AI Financial Advisory

State of play.

• Anthropic has moved from AI tooling vendor to embedded financial-services infrastructure provider, releasing ten pre-built agents for core banking and finance workflows — pitchbook creation, credit memo drafting, KYC, AML, underwriting, and insurance claims — in partnership with FIS, with BMO and Amalgamated Bank as early deployers .
• Public has positioned itself as the first brokerage to offer AI agents for automated portfolio trading, putting agentic execution — not just advisory — directly in retail investor hands .
• The U.S. Treasury has publicly urged caution on autonomous AI deployment in banking, signaling that federal scrutiny of agentic AI in critical financial infrastructure is intensifying ahead of any formal rulemaking .
• The skill-versus-luck framing for retail derivatives trading is entering mainstream discourse, with industry figures like Bill Perkins publicly articulating expected-value frameworks — a framing that regulators and plaintiffs' counsel will both engage in enforcement and litigation .
• For counsel advising broker-dealers, investment advisers, or banks deploying AI agents, the practical baseline is: agentic execution in client accounts creates fiduciary, suitability, and best-execution exposure that existing compliance frameworks were not designed to address, and regulators are watching.

Where things stand.

• Agentic AI is executing — not just recommending — in client portfolios. Public's AI agents place trades autonomously, compressing the gap between advisory and discretionary management in ways that implicate investment adviser registration, suitability obligations, and best-execution duties .
• Major AI developers are embedding directly in bank and insurer workflows. Anthropic's ten-agent suite — co-developed with FIS and already adopted by Goldman Sachs, Visa, Citi, and AIG — means AI is now inside AML investigations, underwriting decisions, and financial statement audits at systemically significant institutions .
• Regulatory compliance frameworks for deployed agents remain undisclosed. Anthropic has not published details on how its financial services agents satisfy BSA/AML, fair lending, or model risk management requirements — a gap that deploying institutions will need to fill through their own governance structures .
• The skill-versus-luck distinction in retail derivatives trading carries regulatory and litigation weight. As retail options volume remains elevated post-2021, how platforms and thought leaders frame the skill-versus-chance question shapes broker-dealer suitability defenses and potential FINRA enforcement posture .
• Treasury-level concern about agentic AI in banking is on record. The Treasury chief's public caution — issued before Anthropic's agent launch — establishes a federal supervisory posture that banking regulators at OCC, FDIC, and the Fed are likely to operationalize through examination guidance .

Latest developments.

• Anthropic releases ten pre-built AI agents for financial services at an invite-only New York event alongside Claude Opus 4.7; FIS co-develops the suite; BMO and Amalgamated Bank deploy the Financial Crimes AI Agent for AML investigations .
• Anthropic discloses a $1.5 billion enterprise services joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs, one day before the agent launch .

Active questions and open splits.

• When does an AI agent executing trades trigger investment adviser registration? Autonomous portfolio execution by a brokerage's AI agent sits at the edge of the broker-dealer exclusion from the Investment Advisers Act — a line the SEC has not drawn for agentic systems .
• Who bears model risk when an AI agent makes a flawed AML determination? Anthropic's agents compress AML investigations "from hours to minutes," but SR 11-7 model risk management guidance places validation and governance obligations on the deploying institution — not the vendor .
• Does agentic AI in underwriting create fair lending exposure? If AI agents are making or materially influencing credit decisions, ECOA and fair lending examination scrutiny follows — and Anthropic has not disclosed how its underwriting agent handles disparate impact .
• How will the skill-versus-luck framing affect suitability defenses in retail options litigation? If regulators adopt an expected-value framework to define "skill," platforms that gamified options access face a higher bar in defending suitability determinations .
• What audit-trail and explainability requirements will banking regulators impose on deployed agents? Treasury's public caution signals examination scrutiny is coming, but the specific documentation and control standards have not been articulated — leaving deploying institutions to design governance frameworks against an uncertain regulatory target .

What to watch.

• OCC, FDIC, or Federal Reserve examination guidance on agentic AI in banking workflows — the Treasury chief's public caution is the leading indicator that supervisory guidance is in preparation.
• SEC staff statements or no-action requests addressing whether AI-agent portfolio execution triggers investment adviser registration or discretionary management obligations.
• Any FINRA enforcement action or guidance on AI-driven retail trading platforms and suitability obligations.
• Anthropic's IPO timeline — a public offering will force disclosure of regulatory compliance frameworks for its financial services agents, surfacing information currently unavailable to deploying institutions' counsel.
• Whether BMO or Amalgamated Bank's deployment of the Financial Crimes AI Agent draws OCC or FinCEN scrutiny as a test case for AI-assisted AML programs.