Generative AI: What’s Next for Financial Services

Emerging use cases

Firms are testing AI for drafting and summarizing reports, supporting marketing material review, coding assistance, data visualization, research, HR screening, and workflow automation. Summarization remains the most common and practical use case so far.

Implementation choices

Many firms are weighing vendor-provided AI features against building custom models. The key considerations are cost, integration, data privacy, and maintenance, especially as models update every few months.

Adversarial use and cyber risk

FINRA warned that bad actors are using AI for deepfakes, fraud schemes, and social engineering. FINRA’s intelligence alerts now include updates on these evolving risks, underscoring the importance of staying engaged with your risk monitoring analyst.

Governance and data controls

The panel stressed understanding where firm and customer data travels, especially in multi-layered vendor environments. Encryption, key ownership, and visibility into how models handle data are essential.

AI hallucinations and human oversight

Firms should design prompts with clear context, require AI-generated outputs to include references, and — importantly — keep a human in the loop before using any output in a client-facing capacity.

Agentic AI

While still experimental, agentic AI (autonomous systems that can complete multi-step tasks) could reshape compliance and operations. Panelists cautioned that agents don’t “know the rules,” so firms must build safeguards, monitoring, and access controls before deployment.