Like 10,000 Interns: How ChatGPT Can Liberate Investment Analysts

by Jonathan Berkow, Director of Data Scienceā€”Equities, AllianceBernstein

ChatGPT is generating excitement about the power of artificial intelligence (AI) to reshape business. For investment firms, AI can help execute many menial functions to free up analysts for deeper research dives, armed with more information than they could ever process alone.

AI has become a hot topic since OpenAI launched ChatGPT last year. Academic and practitioner articles and videos have mushroomed, showing how ChatGPT, which is derived from large language models (LLMs), can be used to analyze stocks and predict markets. Is it coming for your job?

Not so fast. AI isnā€™t going to replace human analysts anytime soon. While ChatGPT reflects major advances in AI, we donā€™t think it can replace human securities analysis or fundamental research aimed at developing a long-term outlook on a companyā€™s business prospects and equity return potential. Yet as companies across industries scramble to benefit from the AI revolution, investment firms will discover new ways to unlock efficiencies.

Processing Mountains of Information

When deployed correctly, ChatGPT and LLMs can help analysts do their jobs better. Equity analysts are often overwhelmed by the sheer volume of material that needs to be processed to gain insight on a company. News reports, official filings and earnings calls generate an endless stream of information. This forces investment teams to focus deep research on a limited number of higher-priority holdings or strong investment candidates.

But what if an analyst had 10,000 highly competent interns at her side? This would be a revolutionary development in processing information that could sharpen investing capabilities. We believe LLM-driven tools such as ChatGPT could perform mundane, time-consuming tasks for investment professionals that currently exhaust disproportionate resourcesā€”or donā€™t get done at all. New LLM technologies have advanced dramatically in recent years and are easily accessible for anyone with basic programming skills. So, once we identify logistical bottlenecks that impede analysts, we can deploy AI creatively to create real efficiencies.

Covering More Earnings Calls and Filings

Equity analysts typically cover dozens of companies in a sector, while also monitoring competitors and potential investment candidates. During earnings season, itā€™s impossible to participate in every hour-long company call. And there simply isnā€™t enough time to dial in to every competitorā€™s call.

Most analysts would never hand over earnings calls of a portfolioā€™s largest-position holdings to an intern. Yet with an unlimited number of interns, an analyst could selectively farm out other earnings calls or conferences that canā€™t be attended to the LLMs, which can produce refined summaries. Weā€™ve tested this on earnings calls and various documents, including ABā€™s recent earnings call, and were impressed by the quality and accuracy of the summaries generated.

Analysts are always curious about themes or questions that may arise outside the coverage universe, for example, issues related to inflation, supply chains or regulation. The LLMā€™s virtual interns can search and extract themes like these automatically and send out alerts.

LLMs can also be used to identify significant changes to text in corporate filings. These documents are often hundreds of pages long and most of the text doesnā€™t change from quarter to quarter. LLMs can comb through these filings to identify changes that may signal a substantive shift in a companyā€™s business or strategy and require further investigation.

Summarizing Meetings, Notes and Internal Text

Portfolio managers and analysts participate in countless meetings and Zoom calls. Sometimes, analysts jot down messy notes and then hand them over to an intern to rewrite.

ChatGPT can condense Zoom transcripts and clean up raw notes into defined user-friendly formats, such as complete sentences or bulleted take-aways or key points. ChatGPT can also be used to extract themes from a longer, unfocused text, helping analysts instantly make sense of a rambling conversation. Polished notes are more digestible for a broader non-expert audience and can be automatically shared with team members and stored in research management systems.

With this type of in-house database, LLMs could be used to generate proprietary insights. In our view, here lies the true power of AI for investment firms. In the future, we may be able to direct ChatGPT to process and interpret an internal collection of analyst notes, for instance, to generate a quick synopsis of the firmā€™s bear case or bull case for specific companies. This shows how Q&A can become a valuable tool for chief investment officers when monitoring investment theses or when automating quarterly commentaries for clients.

Challenges and Risksā€”Hallucinations and Security

ChatGPT is still a nascent technology. Though it is fast improving, it still has a worrying tendency to hallucinateā€”or invent a response that may sound plausible but is totally fabricated. These false positives may be stated with persuasive conviction. Analysts must review ChatGPTā€™s output and verify its accuracy, just as we would work with a motivated-but-inexperienced intern, who has access to unsourced information from the internet.

Data security issues warrant attention, too. ChatGPT and other AI models must be incorporated into a firmā€™s internal IT environment to ensure security and to safeguard intellectual property. Copyright and sourcing are concerns as well, as we consider how to cite machine-generated information for clarity and compliance.

Cultural Shift to Foster Productivity

Beyond the technical issues, ChatGPT poses a cultural challenge. Data-science teams must actively promote the benefits and show tangible examples. Analysts need to discover what they can automate, get comfortable with the technology and understand its limitations. And given ChatGPTā€™s most famous innovationā€”the ability to process commands in natural, human conversationā€”finding the right prompts (prompt engineering) to get high-quality output is critical for success. Crossing these hurdles will allow teams to embrace the technology.

Once they do, we think ChatGPT and LLMs will be a game changer for firms that adopt them strategically and systematically. Itā€™s important not to overstate the power of AI to conduct research or select securities. Instead, we should identify the tasks that could provide the most effective productivity payoffs, as if a team had access to thousands of interns. By combining the human brainpower of investment analysts with the processing power of ChatGPT, we can upgrade an AI arsenal, improve research efficiency and ultimately make better investment decisions.


The views expressed herein do not constitute research, investment advice or trade recommendations and do not necessarily represent the views of all AB portfolio-management teams. Views are subject to change over time.

About the Authors

Jonathan BerkowJonathan Berkow

Jonathan Berkow is a Senior Vice President and the Director of Data Science in the Equities division at AB. He leads the research and adoption of alternative data in equity research and systematic strategies. Prior to joining the firm in 2018, Berkow was a systematic portfolio manager and researcher at hedge funds Element Capital Management and Kepos Capital. He started his career at Goldman Sachs Asset Management, where he managed quantitative research and was a portfolio manager for global equity portfolios. Berkow's research has spanned equities and macro asset classes. He holds a BS in economics from the Massachusetts Institute of Technology. Location: New York


Previous Article

FactSet: S&P 500 Firms with Global Reach See Q1 Revenue Dip

Next Article

Bank Problems Arenā€™t Over, But Itā€™s Not 2008

Related Posts
Subscribe to notifications
Watch. Listen. Read. Raise your average.