AI governance for finance

Grant Thornton said on April 20 that finance teams should treat artificial intelligence tools like financial reporting systems, with tested controls and evidence that can survive review. (grantthornton.com) The firm said written policies and review committees are no longer enough once artificial intelligence moves into live workflows, where risk shows up during use rather than only before launch. It said boards and regulators now want “traceable decisions, testable controls and evidence” that can be reported consistently. (grantthornton.com) In finance, that means documenting what data goes into a model, who can approve its use, how exceptions are handled, how output is checked, and whether performance drifts over time. Grant Thornton framed those controls as part of day-to-day operations, not a cleanup step after deployment. (grantthornton.com) The warning lands as companies push artificial intelligence deeper into accounting, forecasting, close processes and reporting. KPMG said in a November 2024 guide that companies using artificial intelligence in financial reporting need entity-level controls, process controls and general information-technology controls, with management and board oversight. (kpmg.com) Audit oversight bodies have been moving in the same direction. The Public Company Accounting Oversight Board said in a July 2024 staff spotlight that it was studying whether increased use of technology-based tools in preparing and auditing financial statements could require guidance, standards changes or other regulatory action. (pcaobus.org) Internal auditors have also been given a playbook. The Institute of Internal Auditors said in its September 13, 2024 framework that audits of artificial intelligence should cover governance, management and controls over data, algorithms and cybersecurity, with an emphasis on transparency, traceability and accountability. (theiia.org) Grant Thornton tied the issue to a wider “AI proof gap” in a 2026 survey of 950 C-suite and senior business leaders. The firm said 78% lacked strong confidence they could pass an independent artificial intelligence governance audit within 90 days. (grantthornton.com) The same survey said organizations with fully integrated artificial intelligence were nearly four times as likely to report artificial-intelligence-driven revenue growth as companies still in pilot mode, 58% to 15%. Grant Thornton’s argument is that tighter controls are becoming part of how finance teams defend numbers, not just how they adopt new software. (grantthornton.com)