GermanyLogistics and transport2024 – ongoingHapag-Lloyd
Generative AI drafting of corporate audit findings and executive summaries
Hapag-Lloyd's auditors wrote findings in 5 minutes instead of 15 with a generative AI prototype
Problem
Corporate audit reporting at Hapag-Lloyd involved many manual steps. It was slow and produced inconsistent documentation. The company had wanted to cut report-writing time for a couple of years.
Approach
Hapag-Lloyd built a prototype Finding Generation Interface on Databricks that turns auditors' bullet points into written findings and executive summaries. It tested Llama 2 70b and Mixtral, then chose the DBRX model and fine-tuned it. MLflow evaluated prompts and models, and a retrieval-based chatbot lets auditors query documents in natural language.
Results
Takeaway
The time savings came from drafting and review, which are steps the auditors still own. The interface was still a prototype.
Categories
- Size
- Large enterprise
- Industry
- Logistics and transport
- Business function
- Finance, risk and legalInternal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Generative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes Hapag-Lloyd's prototype that generates audit findings and executive summaries with a fine-tuned DBRX model. It reports that writing a finding from bullet points fell from 15 to 5 minutes. Review time per executive summary fell from 30 to 7 minutes.
Full details are in the original sources.
