---
id: "contrarian-ai-commoditization"
type: "contrarian-insight"
source_timestamps: ["§ The Business Consequences of Non-Diversity in Agentic Teams"]
tags: ["strategy", "competitive-advantage"]
related: ["claim-uniformity-compresses-differentiation", "quote-competitive-compression", "concept-correlated-ai-errors"]
challenges: "The assumption that simply adopting AI tools automatically confers a competitive business advantage."
speakers: ["Enver Cetin", "Mark Purdy"]
sources: ["agentic"]
sourceVaultSlug: "hbr-seg-agentic"
originDay: 6
articleStem: "hbr-new-28-agent-teams-different-models"
sourceUrl: "https://hbr.org/2026/06/the-strongest-teams-of-ai-agents-will-be-built-using-different-models"
sourceTitle: "The Strongest Teams of AI Agents Will Be Built Using Different Models"
---
# Uniform AI adoption destroys competitive advantage rather than creating it

**Contrarian insight.**

**Conventional wisdom:** Adopting cutting-edge AI is a way to gain a competitive edge.

**The inversion:** If an entire industry (e.g., retail) adopts the *exact same* AI tech stack for pricing and recommendations, the models converge on the same answers. Instead of gaining an edge, firms **quietly price toward the same equilibrium** (see [[quote-competitive-compression]]), compressing differentiation and commoditizing their market positioning — see [[claim-uniformity-compresses-differentiation]]. This is [[concept-correlated-ai-errors]] expressed as *correlated success*.

**Challenges:** The assumption that simply adopting AI tools automatically confers a competitive business advantage.

**Enrichment — steel-man the other side:** A shared model stack does **not automatically** destroy differentiation. Firms can still differentiate through (1) unique proprietary data and feature engineering, (2) different objective functions and constraints (long-term brand equity vs. short-term margin), and (3) distinct human governance and business processes. The algorithmic-collusion literature shows convergence is *possible*, but also that regulators and firms can design algorithms to avoid collusive outcomes. Outcome depends heavily on *how* models are used, not merely *which* model is used.