HAI Seminar: Industry Conversation with Instacart

HAI Seminar: Industry Conversation with Instacart

by Stanford HAI · View the channel on PSI

Humanities, Social Sciences & Thought AI & Society UBDigital and Information technology: general topicsUBJDigital and information technologies: social and ethical aspects
🎙 Stanford HAI 👥 35K 📅 September 30, 2026 ⏱ 72 min 👁 461 📄 original study 🧭 2026-10-06
Available in: English (current) Français

Keywords

AI adoptionpeer networksproductivityagentic AIorganizational roles

Summary

The seminar presents two complementary research perspectives on AI adoption in firms. Alexandr Lenk from Instacart Economics describes a study on how AI adoption spreads via peer effects within engineering and data-science teams. Using employee-level data from a major undisclosed e-commerce company, they identify ‘AI pioneers’ as early heavy users of agentic AI and find that their arrival significantly increases usage among teammates, accounting for about 23% of overall AI usage growth in 2025. This leads to improvements in productivity outcomes such as experiment creation and ticket resolution, with effects varying by tenure and role. The study emphasizes that for complex tasks, pioneers need both AI enthusiasm and domain expertise. Arvind Karunakaran from Stanford MS&E then discusses organizational consequences after adoption, drawing on fieldwork in a law firm using generative AI. He contrasts how managerial framing of the tool as a productivity improver versus as an opportunity for expansion influences whether employees use AI for substitution or higher-order tasks. He argues that substantive value comes from role expansion rather than acceleration. The discussion covers where these views converge and diverge, and how to assess enterprise value of AI.

188 words

Critical Evaluation

Value of the Information & Strength of the Argument

The presentations offer valuable empirical evidence on AI diffusion, moving beyond task-level proxies to firm-level outcomes. Lenk’s staggered difference-in-differences design provides credible causal estimates of peer effects, and the heterogeneity analysis reveals important nuances about the complementarity between human skill and AI knowledge. Karunakaran’s qualitative work complements this by showing how managerial framing shapes whether employees use AI for substitution or expansion, and convincingly argues that expansion is where true value lies. The argumentation is solid, though some claims remain preliminary, and the law firm study is based on a single site. The discussion touches on limitations but overall the reasoning is sound and well-supported.

Scientific Rigor, Source Quality, Title Accuracy

The presentation maintains rigorous scientific standards, with clear empirical methods and acknowledgment of limitations. Lenk’s study uses internal data from a major tech firm and follows best practices in causal inference, though the company remains anonymous, limiting external validation. Karunakaran’s multi-method approach includes interviews and observations, but findings are based on one firm. Both are grounded in existing literature, and the description accurately frames the content. The title precisely reflects the seminar’s industry conversation and organizational perspective. The video includes a Q&A that adds depth. No public comments are provided, so community feedback is not analyzed.

216 words

Title / Content Match

The title accurately reflects the seminar's focus on industry perspectives (Instacart) and organizational lens, matching its content.

Quality & Reliability

8/10

Presentation of empirical research from two studies: Instacart's quantitative analysis of AI diffusion and a qualitative field study of a law firm. Both are methodologically sound but not yet peer-reviewed, with some limitations such as undisclosed company and single-site focus.

Chapters

Contribution & Novelties

This video provides novel perspectives on AI diffusion in firms, combining quantitative peer effects analysis with organizational qualitative insights. The Instacart study highlights the role of pioneers in agentic AI adoption, an understudied aspect, and demonstrates measurable productivity gains. Karunakaran’s research underscores the importance of managerial framing in using AI for task expansion rather than substitution.

Pour aller plus loin :

99 words

Radar Profile

The radar profile shows high scores in information quantity and quality, moderate technical level, and high reliability, indicating a strong academic seminar with substantial empirical content.

Reliability 8/10