Enterprise systems recover stale business rules through transparent key lookups
Recovering Business Rules as Cascade Lookups for Enterprise Field Prediction
Machine LearningArtificial IntelligenceDatabasesInformation Retrieval
Summary
Many business software systems use written rules to decide details like payment and shipping, but these rules often get outdated or missing, causing errors. The authors found a way to automatically recover these missing rules by analyzing past transactions and using AI to suggest important data connections. Instead of using complicated models during daily operations, their approach creates simple lookup tables that people can see and update easily. This method works best when the business rules don't change frequently and helps spot when rules need updating.
What this means in practice
- •For enterprise software teams: Automate recovery of outdated or missing business rules from past sales data to reduce manual entry errors and improve order processing accuracy.
- •For database administrators: Use drift analysis on lookup cascades to diagnose which business rules need updating based on recent transaction errors.
Authors
Changxu Duan
Abstract
Enterprise sales-order systems rely on manually maintained configuration rules to determine fields such as payment terms, shipping points, and sales ownership. When these rules are incomplete or stale, operators must enter values by hand, and downstream errors spread into billing, logistics, and reporting. This paper studies whether missing or outdated rules can be recovered from historical transactions instead of being relearned as opaque statistical patterns. We introduce a knowledge-graph-guided workflow in which an LLM agent proposes candidate business keys, meaning column combinations such as customer and document type, validates them through SQL profiling, and compiles the resulting evidence into deterministic cascade lookups. The LLM is used only during offline rule discovery. At inference time, prediction is a transparent key lookup with no model-serving requirement. Experiments on enterprise sales-order prediction and public relational benchmarks show that cascades are most effective when fields follow stable business rules, while learned graph models remain preferable when prediction depends on temporal or neighborhood patterns. Drift analysis further turns residual errors into diagnostics for configuration maintenance. Code and data: https://github.com/Fireblossom/CascadER.