How a Multi-Entity Tax Team Cut Data Prep 70%

The client, a multi-entity U.S. enterprise, manages direct and indirect tax across multiple jurisdictions. Its finance and tax data sat in separate ERPs and spreadsheets, and every reporting, provision or compliance cycle started with people extracting, cleaning, standardizing and mapping that data by hand. Book-to-tax reconciliation was slow, classification varied by who did it, and getting consistent, current numbers took far more effort than the tax work itself.

XIT Solutions deployed Complyia, its finance and tax data platform, for the client. The platform pulls data from the client's ERPs and spreadsheet sources into one data layer, then uses locally hosted language models to interpret financial records, classify transactions into tax categories, map accounts and flag discrepancies. Tax calculations themselves run on rules. Anything the models are unsure about, or anything that looks inconsistent, goes to a tax professional for review, and approved corrections improve later runs. Every classification and calculation traces back to the source record, so the tax team can audit how each number was produced.

The team started with the data and the tax logic, making sure source data, tax categories and calculation rules were correct before training anything. Connectors brought the client's ERP and spreadsheet data into a single layer. On top of it, locally hosted models served through Ollama, including Llama 7B and 10B and Qwen 12B and 24B, handle financial data interpretation, transaction classification, account mapping and discrepancy detection. Running the models locally kept sensitive financial data inside the client's environment. Rules handle the tax calculations, and a review queue sends low-confidence mappings and anomalies to tax professionals, whose approved corrections feed later runs. Results were measured in a controlled before-and-after assessment: the same financial datasets and reconciliation workloads were processed manually and through the platform, and the classifications were checked against tax-expert-reviewed answers. The hardest part was accuracy, since tax rules change by jurisdiction, entity and transaction type.

Mid-market and enterprise organizations, tax technology providers and advisory firms managing tax across multiple entities, jurisdictions and financial systems, especially with high transaction volumes in VAT, GST or sales and use tax. It works best where reliable transaction data is available, tax experts and engineers can work together, and the team is willing to standardize existing workflows.