With the rapid development of smart and autonomous systems, the rise of increasingly autonomous software and hardware agents has raised concerns for societies in terms of ethics, regulation, accountability, and law enforcement. Agentic AI refers to those artificial intelligence systems that have some level of autonomy and can act on behalf of a human user in a certain context. Industries are already relying heavily on agentic AI to monitor regulation compliance of businesses through natural language processing and a knowledge graph of regulations. Similar AI agents can be deployed in the financial industry to ensure rules compliance. The financial industry is highly regulated. Financial market and institution supervision is crucial but requires astronomical cost and human effort. To lessen the burden, many regulatory agencies already publish guidelines for horizontal regulation to ease compliance. The resulting regulatory text flood is unprecedentedly big and complex and thus serves as great risk for financial misconduct. Banks, as the major subject of compliance, have already deployed advanced models to extract structured information from these regulations. Rules extraction and representation leverages a knowledge graph on regulations with multi-level granularity, knowledge schema development, and automated rule generation based on reinforcement learning techniques. Nevertheless, implementing compliance measures against this knowledge graph requires tremendous human effort on rule coding that costs institutions millions of dollars annually. Compliance rule coding varies across financial institutions and must be ingrained into internal policies, which rules out a consolidated solution. Thus, a novel approach to autonomous compliance implementation against the knowledge graph of regulations is unmet and essential. To achieve goal-oriented agentic AI agent design, a potential solution needs to serve as a benchmark testing ground with the capped regulation space and benign financial domain for human-like compliant AI agents. Regulators can monitor automated compliance efforts and intervene if exploitations are discovered, and a regulatory sandbox is thus proposed.