From Excel to Agentic Finance

This working paper examines how agentic artificial intelligence could transform Financial Planning and Analysis (FP&A) and controlling processes in European small and medium-sized enterprises (SMEs). It focuses on a central question: under which organizational, technological and governance conditions can AI agents create measurable business value rather than simply automate individual finance tasks? The study uses a structured critical evidence review covering academic research, official European statistics, institutional publications, professional guidance and selected industry evidence available up to 5 September 2026. It distinguishes between experimental findings, observational evidence, survey results, consulting cases, projections and vendor claims in order to avoid treating fundamentally different forms of evidence as directly comparable. The paper evaluates twelve potential applications of agentic AI across FP&A and controlling, including forecasting, budgeting, variance analysis, scenario planning, management reporting, working capital management, cash flow forecasting, cost controlling, revenue forecasting, financial data consolidation, decision support and anomaly detection. Particular attention is given to the distinction between task efficiency, process improvement and actual economic value. The analysis argues that time savings or faster reporting do not automatically translate into financial returns and proposes evaluating agentic systems against both the existing process and a simpler conventional automation alternative. As original contributions, the paper introduces an Agentic Finance Readiness Framework, an Agentic Finance Maturity Model and a Use Case Prioritization Matrix designed to support SMEs and finance leaders in assessing readiness, selecting appropriate pilot applications and determining suitable levels of agent autonomy. The readiness framework considers data, process, technology, governance and organizational capabilities. The findings support selective and supervised adoption rather than broad autonomous deployment. Variance investigations, management reporting, cash forecast updates and financial data exception reviews emerge as particularly suitable areas for controlled experimentation. At the same time, the paper emphasizes that conventional automation remains preferable for many stable and deterministic finance processes. The paper is intended for researchers, finance professionals, controllers, CFOs, SME decision makers and practitioners interested in the intersection of artificial intelligence, finance transformation, management accounting and AI governance.