Investing in Knowledge Automation:

How Generative AI Is Reshaping Market Analysis

Generative AI, powered by large language models (LLMs) like ChatGPT and Google Gemini, is increasingly being used to interpret complex data. These tools can generate detailed responses when trained on extensive datasets, offering a powerful resource for investors navigating a sea of corporate information.

 

 

AI in Investment Research

Researchers Long Chen and Yi Cao from George Mason University are studying how individual investors can leverage LLMs to gain market insights. Their recent paper, co-authored with Jennifer Wu Tucker and Chi Wan, explores how AI can identify “peer firms,” companies that operate within the same market or industry.

 

 

Understanding Peer Firms

Yi Cao uses the real estate market to explain the concept of peer firms. Just as home prices are compared with nearby properties, a company’s value is influenced by similar businesses in its sector. Traditionally, identifying these peers requires extensive effort. However, LLMs can simplify the process by analyzing large datasets and suggesting relevant comparisons.

 

 

Automating Data Analysis

The study suggests that LLMs can handle data gathering and analysis tasks typically performed by experts. These models can generate peer firm lists that are nearly as accurate as those created by human analysts.

“The real benefit lies in using all available data efficiently,” Cao notes. For this study, the researchers used Google’s Bard (now called Gemini), due to its extensive pre-training and large parameter base.

 

 

Ensuring Objectivity and Avoiding Bias

To avoid “look-ahead bias,” where future data might skew results, the team restricted the LLM’s dataset to individual years between 1981 and 2023. They focused on large, publicly traded companies, since data for smaller firms is limited. On average, the LLM identified seven peer firms per focal company, aligning with SEC guidelines.

 

 

Comparing AI With Human Analysts

The AI-generated lists were compared with those compiled by human experts for 40 major software companies. The overlap averaged slightly above 40%, indicating a strong correlation. The study also evaluated AI results against other systems like Standard Industrial Classification (SIC) codes and Text-based Network Industry Classification (TNIC).

Interestingly, AI and TNIC had substantial overlap. Furthermore, peers identified by the LLM had stock returns more closely aligned with the focus firm than those found using SIC or TNIC.

 

 

Limitations and Future Potential

Despite its advantages, the LLM didn’t outperform TNIC in identifying peers for mid-sized companies. This shows that while AI is powerful, it is not infallible. Cao emphasizes the technology’s efficiency and accessibility, especially for individual investors who may lack resources for traditional analysis.

“Generative AI tools are incredibly efficient, low-cost, and widely available,” says Cao. “These are essential qualities for retail investors.”

 

 

Caution and Ethical Considerations

Chen highlights the risks associated with relying on AI for investment decisions. While the technology is transformative, it should be used with caution. “AI can’t be held accountable for its suggestions,” he warns. Users must understand both the benefits and limitations before integrating AI into their strategies.

 

 

Conclusion

Chen concludes, “We must embrace new technology while recognizing its imperfections. The rapid pace of AI development means current tools may soon be outdated. Our findings may only represent a baseline for what’s possible.”

As generative AI becomes more sophisticated, its role in investment research will grow. Investors who learn to use these tools effectively will gain a significant edge in navigating today’s data-rich markets.