DentOP
Client
AI TRANSFORMATION
DentOP AI Transformation
A phone receptionist that answers when the front desk is closed, a knowledge engine built from real calls, and an X-ray workflow for the clinic’s doctors.
Healthcare
Industry
Voice AI, Knowledge Pipeline, X-ray Workflow
Services
Romania
Region
2025-present
Year
Patients call when the front desk is closed.
DentOP’s front desk works set hours. Patients call outside them anyway, before work, late in the evening and at weekends, and some of them call from Germany, Italy, Spain and the UK. Before the AI receptionist, those calls went to voicemail.
During the day, the team had a different problem. Operators repeated the same answers about prices, locations and preparation on every shift, and doctors read panoramic X-rays sent on WhatsApp between their own appointments. The brief was to add AI to both workflows without replacing the telephony, the CRM or the WhatsApp channel the team already uses.
After hours
Calls outside front-desk hours reached voicemail
Repetition
Operators answered the same price and location questions on every shift
Diaspora
Patients calling from Germany, Italy, Spain and the UK
X-ray queue
Doctors read WhatsApp X-rays between appointments
Three connected systems, voice, knowledge and X-rays.
The voice agent runs on OpenAI Realtime and answers DentOP’s real phone number over SIP when the front desk is closed, through the clinic’s existing While1 telephony. It works out what the caller needs, answers in natural Romanian, takes appointment requests and routes anything outside its scope to the team. Every transcript and call summary goes into AmoCRM, next to the patient record, so nobody types anything twice.
Behind the agent, a knowledge pipeline turns thousands of recorded operator calls into what the agent knows, and a redaction loop with two models removes patient data before any text reaches OpenAI. For X-rays, DentOP runs a structured workflow: a case arrives from WhatsApp or from an operator, the system works through the image in stages, and a doctor reviews the result in a dedicated app.
Three systems across the clinic: voice, knowledge, X-rays.
01.
AI phone receptionist
Voice Agent
Answers the clinic’s real number over SIP through While1 whenever the front desk is closed. A call state machine keeps each call predictable, the agent routes anything outside its scope to the team, and it recognises returning callers. Each transcript and summary lands in AmoCRM.
02.
The knowledge engine
Knowledge Pipeline
A four-stage pipeline turns recorded operator calls into the agent’s knowledge. Postgres works as the queue, and four workers process each recording exactly once. A three-pass redaction loop with a two-model check removes patient data before any text reaches OpenAI.
03.
X-ray reading workflow
Radiografii
Operators open one case per patient in a dedicated web app and attach the X-rays. The app assigns each case to a doctor in rotation and sends a push notification to the doctor’s phone. The workflow is built around the doctor: the system prepares a structured draft in stages, and the doctor confirms or corrects it in a guided form, so each review adds to the clinic’s own reading engine.
What the call log shows.
525
Patient calls answered outside front-desk hours, April to mid-September 2026
170+
After-hours calls a month at the current volume
27%
Of summarized calls ended with an appointment request
Zero
Patient PII in the knowledge store
Source: DentOP’s AmoCRM call log, 1 April to 14 September 2026. Appointment requests come from the calls with an AI summary, a feature that went live in August. Covering the same hours with people would take about three full-time operators, including night shifts, and the AI’s running cost is about a tenth of that.
Want this across your business?
We start with the workflow where you lose the most time or revenue, build AI on the systems you already use and run it with your team, as we do at DentOP.
Starts with a free 30-minute discovery call.
Next case study
AI copilot for Romanian public procurement evaluation.