Earlier this week Frank Fabozzi and the CFA Institute Research Foundation hosted an informative webinar that discussed the transformative power of Agentic AI in Decision making within Financial Markets with Alicia Vidler, PhD. Their conversation was very candid and focused on Dr. Vidler's work with agentic AI throughout her professional and academic career. I encourage you to watch the conversation here - and read some key highlights further below.

I'll also draw your attention to a great article, written by Isabelle Bousquette from the The Wall Street Journal late last year, about Citi running a 5,000 person pilot to find out how helpful the new “agentic” technology is to staff in areas like research and client profiling.

The firm's Citi Chief Technology Officer, David Griffiths, said "Citi would run the pilot over a four to six week period." Griffith's also said "the internal Citi platform uses a range of models, including Google Gemini and Anthropic Claude." With the new capabilities, a user can direct a tool inside the platform, known as Citi Stylus Workspaces, to research a specific client, build a profile of them from both publicly available data but also multiple internal data sets, and translate it into a foreign language, all in a single step.

Griffiths said he’ll be looking for information on how people use the new capabilities, how impactful they are and how the cost to value ratio plays out.

It's fascinating to learn how "AI agents will be able to autonomously complete complicated assignments much in the way human workers would, chaining together a series of discrete tasks, sometimes over hours or days, and accessing multiple internal company systems." - as Isabelle Bousquette says in her article, which you can find here.

Summary of Agentic AI in Financial Markets Webinar

by CFA Institute Research Foundation

Demystifying Agentic AI: Agentic AI systems are capable of informed decision-making beyond rigid rules, giving an agent the autonomy to "check the fridge and decide what's missing" rather than just "buy milk." This leap requires new approaches to transparency, trust, and reliability and introduces new risks.

Human in the Loop Remains Paramount: While AI offers significant alpha generation, especially in speeding up processes and uncovering new market angles, "human in the loop" safeguards with kill switches and robust governance are non-negotiable. Yet, training humans to effectively intervene remains a key challenge.

Modeling Opaque Markets: Her PhD research explored multi-agent simulations for OTC bond markets, revealing how diverse agents and thoughtful regulation are crucial for market robustness, challenging conventional wisdom that more players or more rules automatically lead to better outcomes.

From Back-Test to Live Markets: Moving AI models from theoretical back-tests to live portfolios is fraught with challenges. It's less about finding the "secret sauce" and more about meticulously managing assumptions and real-world market dynamics, where risk and P&L considerations often diverge from academic models.

AI for Societal Good: I shared examples of how market-focused AI is driving "Tikkun Olam" (repairing the world), notably through the democratization of financial market access for retail investors and the rise of private credit funding small and medium enterprises where traditional banking falls short.

Cultivating Future-Ready Talent: For aspiring portfolio managers, programming is now table stakes. The true competitive edge lies in patience, curiosity, diligence, and the ability to problem-solve complex, multi-agent systems—viewing them as a dynamic orchestra rather than a simple formula.

The Future of Finance is Bifurcated: A bold prediction: within five years, human traders may need to justify not using an AI for a trade. Furthermore, financial firms will distinctly choose between being tech builders (high capex) or tech consumers (focused on distribution and client relationships).