This thesis presents a governable, end-to-end methodology for validating and interpreting operational expenditure (OPEX) data managed in spreadsheets, combining deterministic data-quality controls with Large Language Model (LLM)–based insight generation. Spreadsheet-centric reporting remains widespread, yet it is prone to structural inconsistencies, semantic ambiguity, and error propagation, which can undermine decision-making. At the same time, LLMs can support interpretation and narrative reporting, but their non-deterministic behaviour and potential hallucinations require explicit safeguards to be acceptable in professional and assurance-oriented contexts. The proposed solution implements a layered architecture integrated with Microsoft Excel and a dedicated backend service. Upstream components enforce a Control and Validation stage based on business rules and objective checks (e.g., schema conformity, units, ranges, completeness, and consistency).