Using AI to Find Investing Stories with Perscient Co-Founder Ben Hunt

Episode Summary

Executive Summary: Barry Ritholtz interviews Ben Hunt about how narrative, not just price data, drives markets, politics, and corporate behavior. Hunt explains how Perscient uses AI and large-scale text analysis to detect rising and dormant storylines across global media, arguing that central banks, CEOs, and politicians now use words as tools to shape outcomes. He also warns about regime shifts in immigration, capital flows, credit, and U.S. influence.

Main Topics: From academia to markets: how Hunt’s background shaped his lens (Priority: 5/5): Hunt explains that his Harvard political science training, game theory, and narrative theory gave him a framework for seeing markets as strategic, story-driven systems rather than purely numerical ones. Narratives as market drivers (Priority: 5/5): The conversation centers on the idea that markets move on stories—especially when a narrative enters the 'discovery phase' and becomes widely adopted by investors, media, and CEOs. How Perscient uses AI to measure stories at scale (Priority: 5/5): Hunt describes Perscient’s method: ingesting global news and public text, constraining AI with human-defined context, and identifying semantic signatures of recurring narratives. Central banks, CEOs, and the power of forward guidance (Priority: 4/5): Hunt argues that post-2008 central banking pioneered the use of language as policy, and that corporations copied this approach by using storytelling to command higher valuations. Regime shifts in politics, immigration, and U.S. power (Priority: 4/5): He says narratives around immigration, tariffs, and America’s global role have changed materially, with implications for elections, capital flight, and the end of Pax Americana. Credit risk and the next major narrative (Priority: 4/5): Hunt flags private credit and related financial stress as an emerging narrative that is intensifying across financial media and could become systemically important. Limits, risks, and responsible use of AI (Priority: 4/5): He stresses that AI must be tightly constrained by human context engineering; open-ended prompting leads to hallucination, while structured inputs can produce useful signal.

Key Arguments: Markets are not driven only by structured data; words, stories, and narratives increasingly determine asset prices, valuations, and policy expectations. The 2008-09 crisis and the Fed’s shift to forward guidance taught markets that language itself can be a policy instrument. Perscient’s value is not predicting the future, but identifying which narratives are rising, fading, or re-emerging in real time. AI should be treated as a constrained linguistic calculator, not an autonomous analyst; human-designed context and output checking are essential. Storytelling has become more powerful because of 24/7 media, social media, and smartphones, which amplify narratives continuously. Political and market narratives often reflect a finite set of recurring story archetypes that can be detected and tracked systematically. The current U.S. policy environment appears to be weakening some long-standing narratives about immigration, trade, and America’s global leadership. A growing narrative around private credit and financial stress may be a significant early warning signal for broader credit issues.

Data Points: Perscient launch year: 2018 - Hunt says he and his partner started Perscient in 2018 to research and track financial narratives. Years of academic work before transition: About 10 years plus grad school - Hunt describes academia as a long way station before entrepreneurship and finance. SPX performance in 2008: Down 37% - Referenced as a benchmark while describing their long-short fund’s strong crisis performance. Fund performance in 2008: Up 20-something net - Hunt says the fund was up materially during 2008 while the S&P fell sharply. Crisis recovery period: March 2009 onward - He says their fund flatlined after the March 2009 bottom and missed the rebound. QE1 size: About $800 billion - Mentioned in discussion of the Federal Reserve’s large-scale asset purchases. Tokens processed: Over 200 billion tokens - Perscient has processed hundreds of billions of language tokens in recent months. Language coverage: About a dozen market-relevant languages - He says only roughly a dozen languages matter materially for markets, including Chinese, Japanese, and European languages. Signal improvement after AI rebuild: Over 100x - Hunt says rebuilding with modern AI expanded their signal capture dramatically versus earlier models. Possible population trend: 2025 could be the first year U.S. population decreases - He cites this as a striking demographic implication of lower legal immigration and deportations. Political polling swing on economy: GOP +14% to -4% - Hunt cites a Gallup poll shift over two years as evidence of changing economic sentiment. Temporal horizon for midterms: About 12 months - He says narrative changes now may cash out in the next midterm elections.

Pivotal Quotes: "You do not let it out, you constrain it. Dramatically." — Ben Hunt: On using AI for narrative analysis and why open-ended prompting fails. "A multiple is a narrative, a multiple is a story." — Ben Hunt: On how CEO storytelling affects valuation and investor perception. "I'm not predicting, I'm observing." — Ben Hunt: On Perscient’s role as a nowcasting and narrative-tracking tool rather than a forecasting system.

Implications: Investors, policymakers, and executives should monitor narrative shifts as leading indicators of behavior, valuation, and regime change. AI is most useful when constrained to reveal story dynamics, not when asked to guess. Future alpha may come from detecting which stories are entering the discovery phase first.