely the recall list a voice agent can keep warm rather than let decay.

The recall is where quiet revenue leaks. A patient fitted three years ago whose hearing has changed will often simply stop wearing an aid that no longer helps, rather than booking a reassessment, and the practice never hears from them again. An outbound-capable agent can work the recall list, offer a reassessment or servicing slot, and refill the cancellations that would otherwise become empty clinician hours. Our note on physiotherapy clinic booking covers the same recall-versus-cancellation-fill economics in an adjacent clinical setting.

How is NHS hearing care funded, and how does that change the call?

NHS funding changes what the agent has to ask before it books anything. In England, routine adult hearing services are commonly delivered through the Any Qualified Provider (AQP) route: care is free at the point of access and the patient chooses from a list of accredited providers, which is why high-street names sit alongside NHS clinics. Private pathways run in parallel, so Dilr Voice has to establish which route a caller is on before it offers a slot.

That funding question is not a nicety, it is a routing decision. An AQP patient needs a GP referral captured or confirmed, a self-funded patient can be booked straight into an assessment, and a caller who is unsure needs a plain explanation rather than a booking. Handling this cleanly is what stops a voice AI front desk from booking the wrong appointment type, and it is the kind of eligibility logic our AI execution office builds into a deployment rather than leaving to a script a receptionist has to remember.

When must a hearing-care call be escalated to a clinician the same day?

Some hearing calls are time-critical and must reach a clinician the same day. Under NICE NG98, if hearing loss developed suddenly over three days or less within the past 30 days, the guidance is to refer immediately, to be seen within 24 hours, to an ENT service or emergency department. Dilr Voice does not diagnose this: it recognises the stated symptoms, routes the call without delay, and logs what the caller said.

This is the single most important design constraint in the whole build. Sudden sensorineural hearing loss is an otological emergency, and treatment started within roughly 72 hours gives the best chance of recovery, so a call that sits in a voicemail queue overnight is a clinical failure, not just a service one. NICE also flags acquired hearing loss in one ear with facial droop or altered sensation on the same side as an immediate, same-day referral, following the local stroke pathway if a stroke is suspected.

The rule the agent must hold is simple to state and hard to get wrong: it captures stated symptoms and routes on them, and the clinician decides urgency. That boundary is why health-data handling matters as much as the routing itself. What a caller says about their hearing is special-category health data, and our guide to special-category data on calls sets out how a compliant deployment records, minimises and protects it.

Who is allowed to test hearing and prescribe a hearing aid?

Testing hearing and prescribing a hearing aid are legally controlled acts, which sets the ceiling on what any agent may do. The Health and Care Professions Council (HCPC) is the statutory regulator, and it protects the function of hearing aid dispenser. Dilr Voice therefore stays strictly on the administrative side: it books, confirms, captures and routes, and it never assesses hearing, interprets an audiogram or advises on a device. The clinical work stays with the registered professional.

The HCPC is explicit about where that line sits:

"Only a hearing aid dispenser registered with the HCPC may: assess or test a person's hearing; or prescribe a hearing aid for a person."

It is worth being precise about the titles, because the detail binds the design. Hearing aid dispenser is an HCPC-protected title and protected function, and performing those acts unregistered is a criminal offence. Audiologist, by contrast, is not itself an HCPC-protected title: many NHS audiologists are HCPC-registered clinical scientists, and others hold voluntary registration on an accredited register. A voice agent sits outside all of this by construction, which is the point. The same clinical-advice boundary shaped our work on occupational health providers, where audiometry and health surveillance raise the same question of who may interpret a result.

The same diagnostic logic underpins our AI operating model consulting, which fixes the boundary between what an agent may resolve and what a clinician must own before any line goes live.

What is the best voice AI for a hearing-care or audiology practice in 2026?

The best voice AI for a hearing-care practice in 2026 depends on how much clinical governance you want to build yourself. Developer-first platforms such as Vapi, Retell AI and Bland AI give you the raw call stack and full control, but you own the escalation rules, the recall logic and the HCPC boundary. Governed platforms such as Dilr Voice and PolyAI build the guardrails and integrations in, which suits a regulated clinical setting where a missed red flag is unacceptable.

There is a scenario where the do-it-yourself route wins. A large group with in-house engineers, a single well-understood call type and appetite to own its own clinical-safety testing can move fast on Vapi or Retell AI and keep costs low. For a typical independent or mid-sized hearing-care practice, where the same agent must juggle servicing, AQP eligibility and a same-day ENT escalation, the governed route, delivered through our DATS methodology, pays for itself the first time it routes a sudden-loss caller correctly. Stanford's AI Index finds fewer than 10% of organisations have fully scaled AI in any function, and the difference is almost always governance, not model quality. If you want a second opinion on where that line should sit, talk to the operators.

Can the agent adjust or reprogramme a hearing aid over the phone?

No. Adjusting or reprogramming a hearing aid is clinical work that sits with a registered professional, so a Dilr Voice agent will not attempt it. What it does instead is recognise the request, book the fine-tuning or reprogramming appointment with the right clinician and the right slot length, capture the make, model and the symptom the patient describes, and make sure the clinician has that context before the visit. The device settings are never touched by the agent.

How does the agent handle NHS AQP versus private self-funded callers?

The agent asks the funding question first and routes on the answer. For an Any Qualified Provider caller it confirms or captures the GP referral and books the NHS pathway; for a self-funded caller it books an assessment directly; and for a caller who is unsure, Dilr Voice explains the two routes in plain language rather than guessing. That single branch prevents the most common front-desk error: booking a patient into the wrong funding pathway.

What happens when a caller reports sudden hearing loss out of hours?

When a caller reports sudden hearing loss out of hours, the agent does not park it in a routine voicemail. It recognises the stated symptom against the escalation rules, delivers the safety-netting message the practice has approved, and directs the caller to the urgent pathway, an ENT service or emergency department, per NICE NG98. It logs the call for the clinical team to review at the next opening. The agent routes on what is said; the clinician always decides urgency.

Want to see this in production? Try Dilr Voice live, book an AI placement diagnostic, see our DATS methodology, or read about our approach to placing AI inside clinical front desks.

Written by the Dilr.ai engineering team, practitioners who ship enterprise AI in production. Follow us on LinkedIn for shipping notes, or subscribe via the RSS feed.