The project stems from the need to automatically answer the numerous questions addressed to the Order by pharmacists, with the aim of reducing the workload on the secretariat and optimizing the internal management of requests. An intelligent chatbot prototype has been developed, based on the Retrieval-Augmented Generation (RAG) technique, capable of retrieving official documents, such as the website of the Pharmacists’ Order, the one of the national social security and welfare institution for pharmacists and PDF versions of the relevant regulations, and of generating appropriate answers to the questions posed by members. The solution is structured into three main components: a cloud folder, where all the information sources are collected and organized by topic; a preprocessing software that processes the documents daily and indexes them into a vector database; and a chatbot system that, through advanced RAG techniques, correctly responds to requests, distributing the service through an API server integrated with the Order’s official systems.
Collecting and organising documentary sources
In cooperation with the Order's operators, all the documentary sources needed to answer the questions were collected. This phase involved a careful selection and digitalization of the material, organized in a shared structure.
Deliverables:
- Folder containing all documents.
System development and infrastructure integration
An AI system based on automatic preprocessing and RAG (Retrieval-Augmented Generation), integrated in a chatbot, was designed and implemented. The system is connected to the colud folder and automatically updates the data. The chatbot exposes an API that can be integrated into the Order's systems.
Deliverables:
- Preprocess software.
- AI chatbot with API server.
Validation and performance measurement of the prototype
The prototype was tested with the support of the Order's experts. A dataset of questions and answers, both human and AI-generated, was constructed to verify the accuracy and consistency of the answers.
Deliverables:
- Validated Q&A Dataset.
- Performance Evaluation Report.
TThe system is now able to respond effectively and autonomously to users' queries regarding the most common administrative procedures such as registrations, payments and standard fulfilments without requiring the intervention of operators. This has made it possible to significantly reduce response times, lighten the load on human resources and allow the Order's staff to concentrate on more complex and high value-added activities, where the human contribution remains irreplaceable.
AI Research Scientist
AIRIC
Piergiuseppe
Pezzoli
AI Research Scientist
AIRIC
Tomaso
Castellani
Director
Ordine dei Farmacisti
Carlo
Salis
Administrative Assistant
Ordine dei Farmacisti
Sara
Fogliani
President
FOFI
Andrea
Mandelli
Director
FOFI
Mariarosa
Tedesco
