The Problem
A financial services client needed to map every transaction back to its statement of record for regulatory reporting — the kind of work BCBS 239 and IFRS 17 demand. It's the sort of mapping engagement that Big 4 consultancies bill out for hundreds of hours, and the firm couldn't keep paying that rate every reporting cycle.
The data existed across multiple ledgers, transaction systems, and downstream reports. The relationships between them did not. They needed lineage, not a dashboard.
The Approach
We built an AI-powered data lineage pipeline that automates the discovery and mapping process end-to-end.
The system ingests transaction data and statement records, classifies and matches them using a combination of deterministic rules and LLM-based reasoning, then produces a fully-traversable lineage graph with confidence scores and exception flags. Every match is auditable. Every exception is flagged for human review rather than silently swallowed.
We built it as a pipeline that can be re-run on every reporting cycle — the firm doesn't pay for the work twice.
The Outcome
Work that previously consumed hundreds of Big 4 consulting hours per reporting cycle now runs as an automated pipeline. The firm's internal team supervises exceptions instead of reconstructing relationships from scratch. Audit trails are generated automatically.
The system pays for itself within a single reporting cycle — and keeps paying for itself every cycle thereafter.