AI-based functional expense classification for nonprofit financial reporting and form 990 compliance

Preparing a Statement of Functional Expenses is an important but often time-consuming task for nonprofit organizations. Accountants must classify expenses not only by type, such as salaries, rent, supplies, and professional fees, but also by purpose, including Program Services, Management and General, and Fundraising. In practice, this process often depends on manual review, spreadsheet workpapers, allocation percentages, payroll records, class codes, grant codes, and supporting documents. As a result, nonprofits may face inconsistent classification, weak documentation, reporting delays, and reconciliation problems.
This study proposes an AI-based functional expense report automation framework for nonprofit financial reporting and Form 990 compliance. The framework uses accounting data from Profit and Loss Detail or General Ledger reports, along with supporting inputs such as vendor information, payroll records, memo descriptions, invoice text, class codes, grant codes, historical classifications, and editable allocation rules. The system classifies direct expenses, allocates shared costs, generates confidence scores, explains the basis for each classification, reconciles totals to accounting records, and produces a draft Statement of Functional Expenses for accountant review.
The framework is not designed to replace professional judgment. Instead, it supports accountants by reducing manual workload, improving consistency, strengthening audit documentation, and identifying transactions that require human review. By combining automation, explainable classification logic, editable allocation percentages, and reconciliation controls, the proposed model offers a practical solution for improving nonprofit reporting accuracy, Form 990 preparation, and transparency for donors, grantors, auditors, regulators, and the public.