Agentic AI can already support research, recommendations, and risk checks in capital markets. The harder challenge begins at execution, where firms need clear accountability, supervision, information controls, and governance architecture before agents can move closer to autonomous decisions.