BrightSmile Network operates nine dental clinics across the city, from small neighborhood offices to a flagship site with orthodontics and implantology. Each location keeps two to three administrators at the front desk, and most of their day goes to the patient standing in front of them: registration, payment, walking someone to the X-ray room. The phone is still how patients book, reschedule, and ask whether a toothache can wait until Monday.
The challenge
Across nine clinics, reception lines were busy at peak hours and silent after closing. A third of callers who couldn't get through never called back, they booked with the clinic next door.
Hiring more receptionists per location was expensive and still left evenings and weekends uncovered.
The peak was visible without any analytics: from 9 to 11 in the morning and again after 5, every line at a busy clinic was occupied. Callers got a busy signal or waited while the administrator finished with the patient at the counter. A third of those who could not get through never called back, they booked with the clinic next door.
After closing, the phones simply went dark. Evening is exactly when a filling falls out or a crown comes loose, and Saturday morning is when parents plan a checkup for their child. Those calls landed on an answering machine nobody listened to until Monday.
Nine locations also meant nine schedules. A patient dialed the nearest clinic, learned the orthodontist only worked there on Tuesdays and Thursdays, and was asked to call a different number. Most did not.
Why earlier attempts failed
More front-desk hires An extra administrator at every location was expensive, and even a second shift ended before the evening calls started.
A shared call center One team answering for all nine clinics did not know individual doctors' rotations, so it transferred patients back to the clinic whose line was already busy.
IVR menu with extensions Callers pressed zero to reach a human or hung up, and the menu could not actually place a booking.
Voicemail and a callback list Overnight messages piled up, and by the time someone worked through the list in the morning the patient already had an appointment elsewhere.
The solution
A MyChatBot voice agent now answers on every clinic line, knows all nine schedules, and books the patient while they are still on the call.
One agent, nine lines Every clinic number runs on MyChatBot Calls SDK, so the agent picks up on the first ring, day or night, in the language the patient starts speaking.
Load the network map Business Domain docs in the Knowledge Base carry doctor rotations, the service list, preparation rules and directions for each site, and Agentic Search returns the answer for that specific clinic instead of a generic one.
Book during the call Google Calendar and Altegio let the agent see real openings by doctor and chair and place a confirmed appointment before the call ends.
Every call into CRM create_crm_lead and add_crm_client_contact write the patient into KeyCRM with the source tag "AI-", and lead_chat_link attaches the transcript and recording so the administrator sees exactly what was promised.
Acute pain goes to a human Hand-off Control routes acute pain, post-treatment complaints and any call with the stop word to a live administrator, while Flight Control alerts that clinic's Telegram group immediately.
Close the ones who hesitated Follow-ups call back people who asked about implants or braces and did not commit, and Lead Processing labels and statuses keep consultations, hygiene recalls and reschedules in separate queues.
Knowledge Base: The Knowledge Base holds a separate Business Domain doc per clinic: doctor rotations and specialties, the service list with what a visit includes and how to prepare for it, which cases start with a consultation (implants, orthodontics), address, entrance and parking notes, and the network's rules for what counts as an urgent case.
Within a month the network booked a third more appointments without adding reception staff.
What was deployed
Agents used
Voice AgentLead QualifierAppointment Booker
Channels connected
VoiceWhatsAppWeb widget
Rollout timeline
Days 1 to 3
The instruction was assembled in the Configuration Wizard, and profiles of all nine clinics plus the service list were loaded into the Knowledge Base.
Week 1
Two clinics went live on the Calls SDK with calendars connected, and recorded scenarios were replayed in battle-testing until booking and triage came out clean.
Weeks 2 to 3
All nine numbers were switched over, CRM tools and Flight Control routing were mapped per location, and hand-off rules were tuned with the head administrators.
Month 1
Full round-the-clock coverage, and the network compared the first month against the same period when only the front desk answered the phone.
Results
Metric
Before
After
Missed calls
Busy signal at peak hours, answering machine after closing
0
New bookings
Capped by how many calls the front desk could physically take
+34%
Phone coverage
Working hours only, nothing in the evening or on Sunday
24/7
Answer time
Patients waited on hold or hung up
Answered on the first ring
Front-desk load
The administrator was torn between the phone and the patient at the counter
Same headcount, the phone handled by the agent
"The phone is never busy now, and nobody loses a patient after 6pm."
Olena Marchuk
Operations Lead
What comes next
Outbound calling on the Calls SDK is next: reminders the day before a visit and check-in calls after treatment. WhatsApp is being connected in parallel so the agent can send the preparation notes it currently reads out loud.
Ship something like this
Set up your first AI agent in 10 minutes. No credit card needed.