The Challenge
BAS processing requires accountants to manually review large volumes of transactions to identify GST discrepancies, misclassifications, and compliance risks. This includes checking for incorrect GST treatments, vague descriptions, misallocated expenses, and edge cases like insurance, government charges, and vehicle apportionment. The process is time-consuming, inconsistent across reviewers, and heavily reliant on individual experience. As transaction volumes grow, maintaining accuracy and efficiency becomes increasingly difficult.
Our Approach
We are developing a rule-based and context-aware system that automatically flags transactions with potential GST discrepancies during BAS preparation.
The system first applies pre-filtering rules to identify high-risk transactions based on account type, value thresholds, supplier patterns, and keywords. It then uses a scoring engine to assess the likelihood of misclassification — such as distinguishing between assets vs expenses, or identifying vague descriptions like 'staff reimbursement'.
If insufficient context is available, the system integrates with Xero to retrieve invoice data, including line items, GST breakdowns, and attachments. This additional context is used to refine classification decisions.
For complex scenarios — such as vehicle-related expenses requiring business-use apportionment, insurance with mixed GST treatment, or government charges — specialised rules are applied based on ATO guidelines.
Transactions that remain ambiguous are automatically flagged with structured reasoning, enabling the accounting team to request clarification from clients efficiently. The system also maintains a clear audit trail for every decision made.