Benedict Evans is a London-based independent technology analyst and writer who publishes a widely-read newsletter at ben-evans.com focused on AI adoption, deployment metrics, market structure, and consumer-tech historical analogues. He is a frequent counterpart to Ben Thompson in the platform-strategy commentariat, generally writing shorter, faster-cadence, more empirical pieces (drawing on charts and survey data) than Thompson's long-form analytical essays.
Background and roles
Role: Independent analyst and writer (2020–present); newsletter at ben-evans.com.
Previous: Partner, Andreessen Horowitz (2014–2020); GroupM (digital strategy); Enders Analysis.
Most of Evans's regular newsletter pieces are treated as supporting sources — quick takes on the news cycle that fold into existing pages — while a handful of his more original-argument essays are tracked as foundational sources with their own summary pages.
Positions and statements
- AI metrics skepticism (AI metrics, 2025) — Evans argues that current AI consumer-adoption surveys (ChatGPT user counts, "have you used AI" polling) systematically overstate engagement because they conflate "tried it" with "regular use." He cites enterprise-rollout failure rates (MIT 95%, NBER ~80% no-impact) as the more reliable signal.
- AI competition framing (How will OpenAI compete?, Feb 2026) — Evans argues that roughly half a dozen organizations ship frontier models of near-equivalent capability, that no known mechanic lets one firm take a lead others could never match, and that competition therefore shifts to brand and distribution — a disadvantage for OpenAI, which unlike Google and Meta has no existing products to act as distribution. He characterizes OpenAI's 800–900 million users as "a mile wide but an inch deep," rejects the company's platform-and-flywheel framing on the grounds that "a 1:1 relationship between capex and revenue is not a flywheel," and names the widget fallacy — the recurring belief that complex products can be abstracted into a simple standard interface — as the reason he doubts agent-protocol standards will produce lock-in. The essay settles on power, defined by his university professor Roger Lovatt as "the ability to make people do something that they don't want to do," as the property OpenAI has yet to demonstrate.
- AI labor / human-in-loop framing (AI, networks and Mechanical Turks, Nov 2025) — Evans argues current AI deployments resemble Mechanical Turk more than autonomous systems, with low-paid humans cleaning up AI outputs in production, and contends the economic surplus from AI is going as much to data-labeling vendors as to model developers.
- GenAI adoption puzzle (GenAI's adoption puzzle, 2025) — Evans applies a productivity-paradox framing in which generative-AI penetration into consumer workflows is high (~30% have tried it) but enterprise impact and revenue lag well behind, comparing the gap to the early-PC era and its productivity-paradox lag of more than ten years.
Across these pieces, Evans is among the empirical-skeptic voices on AI productivity and penetration claims.
Notable connection points
Evans's 2013 essay arguing that the PC market was mostly corporate is cited by Ben Thompson in What Clayton Christensen Got Wrong as the empirical anchor for the consumer-versus-enterprise distinction in disruption theory.
Relationships
- related: Ben Thompson — frequent counterpart in the same analytical space; cite each other regularly
- related: Aggregation Theory — Evans uses Thompson's framework but generally writes shorter, more empirical newsletter pieces rather than original framework-development
- related: AI Labor Disruption — Evans is an empirical-skeptic voice on AI productivity claims
- related: AI Bubble Debate — Evans tends toward empirical-skeptic framing of AI penetration claims
- related: Widget fallacy — named in the February 2026 OpenAI essay