AI Receptionist for DSOs: Central Answering, Per-Office Booking

A single-location practice has one failure mode: nobody picked up. A group has four, and only one of them is answering speed.

• Wrong-office calls. The patient rings the number they found on Google, which is not the office they attend. The desk takes a message and hopes it travels.
• Sibling capacity invisibility. Office A is booked three weeks out while Office B, eleven minutes away, has Thursday open. Nobody on the call can see that.
• Uneven staffing. The busiest site has the same two people on the desk as the quietest one, so overflow concentrates exactly where volume is highest.
• Reporting that stops at the group level. Regional managers see totals, not which office is dropping calls at 8am.

A central human call centre solves the first two and creates a new problem: patients get a stranger who does not know the office. AI answering solves them without that trade-off, because it reads the actual schedule of the actual office rather than working from a script.

Model 1 — Number-per-office

Each location keeps its published number and rolls over to the same AI. The dialled number identifies the office, so the caller never has to state it. This is the cleanest model and the one most groups should start with, because it preserves local search listings exactly as they are.

Model 2 — Single group number

One number for the brand. The AI asks which location the caller wants, or infers it from the patient record when the phone number matches. Suits groups marketing at a regional level rather than per-practice, but it weakens local SEO signals if you retire the per-office numbers.

Model 3 — Hybrid with regional escalation

Per-office numbers into the AI, with human escalation routed to a regional coordinator rather than the local desk. This keeps clinical teams off the phone entirely and is common in larger DSOs with a shared services function.

When a caller wants an appointment this week and their usual office is full, the choice is between offering a sister location and losing the appointment. Groups consistently under-use this, usually because nobody has defined the rules.

Define them explicitly, per location pair, before launch:

1. Which locations may be offered as an alternative to which, and within what travel distance.
2. Whether cross-booking is offered always, or only when the home office has nothing inside a defined window such as seven days.
3. Whether new patients and existing patients are treated differently — many groups allow it freely for new patients only.
4. Which appointment types are portable. Emergency and hygiene usually are; active treatment usually is not.
5. How the home office is notified, so continuity of care and revenue attribution both stay clean.

Attribution disputes kill cross-booking programmes faster than patient resistance does. Agree how the revenue is credited between offices before you switch it on.

Cloud group platforms make this straightforward. Denticon exposes office IDs, so a single AI can read and write against each location independently and report per site. Curve Dental behaves similarly. Groups running server-based Dentrix or Eaglesoft across sites need a bridge per server, which is workable but lengthens rollout.

The practical consequence is sequencing. Start the rollout with your cloud sites, prove the workflow, then take on the server locations with a longer setup window. See the PMS integration guide for how each platform behaves, or the Denticon integration page for group-specific detail.

Reporting should be per office from day one. Group totals hide the exact problem you bought the system to fix — the one location dropping a fifth of its Monday calls.