Why Big Tech wants more economists

Episode Summary

Executive Summary: The episode examines why AI companies are rapidly hiring economists and building research teams, and whether this creates conflicts of interest. Through interviews with economists at Ramp and Anthropic, and FT data journalist John Burn-Murdoch, the show argues that access to proprietary AI data is valuable but incomplete, and that transparent methods, external scrutiny, and mixed methods are essential to judge AI’s true effects on jobs, productivity, and society.

Main Topics: AI labs’ growing demand for economists (Priority: 5/5): The episode opens on the surge in hiring and collaboration between AI companies and economists, driven by the need to explain AI’s economic impact to governments, investors, and the public. Ramp’s economist role and proprietary business-spend data (Priority: 5/5): Ara Karazian explains how Ramp uses aggregated card, bill-pay, and invoicing data to study AI adoption and market trends, while also strengthening the company’s public profile. Evidence on AI and employment (Priority: 5/5): Ramp’s research found that firms using AI intensely saw headcount rise by about 10% over two years, while low-intensity users saw no significant change, with debate over causality versus selection effects. Research quality, transparency, and conflicts of interest (Priority: 5/5): The discussion centers on whether company-funded research is credible, how much labs sanitize data, and whether outside peer review and public methodologies can mitigate bias. Anthropic’s approach to economic research (Priority: 4/5): John Burn-Murdoch describes Anthropic’s and other labs’ efforts to publish methods, release data, and fund independent work, while also noting that the work may still be used to shape narratives. Limits of usage data and the need for mixed methods (Priority: 4/5): The guests argue that logs of what people do with AI are not enough; interviews and surveys are needed to understand effects on work intensity, bottlenecks, job quality, and real-world productivity. Broader societal concerns and global inequality (Priority: 4/5): Anthropic’s research suggests AI adoption is higher in rich countries, raising concerns that gains could accrue unevenly and widen global economic divides.

Key Arguments: Demand for economists at AI firms is extremely high because the technology is new, politically salient, and poorly captured by standard productivity statistics. Company economists can access granular proprietary data unavailable to governments or academics, making their research potentially very valuable. At the same time, AI companies have incentives to frame findings favorably, so their research should be treated cautiously and checked by external economists. Ramp argues its economists are not simply doing PR; they are using anonymized spend data to help businesses understand market trends and make better decisions. Ramp’s study suggests AI adoption correlates with higher headcount growth, but the relationship may reflect early-adopting firms being inherently faster growing. John Burn-Murdoch argues the best studies combine platform data with surveys and interviews, because task usage alone misses job intensity, frustration, and bottlenecks. Anthropic says its research aims to answer big open questions about AI’s economic consequences, publish methodology, and pre-commit to reporting results even if they are negative. The episode concludes that much AI-lab research is credible and useful, but it also serves a broader communications strategy to shape public and policy debate.

Data Points: Demand for economists at AI companies: 10/10 - Ara Karazian says demand for economists in AI is at the highest level he has seen. AI adoption and headcount growth: 10% increase - Ramp’s paper found companies that use AI intensely saw headcount rise over two years after adoption. Low-intensity AI users: No statistically significant change - Ramp found no significant headcount effect for firms using AI with low intensity. Influencer marketing spend share: 28% - Karazian said 28% of influencer marketing spend is coming from AI companies and startups. Global adoption analysis: More than 100 countries - Anthropic’s September Economic Index report analyzed cloud use across over 100 countries. Timeframe for AI as a consequential technology: Last 30 years, if not the past 100 years - Peter McCrory described AI as potentially the most consequential technology in decades or a century.

Pivotal Quotes: "If you have a toolbox that is full of very shiny diamond-encrusted hammers, then all of your economic analysis is going to use nails." — John Burn-Murdoch: On the risk that AI labs’ research agendas are constrained by the data and incentives they control. "How do I make Ramp look smart?" — Ara Karazian: On the economist’s role inside Ramp and the branding/credibility function of company research. "Helping society make sense of this fact and what we should be doing to exercise agency over the future is what really motivates me in my work today." — Peter McCrory: On why Anthropic’s chief economist joined the company and how he frames the purpose of its research.

Implications: AI-lab economics research is valuable but incomplete: expect more proprietary data, more narrative shaping, and more scrutiny. Listeners should value transparency, external replication, and mixed methods before drawing conclusions about AI’s effects.