AI Agents in Markets

Hey, it’s Matt. This Week on AI Street:

🎙️ Podcast: Alicia Vidler on agents

📊 New research from Prof. Lopez-Lira

📰 Latest AI + finance news and analysis

Forwarded this? Subscribe here. Join readers from Bloomberg, JPMorgan, BlackRock & more.

PODCAST 🎙️ 

AI Agents in Markets: A Trader’s View with Alicia Vidler

The latest episode of the The Alpha Intelligence Podcast is out. My co-host Francesco Fabozzi and I chat with Alicia Vidler about using AI agents in trading. Drawing on her experience in both markets and academia, Alicia offers a grounded perspective on what LLMs can and can’t do across asset classes.

Alicia’s background spans equity derivatives at Deutsche Bank, macro trading at Merrill Lynch during the 2008 crisis, and co-founding an AI-driven hedge fund in London. She now advises fintechs, and sits on the board of 6Star Capital while completing a PhD on agentic AI in capital markets at UNSW.

🎧️ Listen: YouTube, Spotify

These are great and amazing tools, but at the end of the day, they’re still under a lot of legal regulations globally, and a human has to make a decision.

But what LLMs don't do is they don't ask good questions. They're not designed to ask questions. That's the part where humans will always have an advantage.

Ask an LLM how it made its decision. Ask it the same question five days in a row, and you will get five different answers.

🎧️ Listen: YouTube, Spotify

RESEARCH

Can LLMs Really Trade? New Research from Prof. Lopez-Lira

I spoke with Alejandro Lopez-Lira last week to talk about his latest research. Since our first conversation for AI Street last summer, many of his predictions are already playing out, like advancements in running LLMs locally. He’s written a trio of new papers recently. Here’s the rundown:

Alejandro Lopez-Lira is a finance professor at the University of Florida and author of one of the most downloaded AI-and-investing papers. In 2023, he showed that ChatGPT could outperform traditional sentiment models in predicting stock returns from headlines—without any specialized financial training. That paper, downloaded over 70,000 times, earned him a book deal.

On Trading Applications

Lopez-Lira’s simulated stock market lets LLMs trade against each other. The results: surprisingly competent, but highly dependent on how you prompt them.

“They act like normal traders,” he said. “But they don’t care about money—they care about predicting the next word and being helpful.”

Prompting matters. Ask a model to act like a retail investor and it will make naive forecasts. Prompt it as a professional trader, and it performs much better.

“You have to be careful what you wish for—because you may get it.”

He’s preparing to deploy real-money LLM agents into crypto markets in the next few weeks:

“In S&P 500, $100 doesn’t move the needle. But in a small token, you can actually observe price impact.”

On Thematic and Bottom-Up Investing

AI is changing how investors identify themes.

“Before, you had brute-force methods. Now you can scan 10-Ks or fund prospectuses to find exposure to any topic—like quantum computing in the next five years but not the past five.”

His work with LinqAlpha enables investors to build custom taxonomies and portfolios with almost any lens they care about.

“You can run hypotheticals and scenario simulations at scale,” he said. “And invest in anything you have conviction on—for whatever reason.”

On Model Capabilities

GPT-4o is now, in his view, comparable to “a second- or third-year PhD student” across fields.

“Any task that doesn’t involve physical-world or person-to-person interaction is basically solved.”

Still, hallucinations remain an issue.

“They will make up a source or say they ran code they didn’t. You have to check their work.”

On the Future

He’s not betting that AI will reach human-level intelligence anytime soon—but he is betting that today’s systems will drive major productivity gains:

“Old LLMs were like humans without pen and paper. The new ones are like humans with pen, paper, and computer access.”

NEWS

JPM Credits AI for Sales Boost Amid Market Decline

JPMorgan’s AI tools boosted sales to wealthy clients and managed numerous customer requests during April's market rout, according to the bank's asset and wealth management CEO. (Reuters)

  • "When you have a tool that pre-populates all the data and the movement in real time, while also remembering clients' old investment preferences and helps in tailoring a plan for them quickly, it also allows advisers to do much more" - Mary Erdoes, CEO of Asset & Wealth Management

  • AI is projected to help advisers expand client rosters by 50% in 3-5 years

  • Gross sales increased 20% between 2023-2024 with AI-driven tools

Citadel, WorldQuant, and Others on How They Use AI

At Milken, hedge fund execs said AI boosts productivity but human judgment still drives investment decisions. (Business Insider)

To speed up decision-making, Subramanian (Citadel CTO) said that building intuitive tools like chatbots that can be talked to are important because people generally prefer asking questions naturally. The firm is also hiring data scientists and AI professionals to embed them in the fund's various groups to optimize various investment workflows.

Andreas Kreuz, WorldQuant's deputy CIO, said the firm was using AI to expand the data it can bring into its models since it can restructure data from images and audio.

"What excites us is beyond the low-hanging fruit," Kreuz said.

Citadel’s Griffin Downplays AI Impact on Investing

Citadel CEO Ken Griffin said he doesn't think AI will revolutionize the investment business.

"Do we use it in our investment business? A little bit, a little bit. I can't say it's been game-changing," Griffin said in an interview published on the Stanford Graduate School of Business' YouTube channel.

AI Becoming Table Stakes for Global Financial Planners

Two-thirds of financial planning firms worldwide either use AI or plan to within a year, according to a global survey of 6,200 planners across 24 countries by the Financial Planning Standards Board. Adoption is strongest at the largest and smallest firms. (FPSB)

  • Top Use Cases: Client communications (41%), data collection (33%), and risk profiling (30%) lead the way. AI is also helping with marketing (35%) and onboarding (34%).

  • Sentiment: Half of planners view AI positively; only 8% view it negatively. Most see it as a tool to enhance advice, improve client service, and reach underserved populations.

  • Benefits: 78% say AI improves client service. 60% believe it strengthens advice quality and increases access. Many planners see a need to boost their data analysis skills.

“AI is not just reshaping financial planning—it’s opening the door to advice for people who historically couldn’t afford it,” said FPSB CEO Dante De Gori.

Impact of AI on Financial Planning.pdf969.94 KB • PDF File

Palantir, xAI and TWG Global team on AI offering for FS sector

Data analytics giant Palantir Technologies is teaming up with Elon Musk's xAI and investment firm TWG Global to help financial services firms to embed AI at the core of their organizations. (FinExtra)

  • Instead of relying on per-seat licensing or headcount margin, the group will operate on an outcome-based business model.

WHAT ELSE I’M READING

Vanguard's AI Tool Helps Advisors with Market Notes for Clients (ThinkAdvisor)

Singapore’s OCBC Starts AI Stock adviser (Asian Banking & Finance)

Visa to Give AI 'agents' Your Credit Card (Yahoo)

The Future of AI in Wealth and Asset management (PWM)

How did you like today's newsletter?

Login or Subscribe to participate in polls.

Reply

or to participate.