Portima: exploring generative AI and RAG for intelligent knowledge assistants
ChatGPT
LLM
RAG
NLP
SpeechToText
Python
For 30 years, Portima has been an IT partner for insurance brokers and insurance companies in Belgium. With different software frameworks like Portima Connect, Brio and E-GOR, Portima supports Insurance brokers and companies to exchange data in a secure and reliable way, manage contracts, claims and all relevant insurance data and documents.
The challenge
Portima has an extensive library of knowledge items for their products made available to their customers. This library includes both video tutorials & manuals.
They wanted to research the feasibility of developing a Large Language Model based user assistant on their knowledge library. This would help the customers to find answers to their specific questions regarding Portima products and to direct them to the relevant materials in the library.
Our solution
To support both the IT & business interest a dual track was created
A Portima IT focused 3-daytrack in which we:
- Introduced various GenAI & chatbot concepts in theoretic context
- co-created a chatbot POC from scratch
A Portima Business focused track in which we:
- Demonstrated an Intellus created chatbot on Portima data
- Allowed users to change the config through a customization screen to judge the accuracy, speed & overall useability of this POC model.
The business value
Successful POC model that can be extended on with in-house resources
Insight gained
- To assess the feasibility for a potential industrialization of the POC based on the current library content
- To formulate a strategic plan outlining the next steps for further development and potential implementation of the chatbot.
- For Data preparation needs to train an LLM
Increase the overall awareness of the potential benefits and challenges of using LLM Technology, LLM based user-assistants and Machine Learning in General.
20%
Project efficiency improvement
100%
Data Automation rate
1 week --> 24 hours
Date refresh rate
