Using AI as a Research Assistant
What AI chatbots and tools are genuinely useful for in investment research, the ways they get things wrong, and a workflow that keeps every fact anchored to a primary source.
The short version
AI tools can save time explaining terms, summarizing long filings, finding where a topic is discussed, and suggesting questions to ask. They can also state false information with complete confidence, which NIST calls confabulation, and they may draw on data that's outdated, incomplete, or deliberately manipulated. Regulators warn against relying solely on AI-generated information for investment decisions. Use AI to explain and to find, then verify every fact in the primary source, ask it for the strongest counterargument, and make the decision yourself. Never ask it what to buy.
By the end of this lesson, you'll be able to
- Name the research tasks AI tools handle well
- Explain why AI output can be confidently wrong
- Follow a workflow that verifies AI output against primary sources
Why this matters
AI chatbots can read a hundred-page annual report in seconds and answer questions about it in plain English. That's genuinely useful for research. They can also invent a number, misread a table, or sound certain about something false. Used well, AI is a fast research assistant. Used carelessly, it's an unreliable secondary source that sounds like an expert.
What AI is good for
- Explaining unfamiliar terms and concepts, like deferred revenue or interest coverage.
- Summarizing long documents, so you know which sections to read closely.
- Finding where a topic is discussed, such as every mention of a major customer in a 10-K.
- Suggesting questions you hadn't thought to ask.
- Arguing the other side, which helps counter the confirmation bias from earlier in this part.
How AI gets things wrong
The US National Institute of Standards and Technology lists "confabulation" as a key risk of generative AI: the production of confidently stated but erroneous or false content, often called hallucinations, which can mislead users.1 A joint alert from the SEC, NASAA, and FINRA adds that AI-generated information may rely on data that's inaccurate, incomplete, or outdated, or that was spread on purpose to manipulate a stock's price. Even with accurate inputs, the output can be faulty or completely made up. And conversations with chatbots can lead people to make emotional or impulsive investment decisions.2 The alert's advice: don't rely solely on AI-generated information, confirm the underlying sources, and review multiple sources before deciding.2
Suppose an AI tool summarized STU Corp's annual report like this. The tool, the summary, and STU Corp are all hypothetical.
| The AI said | The annual report says | Verdict |
|---|---|---|
| Revenue grew 14% to $400 million | $350 million to $400 million | Correct |
| Long-term debt is $190 million | Long-term debt is $90 million. $190 million is shareholders' equity | Wrong line |
| The largest customer is 13% of revenue | 30% of revenue | Wrong number |
| "The company has no significant risks" | Risk Factors names customer concentration, one factory, one main supplier, and a patent expiring | False |
The summary got the headline right and three important details wrong, in the same confident tone throughout. Each error would change your view of the company. None would be caught without opening the filing.
The rest of this lesson is for members
Still to come in this lesson: A workflow that keeps you anchored, the key takeaways, and 2 quiz questions with explanations.
Membership opens every lesson in the curriculum, the capstone projects, the interactive tools, and the members-only channels in the Hyperion Entities Discord. The first lesson of every module stays free.
Go deeper
- Primary vs. Secondary SourcesPart 4: Research and Valuation
- Fact, Interpretation, and Confirmation BiasPart 4: Research and Valuation
- Reading a Real 10-K, Section by Section coming soonResearch Workshop (depth track)
Notes and sources
- National Institute of Standards and Technology. NIST AI 600-1, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile (July 2024). Checked 2026-10-01. Back to text
- Investor.gov, U.S. Securities and Exchange Commission. Artificial Intelligence (AI) and Investment Fraud: Investor Alert (January 25, 2024). Checked 2026-10-01. Back to text