Most SA businesses lose calls outside office hours. The receptionist clocks off at 5pm; the after-hours number rings out; the voicemail box fills up. Customers either leave a message that gets buried, or — more often — they call your competitor. Hiring a 24/7 reception team for a typical SMB call volume costs R30,000 to R60,000 per month. Most owners just decide to live with the loss.
A voice AI receptionist closes that gap. It picks up every call, conducts a natural conversation in English (and increasingly Afrikaans), captures the caller's intent and details, books appointments straight into your calendar, and either routes the call to a human or takes a structured message that arrives on WhatsApp or email within seconds. This page explains what we build, what it costs, and how to know if it fits your business.
What we build
24/7 inbound answering
Every call answered within two rings — including 9pm Sundays and public holidays. No after-hours abandonment.
Natural conversation
Modern speech models trained on SA English (and accented variants). The AI sounds warm and human, not robotic.
Calendar booking
Direct booking into Google Calendar, Microsoft 365, Calendly, or your practice management system. Conflict-aware.
Caller qualification
Asks the qualifying questions you would. Captures budget, urgency, suburb, service interest. Hands you a clean lead.
Smart routing
Urgency keywords ("emergency", "complaint") trigger immediate warm transfer to a configured human number. Context summarised.
Message capture
For non-bookable calls, structured message taking — name, number, reason, urgency — delivered to WhatsApp/email within 30 seconds.
What makes this different from rule-based IVR
Most "AI" phone systems on the SA market are still rule-based IVRs with a text-to-speech voice bolted on. They ask "press 1 for sales, press 2 for support" — which is exactly what you and your customers already hate. Three things make a real voice AI different:
Natural-language understanding. Callers don't pick from menus. They say "Hi, I'm calling about my booking next Tuesday at 3pm — can I move it to Thursday?" and the AI handles it. No menu trees, no DTMF prompts, no "I didn't quite get that — please say sales, support, or other".
Context retention. Halfway through the call the AI knows the caller's name, why they're calling, and what's already been said. Compare to an IVR that resets state every time the caller backs out of a menu.
Real action. The AI books, transfers, sends, schedules. It doesn't just collect input and dump it in a queue. A booked appointment lives in the calendar before the caller has hung up.
Who it's for
Voice AI receptionist delivers clear ROI for SA businesses with these traits:
- Phone-driven customer base. Industries where calls still outnumber web enquiries — healthcare, legal, real estate, automotive, hospitality, professional services, trades.
- 15+ inbound calls per day, or significant after-hours volume. Below this, the payback is slower.
- Repetitive call patterns. Most calls are variations on five to ten common reasons — booking, rescheduling, status check, pricing question, directions, complaint, new enquiry.
- Calendar or system to write to. The AI shines when it can take real action; if there's no system to write to, half the value disappears.
Sectors where this typically applies: medical and dental practices, law firms, estate agencies, vehicle service centres and dealerships, guesthouses and lodges, professional services (accounting, consulting, financial advisors), home services (plumbers, electricians, builders), beauty and wellness clinics, training providers, small B2B suppliers.
What a typical SA deployment looks like
To give a sense of scope, here's a representative deployment for an SA medical practice with three GPs. Figures below are illustrative of a typical project, not a specific named client.
Baseline: around 80 inbound calls per day across the three doctors, with after-hours volume of 15-20 calls per day going to voicemail (mostly script collection, appointment changes, and prescription queries). One full-time receptionist plus an answering service that costs R8,500/month and produces inconsistent message quality.
What we built: voice AI deployed on the practice's main number, answering after-hours and overflow. Connected to their Healthbridge calendar for booking. Integrated with their existing SMS gateway for prescription-request follow-up. Urgent calls (any mention of pain, emergency, chest, breathing) route directly to the on-call doctor's mobile.
Representative results after six weeks:
- After-hours calls answered: 100% (vs. ~35% to voicemail before)
- Bookings made by AI: roughly 60% of after-hours bookable calls
- Prescription queries handled fully without staff: ~80%
- Urgent escalations to on-call doctor: averaging two per week, with full context summary
- Receptionist time recovered during business hours: ~12 hours/week (the AI took overflow calls when the line was busy)
- Answering service replaced — R8,500/month saved directly
The R32,000 build paid back in roughly five months on the answering-service savings alone. The bookings captured after-hours were additional upside that previously would have gone to a competing practice.
What it costs
Implementation. R22,000 to R65,000 depending on integration complexity. A single-number deployment with calendar booking and basic routing starts at R22,000 to R30,000. Multi-number setups, complex routing logic (multiple departments, multi-language, multi-location), or deep CRM/practice-management integrations land in the R40,000 to R65,000 range.
Monthly running cost. R2,500 to R8,000 per month for most SMBs. The cost is mainly conversation time (charged per minute of speech) plus your SIP/telephony provider. A practice doing 60-80 calls/day typically lands at R3,500/month. High-volume operations doing several hundred daily calls scale up but stay well below the cost of a single human receptionist.
No platform subscription. We don't charge a SaaS fee on top of the actual API and telephony costs. If your call volume drops, your bill drops.
POPIA, call recording, and consent
Voice AI captures personal information by design — name, contact details, often health or financial info. POPIA compliance is built in from day one:
- Recording disclosure. Every call opens with a clear notice that the call will be recorded and processed by an AI assistant, with an option to be transferred to a human instead.
- Lawful basis. Recording is explicit-consent-based; the caller agrees by continuing the call, with the disclosure logged.
- ZA-region storage. Recordings and transcripts are stored on infrastructure you (or a vetted processor) control, in ZA region. We document the exact provider in the service agreement.
- Retention. Default retention is 90 days; configurable down to 30 or up to 365. Beyond that, recordings are auto-deleted unless flagged.
- Subject rights. One-click export and deletion of any caller's records on request.
If your current answering service can't tell you where its recordings are stored, you have a POPIA exposure. We name our hosting region in writing.
How to start
A 30-minute call is the fastest way to know whether voice AI fits. We listen to a sample of your real inbound calls (anonymised), look at your current call volume, after-hours patterns, and the systems you'd want the AI to write to. Within a few days you get an honest scoped proposal — or a "this is not the right project for you, here's what we'd do instead".
Email info@faautosolutions.com or use the contact form.