
Case study
Financial reporting
Building a co-pilot for SBB to assist in monthly financial reporting.
Challenge
Solution
Every department of SBB writes a monthly financial report.
All reports are created by manually analyzing hundreds of financial KPIs.
This is a complex and time intensive task.
Impact
Data pipeline to clean and preprocess KPI data.
Feature engineering: Definition of additional compound features from raw KPI variables based on domain knowledge of report writers.
Report generation: Engineered an LLM (GPT-4) to analyze the KPIs and to write reports matching both style and scope of human written reports.
Time savings: Autonomous generation of the financial report, serving as basis for the creation of the final report by the responsible employee.
Additional insights: The reports generated by the AI provide a second perspective, complementing the human point of view.