BlackRock Study Tests AI Agents for Stock Picks
Hey, it’s Matt. This week on AI Street:
🍎 AI Street in the Big Apple Next Month!
🪨 BlackRock Develops AI Agent System: Study
📉 AI Adoption Issues Prompt Share Drop
NYC
AI Street Meetup
I’ll be back in New York next month to attend Cornell’s Future of Finance & AI conference on Sept 19. It has a great agenda exploring many of the topics I’ve been covering and includes some speakers I’ve previously interviewed like Snowflake’s Jon Regenstein and Citadel’s Tharsis Souza. Plus, many more speakers I plan to interview.
I’m only in town for a few days, and there are many folks I want to connect with so I’m thinking of organizing an informal AI Street meetup for drinks that Thursday night (Sept. 18.) Reply to this email if you’re interested and I’ll figure out a spot.
RESEARCH
BlackRock Researchers Develop AI Agent System for Stock Picks: Study
LLMs are often treated as one-size-fits-all tools: they can suggest what to make for dinner from a fridge photo or plan a trip to the Marshall Islands.
But AI struggles with tasks requiring mathematical precision. Because it’s probabilistic, the same prompt can yield different answers — and even the model builders can’t fully explain why.
Instead of a single model, BlackRock built three specialized “agents” that mimic different analyst roles, according to a new study.
Fundamental Agent — parses 10-Ks and earnings reports
Sentiment Agent — reviews news and analyst ratings
Valuation Agent — studies prices, volatility, and volumes
Each agent analyzes a stock independently, then enters a round-robin debate. Disagreements are argued until the agents reach consensus on whether to BUY or SELL — a process designed to mimic an investment committee.
The system runs on Microsoft's AutoGen framework using GPT-4o, with custom tools for each agent: document parsing for 10-Ks, news summarization, and volatility calculators.
The agents' recommendations change based on risk tolerance settings. The same volatile stock might get a SELL from a risk-averse agent but a BUY from a risk-neutral one analyzing identical data.
