Andrey Fradkin is an economist whose work applies market-design and digital-economy methods to AI agents as economic actors. He holds an appointment at Boston University and the MIT Initiative on the Digital Economy and, as of late 2025, is employed by Amazon. He is a co-author of the MarketBench working paper (April 2026, with Rohit Krishnan) and the Coasean Singularity? NBER chapter (2025).
Background and affiliations
Fradkin is an assistant professor at Boston University, in the Marketing group of the Questrom School of Business, and a research affiliate of the MIT Initiative on the Digital Economy. He is an associate of the National Bureau of Economic Research; the Coasean Singularity? chapter forms part of the NBER Economics of Transformative AI volume. As of late 2025, per the disclosure in the Coasean Singularity? chapter, he is employed by Amazon.
Research
Fradkin's contributions to the agentic-AI literature come through two foundational sources: the Coasean Singularity? chapter and the MarketBench working paper.
In *The Coasean Singularity?* (NBER chapter, 2025), co-authored with Shahidi, Rusak, Manning, Krishnan, and Horton, Fradkin and co-authors argue that AI agents lower transaction costs in the Coasean sense, shifting the make-or-buy boundary and enabling new market designs.
*MarketBench* (working paper, April 2026), co-authored with Rohit Krishnan, is an empirical benchmark testing AI-agent market-readiness on SWE-bench Lite tasks. The paper presents a formal proof that, under standard assumptions, market-style allocation weakly dominates every non-market alternative for procurement of AI-agent labor, with the dominance strict whenever there is positive probability of allocation-error or wasteful-payment states. Empirically, the paper finds that frontier LLMs are miscalibrated on both success-probability and token-usage self-assessment, and that a self-knowledge-card intervention only modestly closes the gap.
Relationships
- related: Rohit Krishnan — primary collaborator on AI-agent-market work
- related: John J. Horton — Coasean Singularity co-author
- supports: Principal-Agent Problem Applied to AI
- supports: Agent Architecture Patterns
Confidence is high for institutional affiliations and authorship, verified via author cards on the cited papers.