AI-Powered Clinic FAQ Assistant with Safety Guardrails

Project introduction

Veterinary clinics may repeatedly receive questions about opening hours, appointments, available services, location, payment methods, and current promotions.

I built a portfolio-based Telegram FAQ assistant that connects incoming messages with the Gemini API through Make.com while applying clear scope and safety boundaries.

Operational problem

• Inventing clinic information
• Answering outside the intended purpose
• Giving unsafe medical suggestions
• Presenting uncertain information as fact
• Failing to direct urgent concerns appropriately

System built

Make.com captures the Telegram message and sends it to Gemini together with approved clinic information and system instructions. The assistant answers routine administrative questions, uses fallback language when information is missing, and redirects medical or emergency concerns to professionals.

Workflow diagram

Tools and technical capabilities

• Automated clinic FAQ responses
• Approved-information-only instruction
• Missing-information fallback
• No diagnosis or medication recommendations
• Emergency redirection
• Structured AI response mapping

Actual project evidence

Key learning or next improvement

The key learning is that AI assistants need clear scope boundaries, fallback behavior, and professional escalation rules before they can be trusted in sensitive service contexts.

Have a similar workflow bottleneck?

I can help map the repetitive steps and identify where practical automation could make the process easier.