---
id: "lib-2026-04-02-coase-vs-claude-and-the-future-of-the-firm"
type: "library"
title: "Coase vs. Claude and The Future of the Firm"
lang: "EN"
org: "IMD (Howard Yu)"
source_type: "Practitioner post"
date: "2026-04-02"
week: "2026-W14"
url: "https://nosov.ai/library/lib-2026-04-02-coase-vs-claude-and-the-future-of-the-firm/"
original_url: "https://howardyu.substack.com/p/coase-vs-claude-and-the-future-of"
tier: "B"
tier_name: "Useful, limited"
tier_reason: "IMD professor Howard Yu applies Coase's transaction-cost theory to AI agents in a clear, well-told essay; the argument is useful framing but rests on anecdotes and relayed figures, with no evidence of its own."
summary_basis: "full-text"
original_minutes: "14"
tags: ["theory of the firm", "agentic ai", "organization design"]
---

# Coase vs. Claude and The Future of the Firm

Tier B (Useful, limited): IMD professor Howard Yu applies Coase's transaction-cost theory to AI agents in a clear, well-told essay; the argument is useful framing but rests on anecdotes and relayed figures, with no evidence of its own.

Summary based on the full text.

## In short

Howard Yu (IMD Business School) argues that AI agents push Coase's 1937 theory of the firm toward a limit he calls the "Coasean Singularity". On the supply side, agents cut the transaction costs that justified hierarchy. On the demand side, mass markets are splitting into a barbell of cheap-algorithmic and premium-personal. Together these shrink the minimum viable firm and push large companies toward platform forms such as Haier's micro-enterprises. The saved copy is the LinkedIn repost dated 8 April; the essay first ran on Substack on 2 April.

## Key numbers

- roughly $830 billion: market value shed by software and services stocks over six trading days in early February 2026
- 75%: knowledge workers using AI at work, per Microsoft and LinkedIn
- 78%: AI users bringing their own tools to work, per the same Microsoft and LinkedIn work
- over 4,000: self-managing micro-enterprises Haier reorganised its staff into

## What it reports

- Coase: a firm grows until organising one more transaction inside costs as much as buying it on the market. AI lowers coordination costs on both sides of that line.
- Sangeet Paul Choudary (author of Reshuffle), interviewed: the main payoff of AI is cutting the cost of turning one team's output into another's input, not speeding up single tasks. He frames this around the "atomic unit of work".
- Cases on shrinking units of work: Zara (collection) vs Shein (single design decision); Adobe (file) vs Figma (element), ending in Adobe's blocked $20 billion acquisition attempt.
- Middle-market decline examples: Red Lobster bankruptcy in May 2024 with more than 130 locations closed; TGI Fridays shut 185 locations, Denny's up to 150.
- Claim that a software company could need 500 customers and 2 people instead of 10,000 customers and 50 staff.
- Haier: Zhang Ruimin reorganised the company into over 4,000 micro-enterprises with their own P&L; GE Appliances' share, near 2% before the 2016 acquisition, is said to have risen tenfold.
- Closing concern: a micro-enterprise economy may work only for the relentless and become a tournament with few winners.

## How they got it

Essay drawing on history (Coase's 1931 US trip), one interview (Choudary), press reports and company case material. Figures are relayed from Microsoft and LinkedIn, market data and Haier accounts without citations in the saved copy.

## Limits

- No original data; the barbell thesis is illustrated by selected examples, not measured.
- The software start-up economics (500 customers, 2 people) are asserted, not sourced.
- Haier's results are presented from company-friendly accounts and contain no discussion of failed micro-enterprises beyond the general statement that most bets lose.
- Saved copy is a LinkedIn repost with an appended note section that is not part of the original essay.

## Why it matters for companies

Transaction-cost thinking gives a practical lens for operating-model choices: for each hand-off between teams, ask whether agents now make it cheap enough to source outside or to reorganise into smaller units. The essay's point about bring-your-own AI tools is worth taking seriously, since it reframes shadow IT as a demand signal from the most diligent staff. The Haier case shows that smaller units still need a shared platform for brand, trust and capital, which is where central functions keep a role.

Original: https://howardyu.substack.com/p/coase-vs-claude-and-the-future-of