When Wall Street Catches Up to Academia

Hi, I'm Matt. Welcome back to AI Street, your weekly brief on AI + finance.

ANALYSIS  
When Wall Street Catches Up to Academia

Wall Street often dismisses academics, but history has shown that professors sometimes outpace the trading floor.

Take 1965, for instance. Michael Milken, then an undergraduate at UC Berkeley, stumbled upon a book that hardly screamed “get-rich-quick”: Corporate Bond Quality and Investor Experience by W. Braddock Hickman.

The book revealed that markets consistently overestimated the default risk of what would later be called “junk bonds.” Milken, of course, went on to pioneer the high-yield bond market.

A more recent study from University of Pennsylvania professor Daniel Taylor hit close to home for me. His research exposed how an SEC rule—meant to shield executives from accusations of insider trading—was being exploited.

I covered SEC enforcement for years and actually knew about the so-called “10b5-1 plans,” yet I missed how executives could “schedule” trades mere days in advance. Taylor’s work led to reforms like mandatory cooling-off periods.

A New Frontier: AI’s Academic Roots in Finance

Last week in NYC, I got a similar sense of “something brewing” while attending an academic conference on Large Language Models’ impact on finance. The event, the Association for Computing Machinery’s International Conference on AI in Finance, brought together academics and industry professionals.

Demand for the LLM workshop was so high that the organizers expanded capacity twice and ended up hosting over 300 attendees.

Here are some ideas that stood out:

Three Likely Trends in AI and Finance

1. Analyst Coverage Expansion

• Now: Analysts typically cover a limited set of companies due to manual processes.

• Soon: LLMs will automate repetitive tasks.

• Result: Analysts could cover 2-3x more companies without sacrificing quality, potentially boosting liquidity.

2. Trading Signal Testing

• Now: Testing novel data ideas is limited by time and cost constraints.

• Soon: LLMs will expand capacity for testing a broader range of signals.

• Result: Novel signals (e.g., behavioral trends among executives) could open up new trading strategies.

3. Data Integration

• Now: Structured and unstructured data (like earnings reports and market data) are analyzed separately.

• Soon: LLMs will bridge this gap, integrating text and quantitative data.

• Result: Holistic insights for decision-making by combining qualitative insights (e.g., sentiment analysis) with quantitative metrics.

Looking Ahead

History has shown academia sometimes opens up new opportunities for Wall Street. The same could be happening right now in AI.

THANKSGIVING
AI Street will publish on Wednesday next week for Turkey Day 🦃 

I will be spending the next week trying to source a Thanksgiving dinner in Milan. 🇮🇹 

INTERVIEW: Five Minutes With…
Aakarsh Ramchandani, chief strategy officer at RavenPack | Bigdata.com

While in NYC last week, I caught up with Aakarsh Ramchandani at Ravenpack’s office in Tribeca.

RavenPack has been developing natural language processing (NLP) products for traders since the early 2000s and recently launched a new platform, Bigdata.com, to combine LLMs with traditional NLP.

I really enjoyed our conversation about the future of AI prediction, which feels like a step toward the world imagined in Minority Report (2002). Instead of predicting crimes, we’re heading toward AI agents alerting investors to meaningful changes as they emerge.

The key word is "meaningful." The goal is to have AI agents that surface what matters - spotting early warning signs in supplier comments, or identifying how changes in one sector, like energy demand, might ripple through seemingly unrelated industries.

Hope you enjoy our chat!

This interview has been edited for clarity and length.

What about tracking events or price movements—can your platform help explain market behavior?

Absolutely. We've experimented with workflows where you can hover over a price spike on a stock chart, and the system explains the narrative driving that movement. Instead of drowning in unrelated news articles, it delivers a concise reason. For example, it might say, "This spike correlates with a supplier announcing delays that impact the company." Over time, this approach can even help identify recurring patterns and predict how certain narratives might move prices.

We're also exploring how these insights can extend to monitoring entire market sectors or thematic trends. For example, let's say there's a tariff announcement impacting toy manufacturers. Our system can identify not just the direct impact on specific companies but also downstream effects on suppliers, logistics providers, or even retailers. It's about connecting the dots between events and their broader implications.

This capability goes beyond just reporting on what happened—it can actively assist in strategy development. Users can set up alerts to track specific types of events or narratives, enabling them to respond faster and more effectively. For example, an investor could set a tracker for "earnings downgrades due to supply chain disruptions" and receive a real-time summary whenever this narrative surfaces. It's a way to systematically manage market noise and focus on actionable intelligence.

Read the full interview here

ICYMI
Recent Interviews:

• PhD candidate Alex Kim on executive tone in earnings calls

• MDOTM’s Peter Zangari, PhD, on AI For Portfolio Management

• Arta’s Chirag Yagnik on AI-powered wealth management

• Finster’s Sid Jayakumar on AI agents for Wall Street

• Sov.ai's Dr. Derek Snow on AI for fundamental Investors

• Bain’s Richard Lichtenstein on AI adoption in private equity

REGULATION
Fed's Powell: AI Productivity Gains Still to Come

Federal Reserve Chair Jerome Powell said last week that while traditional AI is widely used by companies, generative AI adoption remains nascent, including among banks.

BIG PICTURE

Powell’s comments align with what I’ve heard from many professionals—despite the generative AI hype, actual implementation remains in its early stages.

FSB Warns on AI Risks in Finance

The Financial Stability Board (FSB) warns of systemic risks from AI adoption, highlighting a shift from focusing on opportunities to addressing potential threats.

Key Trends:

• AI spending in finance to hit $400B by 2027, up from $166B in 2023.

• 8 of the 10 largest U.S. banks now integrate AI.

• Preference for specialized models over large generative AI.

• Focus remains on internal efficiency rather than customer-facing apps.

Main Risks:

1. Over-reliance on a few providers (vendor concentration).

2. Market correlation from similar models/data.

3. Enhanced cyber threats due to AI’s capabilities.

4. Model risk and data quality concerns.

BIG PICTURE

Yet another report from regulators on the potential risks of AI and the need for uniform disclosures. I’m unaware of any concerted effort at the global level to remedy these concerns.

MARKET MOVERS
Nvidia’s AI Chip Sales Soar, but Slower Growth Worries Investors

Nvidia shares fluctuated Thursday morning in New York trading. The world’s most valuable company reported a 94% year-over-year sales jump to $35.08 billion, driven by strong GPU demand. However, growth showed signs of slowing compared to prior quarters.

AI Stock Movers Through Nov. 20

Here’s a snapshot of the week’s top stock movers:

- Applied Materials (AM): -6.7%; Shares declined as earnings missed expectations. Mgmt identified AI as a key industry catalyst, emphasizing the need for innovation in device architecture.

- Walmart (WMT): +2.0%; Delivered 3Q24 revenue that beat consensus and issued an upbeat FY25 outlook. Mgmt highlighted GenAI’s role in enhancing the product catalog and mentioned ongoing development of a personal shopping assistant.

- Keysight Technologies (KEYS): +5.2%; Surpassed 4Q24 EPS expectations and provided strong FY25 guidance. Mgmt credited AI-driven momentum in the wireline business, with orders exceeding $1B this year, largely tied to major infra upgrades for hyperscalers.

- Vertiv Holdings (VRT): +10.4%; Shares rose following an investor day and tech showcase at SC24. Positive investor sentiment stemmed from (1) the company reaffirming 2024 targets and providing a 2025 sales outlook above consensus, and (2) raising its long-term sales growth and profit margin targets through 2029.

Data provided by Linq Alpha, an AI copilot for hedge funds. 

FUNDRAISING
Cardo AI Raises $15M For Private Credit Tech

Cardo AI, a fintech platform focused on asset-based finance and private credit, has raised $15 million in Series A funding from investors including Blackstone Innovations Investments, FINTOP Capital and JAM FINTOP. The company helps manage over $40 billion through its technology platforms.

BIG PICTURE

Generative AI is unlocking insights from PDFs, contracts, and social media, helping streamline private credit deals.

MORE TOP NEWS
BlackRock’s AI Searches for Stock Signals

When IR Met AI: How the Technology Is Shaping Earnings-Day Prep

Generative AI is being increasingly used by investor-relations departments to predict and ready answers to analyst questions on earnings calls, refine word choice in prepared remarks and more. WSJ