Afford Anything Podcast Summary

Episode: Why AI Misleads Investors and How to Fix It
Host: Paula Pant
Guest: Brian Feroldi, financial educator and author
Date: November 25, 2025

Episode Overview

This episode dives deeply into how artificial intelligence (AI) can both mislead and empower investors. Host Paula Pant speaks with Brian Feroldi—seasoned stock analyst and author—about the dangers of AI "hallucinations," the optimism bias of language models, and how to use AI responsibly as a tool for analyzing investments. The conversation is both practical and philosophical, providing step-by-step strategies for using AI as a "junior financial analyst" while remaining vigilant against its well-known flaws.

Key Discussion Points and Insights

1. Why Should Index Investors Care About AI for Stock Analysis?

• Even for predominately index investors, knowing how to evaluate individual companies has value both for "fun money" investing and general business literacy.
• Learning fundamental analysis improves career and entrepreneurial skills.
• “Learning business valuation, learning company analysis is a highly adaptable and transferable skill set.” —Paula Pant [02:12]

2. Pros and Cons of Using AI for Investment Analysis [03:12]

Pros:

• AI rapidly synthesizes large volumes of data.
• It is effective at summarizing and filtering information for first-pass analysis.

Cons:

• AI frequently "hallucinates" by fabricating facts or overconfidently providing incorrect data.
• Built-in optimism: ChatGPT and similar models tend to err on the side of positivity.
• Reliance on unchecked output is risky since investment decisions require factual accuracy.

“You have to put in strong guardrails in place because of that bias that's built into ChatGPT to be so positive and to, and to essentially make up stuff.”
Brian Feroldi [04:18]

3. Building Effective Guardrails to Reduce AI Hallucinations [08:40]

• Rule #1: Insist that AI only uses original, trusted sources (e.g., SEC filings, company reports, specific financial databases).
• Prompt Design: Explicitly restrict AI to use only the information from accepted sources, and have it cite those sources with direct links.
• Verification: Random spot checks or full verification, especially if considering an actual investment.

“Anything that you can do to restrict the information that AI is pulling from to a source that you trust, again will dramatically increase it.”
Brian Feroldi [09:46]

4. Prompt Engineering: Treating AI Like a Junior Analyst [13:17]

• Assign a clear, specific role to the AI (e.g., “Act as Warren Buffett,” “Act as a forensic accountant”).
• Break the analysis down into explicit, step-by-step instructions.
• Require structured answers formatted as you would expect from a human research assistant.

“You have to think of AI like a junior analyst or a junior intern that is eager to do whatever you tell it to the instant that you tell it to.”
Brian Feroldi [13:43]

5. Sample Prompt Technique and Structure [15:02]

1. Assign a role to the AI.
2. Specify trusted sources (SEC filings, company reports, etc.).
3. Create a sequential, stepwise structure for gathering and presenting information.
4. Demand direct links or references for any cited data.

Example Questions for Company Analysis:

• What does the company do?
• How does it make money?
• Who are its customers?
• Where does it operate?
• How often do customers buy?
• Can it raise prices?
• What happens during a recession?

“The clearer you can be with the instructions you want it to follow, the better the information you will get back.”
Brian Feroldi [18:16]

6. How to Learn the Steps If You’re Not a Stock Analyst [19:22]

• Fundamental analysis can be learned gradually—even if you're new.
• Start with high-level, foundational questions to understand any business you invest in.
• Avoid "vibes investing": Don’t just buy because someone suggested it.

“If you're the type of person that bought a stock because a friend told you about it and you did zero research on it, that's not investing. That is just pure gambling.”
Brian Feroldi [21:47]

7. Live AI Prompt Walkthrough: Analyzing CAVA Group [31:19]

• Demonstrated how an AI, using Brian’s prompt structure, can answer foundational questions about a public company.
• Ensured all numbers provided by the AI were directly linked to original SEC filings for transparency.
• Discussed context for the company’s performance, pricing power, customer base, and recession resilience.

“Kudos to ChatGPT for not making stuff up.”
Brian Feroldi [36:02]

8. Assessing Management and Valuation via AI [49:43]

• Use frameworks like OATS: Ownership, Allocation, Tenure, and Stewardship to evaluate management from filings and public datasets.
• For company valuation, tailor the method to the business stage—avoid using Price/Earnings for high-growth or pre-profit companies.
• Always compare management’s forecasts to actual outcomes.

“I want management teams. That net worth is going to be impacted way more than my net worth is going to be impacted if the stock goes down.”
Brian Feroldi [49:56]

9. Combating AI’s Optimism Bias and Stress-Testing Your Thesis [59:02]

• Don’t delegate judgement to the AI—use it for gathering, not deciding.
• Assign the AI a "devil’s advocate" role (e.g., “Act as a short seller”) to surface potential flaws in an investment thesis.
• Diversification and critical thinking are still necessary.

“By prompting it, by giving it the role of think of a short seller, it's going to by definition, go out and look for everything that it could possibly go wrong.”
Brian Feroldi [60:49]

10. Aligning Investments with Personality & Risk Tolerance [69:17]

• The “best” investment for one person may not be right for another—personality, risk tolerance, and emotional disposition all matter.
• AI can help design processes or assessments to help you discover your personal investing style.

“Mathematically we might know that. We might be able to pull data that shows that. But that doesn't change how people feel. And feelings turn into behaviors that become suboptimal.”
Paula Pant [68:14]

Notable Quotes & Memorable Moments

• On Guardrails:
  “If you just first and foremost insist on the sources that the prompt use, that alone will dramatically increase the quality of information that you get back.” [09:16]

• On Trust and Spot-Checking:
  “I think AI is a wonderful source for a first pass… But if I'm going to go deeper on it, there's other sources that I use.” [10:33]

• On Iterative Prompting:
  “Sometimes building the prompts… have taken me hours upon hours upon hours… But I used AI the entire way to help me build those prompts.” [64:59]

• On Stress Testing:
  “Assigning it the devil’s advocate role… makes a lot of sense because you want to stress test all of your assumptions.” Paula Pant [61:13]

Timestamps for Important Segments

• 03:12 — Pros and cons of using AI in investing
• 08:40 — How to build anti-hallucination guardrails
• 13:17 — Treating AI as a junior analyst
• 15:02 — Building stepwise prompts and assigning roles
• 19:22 — Advice for non-analysts and index investors
• 31:19 — Walkthrough: Using AI to analyze CAVA
• 49:43 — Frameworks for analyzing company management
• 59:02 — Handling optimism bias and devil’s advocate prompts
• 66:06 — Three key takeaways

Three Key Takeaways

1. Treat AI Like a Junior Analyst, Not an Oracle

“The key to good use of AI is constraining AI, limiting it, for example, to only using very trusted sources… and then also giving it detailed step by step instructions.”
[66:06]

2. Demand Transparency from AI: Force It to “Show Its Work”

“Whenever you present with a number, you also have to simultaneously give me the link to the information that you got that number from.”
[67:43]

3. The Best Investment is Personal

“Mathematically… the optimal asset allocation may not fit your personality. Make investment choices that let you sleep at night, not just win on paper.” [69:17]

Resources Mentioned

• Brian’s prompt and YouTube tutorials: longtermmindset.co/ai
• SEC filings for company research
• Paula Pant’s community and newsletter: affordanything.com/community | affordanything.com/newsletter

Final Thoughts

This episode demystifies the use of AI in investment analysis with actionable, nuanced advice. The framework provided helps listeners—regardless of experience level—think critically about both company fundamentals and the tools they use. Most importantly, Brian and Paula emphasize never abdicating responsibility: AI may be a powerful assistant, but human judgment, tailored discipline, and self-awareness remain central to successful investing.