A note on what this page is, before anything else: it is not a client case study. DoubleDown AI has no published dental client, so we have no results to report and will not invent any. What follows is an analysis of the problem a Sandton dental practice actually has, the arithmetic for deciding whether automation is worth it, and an honest account of what we can and cannot demonstrate.
Why publish this at all if there is no case study?
Because the underlying question — "should a busy Sandton practice automate its front desk?" — is a real one, and the internet's answer to it is almost entirely marketing. Most pages of this shape carry precise percentages attributed to research houses, and when you go looking for the primary source it is not there. This page previously did the same, quoting figures attributed to a healthcare consumer survey, a healthcare industry report and a dental journal. We could not verify them to a primary source, so they are gone.
What is left is the part we can actually stand behind: the shape of the problem, and how to price it.
What does a Sandton practice's front desk actually deal with?
Sandton is a specific case, and the specifics matter more than any national average:
- Corporate patients on corporate hours. People working in the CBD book around meetings — early, late, or from a desk during the day when phoning is awkward and messaging is not.
- High medical-aid density, which means a large share of first contacts are really benefit questions: which schemes you are contracted to, what a code costs, whether you bill directly.
- Substantial competition within a few kilometres. A patient who cannot get through has several alternatives inside a ten-minute drive, which makes reply speed a genuine differentiator rather than a nicety.
- A front desk that is also doing everything else — greeting, billing, stock, authorisations — while the phone rings.
How would a practice work out whether this is worth it?
Four numbers, all from your own practice management system, none from a blog:
| Number | Where it comes from |
|---|---|
| Missed or unreturned calls last month | Your phone system's log |
| Enquiries received outside 08:00–17:00 | Email and message timestamps |
| Unfilled no-show slots last month | The diary. Count the gaps, not the cancellations |
| Average value of a first appointment | Your own billing |
The first two, times the fourth, times an honest conversion rate, is the ceiling on what a chatbot could recover. The third times the fourth is what better reminders could recover. Compare both against R299 a month. If neither clears it, do not automate anything — and note that a practice with a waiting list genuinely does not have this problem.
What would it actually handle?
The routine and nothing else. Hours, location, parking, which medical aids you are contracted to, what a first visit involves, and booking a check-up or a clean. Anything clinical — symptoms, whether a tooth needs a root canal, what a specific patient's treatment will cost — is refused and escalated, as a hard configured rule rather than a preference.
Somebody in pain at 21:00 is not a lead. They are told what to do and put in front of a person.
What does POPIA require of a dental practice?
Health information is special personal information under the Act and carries a higher bar than ordinary business data. In practice: collect the minimum at first contact — a name, a number, a preferred time, never a symptom description — keep it in a store belonging to the practice rather than pooled, tell the patient plainly what you are collecting, and be able to delete it on request. Our piece on AI and POPIA goes through the Act.
What can we actually demonstrate?
The system, not the outcome. There is a free trial on the Website Chatbot, which means a practice can put it on its own site, ask it its own awkward questions, and watch how it handles a patient in pain — before paying anything. That is a weaker claim than a case study with a number in it, and it is the true one. When we have a dental client willing to have their results published, this page will say so and name them.
We could publish a case study with a convincing number in it this afternoon. What we cannot do is publish one that is true, and the difference is the whole point of the page.
What is the shortest version?
Count your own missed calls, after-hours enquiries and unfilled no-show slots. Multiply by what a first appointment is worth. If the answer comfortably exceeds R299 a month, trial it on your own site and judge it on your own patients' questions. If it does not, keep your money.