AI - LLM: Data models reconciliation
AI-Driven Data Model Reconciliation for Atradius: Simplifying complex data integration
1. Business objectives of the client
Atradius, a global leader in Credit Insurance, relies on high-quality, accurate information about companies to assess risk and support their clients’ business decisions. The company works with a vast network of dozens of information providers, each with its own information data model, which creates complexity in integrating and mapping data from diverse sources. To streamline their operations and improve data exchange with these providers, Atradius is switching to a new internal data model. This API will simplify communications, enhance efficiency, and create a more scalable process for managing data. The company sought to leverage Generative AI, specifically Large Language Models (LLMs), to automate and simplify the reconciliation of different data models, reducing manual labor and increasing operational efficiency.
2. The AI / Data challenges faced
Atradius was facing significant challenges in managing and reconciling data from its numerous third-party information providers. Each provider operated with its own unique data model, leading to discrepancies and inefficiencies when merging this information into Atradius’ centralized system. Currently, the reconciliation process is carried out manually by a subcontractor, which is time-consuming and resource-intensive. The manual mapping of various data models requires constant upkeep and consumes substantial resources, with the potential for human error in the reconciliation process. This inefficiency was impeding Atradius’ ability to scale its operations and rapidly process incoming data, which is critical to maintaining competitive advantage in the credit insurance industry.
3. How Bubo helped on this
Our team worked closely with Atradius to design an AI-powered solution that would automate and streamline the data reconciliation process. Here’s how we approached the challenge:
- Detecting unmapped fields: The first step was to detect these yet unmapped fields from various providers data models.
- Automating process and proposing a data mapping: We developed an automated system that could reconcile and map the data from various sources directly into Atradius’ internal data model. The AI system was designed to identify similarities and differences in the data structures and suggest automatic mappings.
- Continuous learning and adaptation: As new data models are added by providers or as existing models evolve, the system is evaluated by underwriters ensuring that system performance is stable.
- Efficient data integration and API usage: Bubo worked to integrate the AI-driven reconciliation system into Atradius’ new data model API, ensuring that the system operates seamlessly with their existing infrastructure.
4. Results
The AI-based reconciliation solution delivered significant improvements for our customer :
TBC…