E388: The Future of Investing: AI, Expert Networks, and Information Alpha

Episode Summary

Executive Summary: The episode argues that modern investors are overwhelmed by abundant data but lack conviction, and that the solution is not more generic AI, but decision-grade AI grounded in trusted, differentiated sources like expert calls, channel checks, filings, and transcripts. The conversation explains how AlphaSense combines proprietary content with AI to surface cited, verifiable insights that improve diligence, thesis-testing, and investment decisions across public and private markets.

Main Topics: Data overload vs. conviction (Priority: 5/5): The discussion opens with the idea that investors and consumers alike face too much information and not enough clarity. More data has not created conviction; it has made it harder to sift signal from noise. Decision-grade AI and hallucination risk (Priority: 5/5): The speakers contrast generic LLMs with grounded AI, arguing that high-stakes investing requires source-cited outputs built from reliable content rather than broad training data that can be wrong or overconfident. Expert calls as differentiated research (Priority: 5/5): Expert calls are presented as a way to access messy, ground-level intelligence that can’t be found in public filings alone. Their evolution from informal networks to transcript libraries and AI-driven search is central to the thesis. Channel checks and meta-analysis (Priority: 4/5): Channel checks are described as breadth-oriented research used ahead of earnings, while AI can synthesize call sentiment over time and compare signals across multiple sources to reveal patterns. Qualitative inputs behind quantitative models (Priority: 4/5): The episode emphasizes that quantitative outputs in models often originate from qualitative observations gathered through interviews with suppliers, customers, competitors, and former employees. Conviction, thesis-testing, and investment discipline (Priority: 4/5): A recurring theme is that strong investors seek disconfirming evidence, not just confirmation. Conviction helps investors hold through volatility and avoid fear-based decisions. Changing role of analysts and future workflows (Priority: 3/5): The analyst role is shifting from data-crunching to verification, workflow design, and prompt/agent orchestration, with AI taking over much of the manual synthesis work.

Key Arguments: Investors are drowning in data because information is abundant, but conviction comes from interpretation, not volume. Generic AI can be overconfident and may produce incorrect answers, which is dangerous in high-stakes investing. Decision-grade AI must be grounded in trusted source data and able to cite exactly where each fact came from. Differentiated data—not generic public information—is where real edge still exists. Expert calls provide ground-level intelligence that can validate or challenge company narratives. A library of many expert calls is more valuable than any single call because signal emerges through aggregation and cross-checking. Channel checks are useful for breadth and pre-earnings ecosystem mapping, while expert calls are best for deep diligence. All quantitative data in models begins as qualitative insight gathered from people in the market. The best investors seek to disprove their thesis, not confirm it, because that process reduces downside mistakes and increases conviction. AI will increasingly shift analysts away from manual work toward verifying sources and building automated workflows. Private markets are especially suited to this approach because information asymmetry is even greater than in public markets.

Data Points: Expert transcript library size: 250,000+ transcripts - The platform’s library is described as a large repository of differentiated expert-call transcripts. General AI error rate floor: 3% to 25% - A Gartner study is referenced to illustrate that general AI systems can still be meaningfully wrong. Minimum error rate cited: 3% - Used to emphasize that even a small base error rate can undermine trust in investment contexts. Private market asymmetry: Even greater than public markets - The speaker argues private markets suffer from more severe information gaps and scattered data. Historical large-fund spend on channel checks: Millions of dollars per year - Household-name funds are said to have historically spent heavily on channel-check processes. Bitcoin example appreciation: 16x - Used as an illustration of how conviction changes behavior when an asset moves strongly. Bitcoin example decline: 20% - Used to show how weak conviction can trigger fear-based selling after a pullback. Human analysis time split: 80% research / 20% decision-making - The speaker suggests AI will flip this ratio by automating more of the research and synthesis burden. Future analyst time split: 20% research / more time on thesis testing - AlphaSense’s vision is that analysts will spend less time gathering data and more time deciding and stress-testing ideas.

Pivotal Quotes: "“Data isn't the thing. I've had this recurring, I guess, toxic relationship with AI where I ask a very important question.”" — Speaker 1: Introduces the core problem: abundance of data and overconfident AI do not automatically create conviction. "“Decision-grade AI that’s grounded in reliable source data, not just general training data.”" — Speaker 2: Defines the standard required for investing and other high-stakes decisions. "“Tell me what I'm missing, tell me why I'm wrong, disprove this thesis that I have.”" — Speaker 2: Explains the best practice of using expert calls and channel checks to test, not confirm, an investment thesis.

Implications: Investors will increasingly need trusted, cited AI over generic chatbots. Competitive advantage shifts toward proprietary data, thesis-testing workflows, and faster synthesis of qualitative signals into conviction.