How we deploy GenAI to streamline complex data processes for a global insurer.
GenAI will revolutionise data management like autopilots transformed air travel. That is the result of our breakthrough project with a global insurance company.
Put into the hands of an experienced pilot, aviation gadgets boost the efficiency of planes. Similarly, guided by data experts, GenAI can greatly advance data management.
Growing data challenges
That became apparent in our project with a leading insurer. Our client had been experiencing typical challenges:
- other companies were acquired
- operating procedures needed to be updated
- new data sources were integrated into their systems
- technology kept progressing
These issues complicated ETL processes. Extracting data from new sources, transforming it and loading it into target databases had become:
- more complex
- less transparent
- slowed by inefficient approval processes
Here, we came in as a trusted partner to apply new technologies such as GenAI to the challenge.
Since previous attempts had stalled, our client’s experts were initially sceptical whether AI could help set up a scalable, flexible and business-oriented data platform. But given our track record, we were asked to develop a proof of concept.
The aim was to accelerate time-to-market for data-driven products. That would not only free up skilled resources from repetitive tasks for higher-value work. It would also empower business departments to independently create and consume data insights.
To achieve that we advanced in three steps:
I. Holistic analysis
We assessed the challenge based on our client’s enterprise architecture with its four levels:
- business capabilities
- data structures
- solutions & applications
- technical architecture with software and hardware.
II. Strategic LLM Selection
We benchmarked multiple large language models and identified Claude as optimal for our code generation requirements.
III: Automated source-to-target mappings
We used Claude to analyse complex ETL transformation logic and data patterns to classify data and generate business glossaries. With the right guidance, Claude understood the context of historical mappings, suggested intelligent transformations and helped migrate legacy systems to modern data platforms.
The results speak for themselves:
- 20-30% efficiency gains on first system integration
- Up to 50% gains on subsequent systems
Plus, we set up a new self-service-architecture which allows colleagues without a background in data management to analyse complex data and drive innovation.
The project shows how experienced data experts can apply GenAI to huge effects in data management. That is not only true for insurance companies, but for any data-driven organisation.
