I came from medicine and I build the AI that clinics can trust
Grigorii owns the AI side of BRANDAY, from call transcription and scoring models to the data structures behind MedBook. He implemented five speech analytics platforms in working clinics before building AI Call Manager and MedBook, both now live across three clinic networks and 10 branches in Russia, Kazakhstan and the UAE.
Qualifications
Where the expertise comes from
AI in Healthcare Specialization
Five courses covering clinical data, machine learning for healthcare, evaluation of AI applications in clinical settings, and a capstone. Completed July 2026 under Nigam H. Shah, Associate Professor of Medicine in Biomedical Informatics.
- Introduction to Healthcare
- Introduction to Clinical Data
- Fundamentals of Machine Learning for Healthcare
- Evaluations of AI Applications in Healthcare
- AI in Healthcare Capstone
BA International Business and Management, First Class Honours
Business foundation applied to clinic economics, pricing structures and market entry. Currently studying Executive MBA in Healthcare Management.
Meta Certified Digital Marketing Associate
Paid acquisition and measurement, applied to clinic patient flow rather than generic lead volume.
Verify badgeWHY CLINICS AND NOTHING ELSE
I have been on both sides of the clinic door
I studied medicine at Sechenov University in Moscow, Russia's oldest and leading medical school, and spent two consecutive years in clinical placements there before moving into business. What I took from those years was the floor: what a clinic day looks like, why a doctor has no time for a dashboard, why a receptionist under pressure gives the answer she remembers rather than the correct one.
The business side came after, a First Class degree at Middlesex University Dubai and years spent on clinic marketing and unit economics. Most healthcare software fails on one side or the other. Built by clinicians it ignores revenue. Built by marketers it ignores how care actually runs.
I build in the middle. Not analytics for analysts. Tools that a receptionist opens on her first day, a clinic manager reads in two minutes, and an owner can put a number against.
Expertise, stated as facts
Education, markets and projects
- Products
- Author of AI Call Manager, our speech analytics system for clinic contact centres, and of MedBook, our service card platform.
- Platform experience
- Implemented and compared five different speech analytics platforms in live clinic environments before building our own. That comparison defined what AI Call Manager includes and what it deliberately leaves out.
- Technical scope
- Call transcription, scoring models on 15 criteria, refusal reason taxonomies, clinic data modelling and reporting pipelines.
- Markets
- Three markets, the UAE, Russia and Kazakhstan. Systems run in English and Russian, with Arabic in development.
- Delivery
- Implementation across 10 clinic locations, including contact centre staff training.
- Reporting
- Built over 15 report types that clinics had never seen from their own call data.
Work inside clinics
Three cases
Speech analytics across two clinic networks
Alfa Clinic, Kazakhstan. DNK Clinic, Russia.
Both networks recorded every call and read none of them. Management could see call volume by branch and could not answer the only question that mattered, which is why a patient who called about a specific service never appeared in the schedule.
What we did
- Deployed speech analytics across contact centre operations in both networks
- Built reporting by branch so locations could be compared on identical metrics
- Built doctor level conversion reporting showing which specialists turned inquiries into appointments
- Applied a fixed refusal reason taxonomy to every call that did not end in a booking
- Segmented patients into first time and returning
- Tracked topic level demand, including vaccination inquiries
- Tracked promotion mentions, showing whether operators actually raised active promotions on calls
- Configured classification rules separating calls and chats into the correct categories
- Mapped operator address books to telephony so every call was attributed to a named operator
- Trained contact centre staff on the findings rather than handing over a login
Operator performance improved by 40 percent. Bookings increased, script command improved, and dialogue quality rose against the scoring criteria. We delivered over 15 report types built on data the clinics already owned and had never been able to read.
MedBook in three markets
Alfa Clinic Kazakhstan, DNK Clinic Russia, pilot clinics in Dubai.
Contact centre operators answer patient questions from memory, from scattered PDF files and from group chats. When a patient asks what a package includes, what each part costs separately, and which doctor performs it, the operator either guesses or puts the patient on hold. Both outcomes lose the booking.
What we did
- Built three card types for live conversations. Doctor cards carry photo, specialty, experience and price tags, biography, branches where the doctor accepts patients, and booking slots by branch. Service cards carry inclusions, how it works, who it is for, who performs it, where it is available, a pricing breakdown comparing bundle against separate purchase, an operator script tip, upsell options and a loyalty note. Promotion cards mirror the service structure with a pricing breakdown comparing promotional against standard price.
- Built fuzzy search, because operators do not type correctly under time pressure. A misspelled doctor name, a partial service title or the colloquial term the patient used still returns the correct card.
- Deployed in the CIS networks first, then opened the UAE market.
MedBook is running in Kazakhstan and Russia and is now in pilot with clinics in Dubai.
Five platforms, one conclusion
Comparative implementation work, UAE and CIS.
Clinics buy speech analytics built for banks, telecoms and outbound sales floors. The taxonomies do not fit medicine, the scoring criteria measure the wrong behaviour, and the reports answer questions a clinic manager never asks.
What we did
- Implemented and ran five different speech analytics platforms inside working clinics
- Mapped where each one broke on medical vocabulary, on multi branch structures and on doctor level attribution
- Used that mapping to define the Phase 1 scope of AI Call Manager
AI Call Manager labels every call booked or not booked, assigns one refusal reason from a fixed taxonomy, and attaches the transcript evidence behind that label. Filterable by branch, operator, doctor and period, with a custom report builder and Excel export.
What I believe in this profession
Four positions we work by
A tool nobody opens changes nothing
We design for the receptionist and the doctor first, and for the report second.
Recordings hold the answers analytics miss
Ad platforms show clicks. The call shows the moment the patient decided not to book.
One record per service, one source of truth
When marketing, doctors and the contact centre read the same record, the clinic stops contradicting itself.
AI reports facts, people make the decision
Our models score and flag. The clinic manager still owns the action that follows.
Articles by this author
Written by Grigorii Chigrinets
I will not walk into a clinic, I will ask AI first
Being named in an assistant answer follows different rules than ranking on a search page.
Why clinics lose patients on the phone, not on advertising
The call is where the booking is won or lost, and nobody listens to it.
Why every clinic needs one knowledge base for its services
One source of truth for prices, indications and doctor names.
Start with the diagnostic
We measure what happens on your calls before we propose anything. The diagnostic is free and it produces a written finding, not a sales call.
All insights