Compliance monitoring · Production

Hapag-Lloyd cuts audit review time 66–77% using Databricks Mosaic AI GenAI prototypes

The problem

Hapag-Lloyd's corporate audit processes relied on manual, time-consuming report writing and documentation, leading to inefficiencies and inconsistencies that persisted across several years despite leadership recognising the need to change.

First attempt

Hapag-Lloyd's prior infrastructure — including vector databases and an AWS SysOps account — did not support the rapid setup and deployment of AI models needed for audit optimisation.

Workflow diagram · grounded in source
1
Data ingested and prepared
integration
“starting with data ingestion, preparation and prompt engineering and ending with model evaluation and deployment”
2
Finding Generation Interface produces findings
ai_action
“making it suitable for generating high-quality audit findings and summaries. Daniel's team dubbed this prototype the Finding Generation Interface”
3
RAG chatbot queries documents
ai_action
“The chatbot used retrieval augmented generation (RAG) to provide accurate and contextually relevant responses”
4
MLflow automates prompt and model evaluation
validation
“Using Databricks MLflow, Hapag-Lloyd was able to automate the evaluation of prompts and models”
5
Findings stored in Delta table
output
“generating findings, summarizing results and storing them in a Delta table for easy access and retrieval”
Reported outcome

Hapag-Lloyd achieved a 66% decrease in review time per finding (from 15 to 5 minutes) and a 77% reduction in executive summary review time (from 30 to 7 minutes), freeing auditors from administrative tasks.

Reported metrics
Review time per finding66%
Executive summary review time reduction77%
Time to create a finding (current)5 minutes
Time to create a finding (previous)15 minutes
Show all 6 reported metrics
review time per finding66%
executive summary review time reduction77%
time to create a finding (current)5 minutes
time to create a finding (previous)15 minutes
executive summary review time (current)7 minutes
executive summary review time (previous)30 minutes
Reported stack
DatabricksMosaic AIDBRXMLflowGradioDatabricks Model ServingRAGDelta table
Source
https://www.databricks.com/customers/hapag-lloyd
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Hapag-Lloyd achieved a 66% decrease in review time per finding (from 15 to 5 minutes) and a 77% reduction in executive summary review time (from 30 to 7 minutes), freeing auditors from administrative tasks.

What tools did this team use?

Databricks, Mosaic AI, DBRX, MLflow, Gradio, Databricks Model Serving, RAG, Delta table.

What results were reported?

Review time per finding: 66%; Executive summary review time reduction: 77%; Time to create a finding (current): 5 minutes; Time to create a finding (previous): 15 minutes (source-reported, not independently verified).

What failed first in this deployment?

Hapag-Lloyd's prior infrastructure — including vector databases and an AWS SysOps account — did not support the rapid setup and deployment of AI models needed for audit optimisation.

How is this compliance monitoring AI workflow structured?

Data ingested and prepared → Finding Generation Interface produces findings → RAG chatbot queries documents → MLflow automates prompt and model evaluation → Findings stored in Delta table.