Financial institutions are currently struggling to move artificial intelligence from isolated experimental pilots into trusted, production-level environments within their regulatory reporting cycles. This friction, which Regnology has termed the "agentic gap," represents a significant hurdle for banks attempting to scale automated workflows. While 87% of practitioners surveyed across 22 countries are actively exploring or piloting AI, only 16% have successfully reached embedded, production-level use. This disparity suggests that while the technological capability for agentic AI exists, the institutional ability to govern its outputs remains the primary bottleneck. For the banking sector, closing this gap is not merely a matter of technical adoption but a strategic necessity to address the massive operational costs currently tied to manual reporting processes.
The Economic Burden of Manual Reporting Workflows
The financial motivation to bridge the agentic gap is driven by the high cost of maintaining legacy reporting structures. According to research by Oliver Wyman, commissioned by Regnology, regulatory reporting typically consumes between 1% and 3% of total expenditure at the profiled banks. A significant portion of this cost—estimated between 30% and 50%—is dedicated to running the reporting process internally, a figure that reflects the heavy reliance on manual labor. In Tier 1 case studies, Oliver Wyman estimates that agentic workflows could potentially address 15% to 25% of this reporting spend.
Regnology’s research, which surveyed 276 practitioners between February and July 2026, indicates that the constraint to scaling AI is rarely the maturity of the technology itself. Instead, the "agentic gap" is defined by deficiencies in data quality and the governance frameworks required to manage autonomous agents. As Linda Middledith, Chief Product & Engineering Officer at Regnology, notes, the level of authority granted to AI is not a universal setting but a process-specific decision. Institutions must decide, on a case-by-case basis, whether an agent is simply explaining data, recommending actions, or executing defined work under human oversight.
Navigating Governance and the EU AI Act
Transitioning from experimentation to production requires a structured approach to risk management and regulatory alignment. Regnology suggests that moving toward scaled deployment involves matching the authority given to an AI agent with the specific risk and repeatability of the task at hand. This ensures that all outputs remain traceable and reconstructible, a critical requirement for maintaining audit trails in highly regulated environments. Furthermore, any deployment of agentic workflows must align with emerging legal frameworks, such as the EU AI Act, to ensure compliance.
To address these complexities, Regnology has introduced its RGI (Regnology Intelligence) layer, which aims to integrate explainability and AI-assisted decision support with agentic workflows under human oversight. The company’s findings highlight that the difficulty in adoption lies in the scarcity of professionals who can translate complex regulatory logic into safe, systemic applications. By providing a diagnostic framework that maps the cost of inaction against current AI readiness, the report encourages institutions to baseline existing processes before attempting to pilot new technologies. This methodical approach is designed to prevent the errors that conservative banking cultures naturally seek to avoid in the zero-tolerance environment of regulatory reporting.
Key Takeaways
- Regulatory reporting accounts for 1% to 3% of total bank expenditure, with 30% to 50% of that spend dedicated to internal process management.
- Only 16% of financial institutions have achieved embedded, production-level AI use, leaving 89% of the industry still in the experimentation phase.
- Oliver Wyman estimates that agentic workflows could address between 15% and 25% of reporting spend in Tier 1 banking case studies.
FinanceInsyte's Take
In our view, the "agentic gap" identified by Regnology is a clear signal that the era of "AI for the sake of AI" in fintech is ending, replaced by a rigorous demand for functional, governed utility. The fact that 89% of institutions are stuck in the pilot phase suggests that the industry is hitting a wall where technical novelty meets regulatory reality. Banks are not failing to adopt AI because the models are weak; they are failing because their data architecture and governance protocols are insufficient to support autonomous decision-making. For institutional leaders, the strategic priority must shift from testing LLM capabilities to hardening the data pipelines and "explainability" layers that allow these agents to operate within the strict confines of the EU AI Act and other global mandates. The real winners will be those who treat AI integration as a governance project rather than a software upgrade.
Questions & Answers
How much of a bank's total expenditure is currently consumed by regulatory reporting?
According to research by Oliver Wyman, regulatory reporting typically absorbs between 1% and 3% of the total expenditure at the banks profiled in the study.
What is the primary barrier preventing banks from moving AI into production?
The research suggests the binding constraint is not the maturity of the technology, but rather the quality of the data and the robustness of the governance frameworks surrounding it.
What is the estimated financial impact of implementing agentic workflows in Tier 1 banks?
In illustrative Tier 1 case studies, Oliver Wyman estimates that roughly 15% to 25% of reporting spend could be addressable by agentic workflows.
How should institutions determine the level of authority to grant to AI agents?
Institutions should match the authority given to an AI agent to the specific risk and repeatability of the individual process, ensuring that all outputs are traceable and comply with regulations like the EU AI Act.
Source: Regnology
