Over the years, Fincons Group has constantly invested in running training for new technologies such as the cloud and new programming languages.
GenAI technology, however, is evolving too fast and proving too complex for resources to self-train adequately. In our role as innovators, we want to bridge this gap around technology, around regulations and process skills. We therefore came up with the idea to experiment with a new training method: the Hackathon. Usually, hackathons are externally facing activities but this time we wanted to design a hackathon that actively involved our internal resources and offered them a non-traditional and even playful opportunity to up-skill. The topics addressed by the Hackathon, similarly, were not general solutions but ones that address concrete issues brought to us by the Fincons Business Units dedicated to specific industries, clients or the market.
Usually, training sessions and even work teams will be small groups composed of staff specialised in a specific vertical sector. For the Hackathon, however, we decided to create heterogeneous groups of people from different BUs and departments, involving both people with technical backgrounds and not. This has proven to be a really efficient way of strengthening team spirit across different functions. It has also greatly contributed to helping staff form leadership and autonomous work management skills, regardless of their background.
To support all the groups, we provided four transversal training sessions. These were focused on regulations, ethics, tools for GenAI, prompt engineering, and Retrieval-Augmented Generation (RAG) techniques. The playful nature of this non-traditional training method meant that staff felt really engaged and wanted to take up the challenge to the best of their ability, so they were keen to stock up on all the added value we could provide. In addition to these introductory sessions, we also developed a playground technology platform where teams can upload their documents, tailor how AI works on their documents, select AI models and design their proposals in presentation format. Along with all these tools and sessions, we provided a tutor to mentor each group throughout the process.
The hackathons objective was to develop one of four following conversational AI assistants:
1. RAG Documentation Agent: develop a conversational AI to access project documentation from various sources and answer user queries.
2. HR Assistant: create an AI agent for HR use, capable of parsing resumes, generating conceptual maps, and recommending the best candidates.
3. Image Storyteller Assistant: design an AI tool to describe images for visually impaired users, including an interactive Q&A functionality.
4. Query Wizard: build an AI assistant that translates natural language questions into SQL queries to facilitate database interactions.
A special bonus has been awarded for the use of Small Language Models, a new linguistic model requiring fewer parameters compared to Large Language Models, that is becoming increasingly important.
Finally, the objective is for what has been learned to be reshared among the organization benefitting everyone, reinforcing the collective nature of learning in Fincons and the ethics of sharing skills and knowledge.