Agentic Finance

The Agentic Finance panel explored how AI agents are evolving from analytical assistants into systems capable of taking financial actions. Mohammed framed the discussion around autonomy, data, institutional adoption, and verification; Melitza defined agents as language models connected to tools and external systems, while Sam described them more broadly as systems that sense, reason, and act, including trading, market analysis, and automated investment strategies.

Core Themes or Shifts

Sam highlighted that an agent’s effectiveness will increasingly depend on the quality of its data and the actions it is permitted to take, rather than model performance alone. Dev emphasized the progression from human-directed tools to assistants and eventually sovereign applications that can operate independently, while Milica noted that the appropriate level of autonomy, security, and infrastructure will vary significantly by financial use case and regulatory environment.

Key Strategic Insights

Institutions managing significant capital may require near-perfect system reliability before deploying autonomous agents into production. Achieving that standard will depend not only on stronger models, but on the design of the entire system around them. Deterministic guardrails, secure execution, verifiable inference, audit trails, regulatory compliance, data integrity, and continued human oversight can help ensure that agents act within clearly defined limits. The most credible agentic finance platforms will therefore be those that combine adaptability with transparency, control, and measurable accountability.

Implications for Builders or Investors

Trusted execution environments, multi-party computation, encrypted data, cryptographically signed code, and verifiable inference will be critical infrastructure for proving that agents operate as intended. For builders and investors, the opportunity lies in creating agentic financial systems that combine useful automation with transparent execution, strong controls, reliable data, and clearly defined accountability.