A Tebache

Hospital Medical Coding AI Project

An AI solution that solves a real hospital problem

Hospitals lose millions each year due to medical billing errors and insurance claim rejections. These errors occur when complex surgical procedures do not match official reimbursement codes, or when pricing anomalies go unnoticed.

Consequences:

- Millions of euros not reimbursed or lost

- Overwhelmed administrative staff manually correcting claims

- Delayed reimbursements impacting hospital cash flow

- Risk of sanctions for erroneous billing or unintentional fraud

I developed an AI copilot for a hospital that ingests patient records, automatically matches procedures with reimbursement codes, and instantly detects risks of fraud or pricing anomalies.

Results: Drastic reduction in errors and rejections, significant time savings for teams, and assured GDPR/NIS2 compliance.

A healthcare organization contacted me to train its team—doctors, administrative staff, and paramedical personnel—on the proper use of AI and its associated risks, particularly phishing and data breaches.

The initiative stemmed from practical needs, as some individuals were already using consumer AI tools to save time, without fully understanding the implications when handling health data. Therefore, concrete awareness, not just theoretical cybersecurity lessons, was necessary.

I structured the session around the daily activities of these teams: patient records, medical confidentiality, and communications with insurance providers and laboratories. A dedicated module focused on phishing, featuring realistic examples of attempts specifically targeting medical facilities (fake social security notifications, false lab reminders, etc.). The goal was for everyone to leave knowing how to recognize warning signs, without unnecessary jargon—a doctor, a secretary, and a nursing assistant have different digital reflexes, and the training needed to resonate with everyone.

At the end of the session, I provided a reference guide, designed as a quick reminder rather than a document to be filed away and forgotten.

The result: a team capable of identifying risky AI usage or a phishing attempt before it becomes a problem. Not security experts, but a shared, practical, and useful awareness of risk in their daily work.

This kind of vigilance is particularly crucial in the healthcare sector, where sensitive data is handled with increasingly used AI tools, often without a clear framework.

I designed and delivered Habla ia—a comprehensive conversational AI agent that handles medical appointments, replacing the need for administrative staff, 24/7.

The center serves 10 doctors (general practitioners, pediatrician, cardiologist). Before Habla ia: phone queues, unreachable clinic in the evenings, underutilized appointment slots.

After: Patients book their appointments in under 2 minutes, at any hour—instant SMS confirmation included.

What I delivered:

Conversational AI agent in natural French, integrated with the Claude API. The agent guides the patient step-by-step: reason for visit → suitable doctor → available slots → identity → confirmation. It detects medical emergencies and redirects to 112.

Real-time medical scheduling—each confirmed appointment is automatically added to the relevant practitioner's calendar, with no double-booking possible.

Full-stack web application—patient interface (chat), doctor dashboard (filtering by practitioner, history), Google Business page with reviews and hours.

GDPR by design architecture—minimal data, explicit consent, AES-256 encryption, EU hosting, automatic deletion after 3 years. Compliant with GDPR (EU 2016/679) and Belgian law of July 30, 2018.

Technical Stack:

Claude API (LLM) · React / JSX · Node.js · REST API · AES-256 · Google Business API · SMS webhook · Healthcare prompt engineering

Measurable Results:

+40% of appointments booked outside of office hours

< 2 min for a complete appointment booking

4.9 ★ patient satisfaction on Google (127 reviews)

0 medical data stored by the AI