Work

Who Answers for the Agent? Agentic AI, Delegated Execution, and Accountability in the Corporate Finance Function

Year: 2026
Type: article
Abstract: Purpose – Agentic artificial intelligence (AI) systems can now interpret goals, select software tools and execute financial transactions with limited human involvement. This paper asks how delegating execution authority to such systems affects financial decision quality, risk exposure and accountability in the corporate finance function, and which organizational conditions keep accountability intact. Design/methodology/approach – This is a conceptual, model-building paper. It synthesizes agency theory, management control theory, dynamic capabilities theory and research on human–AI organizing, and draws on supervisory guidance on AI in the financial sector to identify risk categories. Findings – The paper develops the Algorithmic Financial Accountability Framework (AFAF). AFAF distinguishes permission to advise, to decide and to execute. It proposes that agentic autonomy has an inverted-U relationship with decision quality, increases algorithmic risk exposure, and widens an accountability gap between nominal and effective responsibility. Oversight effectiveness, governance maturity, decision irreversibility and third-party concentration condition these effects. Nine propositions are developed. Originality/value – The paper separates the technical locus of action from the organizational locus of accountability. It introduces the autonomy–governance gap as a measurable, dynamic construct and specifies indicators and research designs for testing the framework. Practical implications – Chief financial officers, audit committees and supervisors can use the framework to tier decision rights by monetary exposure and reversibility, and to require evidence that oversight can actually stop and correct an agent’s actions. Keywords: agentic AI; corporate finance function; financial accountability; managerial oversight; AI governance; agency theory; management control; dynamic capabilities
Language: English
Subfield: Accounting
Open Access status: diamond
OpenAlex ID: W7225437462
FWCI:
Cites: 29
Cited by:
Related to: 10