- Client
- Large electronics group
- Type
- Technology transfer project
- Pillars
- De-riskConsolidate
- Duration
- December 2023 to April 2024
The challenge
This large electronics group wanted to make a business application easier to learn and more accessible. The idea: let users control the software by talking to it in everyday language.
CATIE’s role
- Design of an autonomous AI agent based on a large language model
- Function calling: user requests become actions in the software through its APIs
- Retrieval-augmented generation (RAG) to answer questions from the documentation
- Automatic evaluation of the model’s answers to make the assistant more reliable
Results
- Features beyond the initial request, such as searching the client’s documents
- Code easily taken over by the client’s teams
- The RAG component prioritized for integration into the application
- New avenues for collaboration around generative AI
