Measuring agentic AI adoption and control frameworks in finance

Agentic artificial intelligence (AI) systems can execute actions rather than merely generate content, raising distinct governance and operational risk questions for financial institutions. This study measures how agentic AI is entering U.S. finance firms’ annual filings by treating disclosures as text-as-data. We assemble a balanced panel of 2,500 firm–year observations (500 firms per year) from 2021–2025 and implement an auditable dictionary-and-context approach that flags agentic references and then quantifies the surrounding “controls density” (governance and safety language) within the same local disclosure window. Agentic disclosures are absent in 2021–2023, appear in 2024 (0.4% of firm-years), and increase in 2025 (1.6% of firm-years), indicating a late but accelerating diffusion phase. Within the set of agentic-mention filings, autonomy evidence remains rare. However, it focuses on regions with higher control density, consistent with governance maturity serving as a prerequisite for action-taking deployments. The analysis provides a transparent measurement framework and baseline statistics for tracking the emerging shift from AI discussion to action-oriented, agentic deployments in finance.