PICC-AI - Transforming intelligence into actionable knowledge
Ingénierie et Architecture

Today, companies are facing an increasing local and international competition. They need to improve and accelerate their innovation process. Greater knowledge management is directly linked to innovation. Lots of time is spent in searching or waste in testing unnecessary or redundant experiences. Poor visibility into R&D information combined with siloed data leads to delays or even no results. Also, employee turnover causes permanent loss of unrecorded knowledge. Moreover, existing solutions fail to keep people motivated to share and update knowledge.
In this context, PICC software, is a collective intelligence platform to increase the overall performance through faster development and best practices sharing. PICC consists of collecting scattered know-how and individual experiences of employees, and extracting knowledge contained in various documents. Currently, the process of making knowledge accessible and actionable through PICC is still time-consuming and not attractive for users, as knowledge is manually extracted offline by experts.
To improve this process and enhance user experience, a solution based on the combination of user engagement with AI algorithms is proposed. Our solution takes the shape of a user-friendly module: PICC-IA, an Intelligent Agent capable of extracting and sharing knowledge among users in a collaborative and personalized environment. Based on a bottom-up approach, we integrate intelligent guidance using machine learning and user engagement mechanisms. The combination of these elements guides and encourages the employees to share the knowledge related to their R&D issues. Our solution aims to improve users’ performance by 50% during the task of knowledge construction, search or exploitation.
In practice, PICC-AI allows users to automatically generate Knowledge map from various kinds of documents. This is done using a pipeline starting from text extraction and performing multiple language processing tasks such as text classification, topic modeling, name entity recognition and parameter extraction. The knowledge map automatically extracted can then be accessed and, if needed, edited by the users through the existing PICC interface. Our poster will focus on the inner workings of this pipeline.