Phone, video and AI front-desk where caller audio never has to leave the practice.
Clinics, GP surgeries, dental practices and small hospitals deal with phone-system vendors who treat PHI as someone else’s problem. CodeB takes the conservative route by default: self-hosted on a Windows box you control, with an on-premise AI option that keeps voice on the machine.
Why generic cloud-voice products don’t quite fit.
PHI in the call audio
Patients name conditions, medications and symptoms on the phone. Cloud voice products process that audio in their own infrastructure under their own DPA. Some of them are HIPAA-friendly, most aren’t out of the box.
Reception is the bottleneck
Monday morning, all phone lines lit. Patient cancels, the slot doesn’t get rebooked, the practice loses revenue. Recall calls for annual checks never go out at all.
Out-of-hours coverage
Patient calls at 9pm to confirm tomorrow’s 8am appointment. Voicemail. They no-show. The clinic has paid for a duty manager to wake up for nothing.
Multi-site practices
Three clinics, three numbers, one PMS. Calls to clinic B end up at clinic A’s reception because routing is hand-coded in the PBX.
A healthcare-specific deployment posture.
On-premise AI option (recommended for PHI)
On-premise AI Voice Engine or on-prem on-premise speech-to-text + an open-weights model. Caller audio never leaves the Windows host. Latency higher than cloud, accuracy lower, but data-residency is binary — nothing leaves.
Cloud AI per workflow (opt-in, scoped)
Use a cloud AI Voice Engine for non-clinical lines (general FAQ, opening hours, public-information DID). Local backend for any line that touches a patient identifier.
Outbound recall campaigns
Dial the annual-recall list. The AI explains the appointment type, asks if the patient wants to book, transfers to reception or schedules a callback. Signed transcript per call.
Video consultation as a side-effect
Same install gives every practitioner a private browser-meeting URL. No separate Zoom subscription, no per-clinician licence, no SaaS media path. End-to-end DTLS-SRTP.
A three-site GP practice in the UK.
Three sites, 18,000 patients on the list, an Asterisk PBX they’ve owned for years. The bottleneck is Monday-morning reception — calls drop, recall lists slip, patients drift to the local GP at Boots.
- AI on a public-information DID handles “what are your opening hours” / “is the clinic open today” with the local backend. No PHI ever leaves.
- Outbound AI runs the recall list every Tuesday evening. Polite, declines aggressive scripts, transfers to reception when the patient wants to book.
- Browser softphone for each receptionist — rings on their laptop, hangs up properly when they end the call, no per-handset licence.
- Video consultation room per practitioner — private link per booking, joined from the patient’s browser, no app install.
Practitioner time spent on reception drops. No-show rate falls. The PMS and the PBX both stay where they were.
Three reasons clinical procurement teams push back on the cloud options.
BAA / DPA scope
Healthcare-grade DPAs are not the default tier from cloud voice vendors. CodeB is processed entirely on hardware you own — Aloaha Limited is the processor for the software, not for the call data.
Vendor lock-in via patient identifiers
Once a SaaS phone vendor holds two years of patient call recordings, switching is a procurement problem. With CodeB, the data lives on disk on a server you own; export is a folder copy.
No-cloud-media-path is a contractual question
Some sector regulators (NHS DSPT, certain EU competent authorities) push for media that demonstrably stays on national infrastructure. CodeB’s media path is local by construction.
Per-clinician licensing math
SaaS voice and SaaS meetings both scale per seat. CodeB scales per-host. Two doctors or twenty — same install, same licence model.
Three patterns we’ve seen work.
Mini-PC in the server cupboard
Fanless NUC. Windows + IIS, CodeB bridge as a service. PBX integration via SIP trunk.
Repurposed Windows desktop
Quiet, off-hours desktop the practice already owns. Same software, no new hardware.
Private hosted tenant
If self-hosting is too operational, we run it on isolated infrastructure under your domain. Single-tenant, no shared media path.
Phased
Start with one non-clinical DID + AI on local backend. Add browser softphones for reception. Roll out video to clinicians.
Want a healthcare-specific scoping conversation?
Tell us about your PMS, the lines you want AI on, and the residency constraints you have to honour. We’ll tell you which CodeB pieces fit, in what order — and where you should not use the AI at all.
Get in touch →Related: AI-call privacy AI receptionist Outbound AI use-cases Privacy manifesto vs. Zoom / Teams Buy / host