Lenny's PodcastMay 31, 20261h 19m

The most rational take on AI you’ll hear this year

TL;DR

AI is huge, but not mystical: Benedict Evans says AI is comparable to the internet or smartphones, not the Industrial Revolution, and compares today to 1997 when the technology was clearly important but most products and business models had not been invented yet.
The job apocalypse story is too simplistic: Evans calls out doomer claims that companies will buy ChatGPT and fire everyone two weeks later, arguing that every major technology wave automates tasks, creates new work, and unfolds through slow enterprise adoption cycles that often take 3 to 10 years.
Consultants are benefiting from AI, not being erased by it: OpenAI, Anthropic, PE firms, and enterprises are putting more money into forward-deployed engineers and services because redesigning workflows, integrating systems, and retraining organizations is a messy human project, not a one-click software install.
The real question is task versus job: AI may write code, draft slides, or produce a rough McKinsey-style deck, but Evans argues the valuable part of many professions is deciding what to build, navigating politics, talking to customers, and figuring out what problem actually matters.
Foundation models may not capture most of the value: His thesis is that if models remain interchangeable and lack strong network effects, companies like OpenAI and Anthropic could end up looking more like cloud utilities or telecom providers, while more durable value accrues in apps, distribution, and product layers above them.
The practical advice is brutally simple: Do not reject AI out of moral disgust and call it a strategy. Get immersed, test tools, understand what they can and cannot do, and show employers you know how the work is changing.

The Breakdown

AI is as big a deal as the internet or mobile, and only that big, Benedict Evans argues, which is exactly why both the hype and the job apocalypse takes miss the point. His core message is calming but not complacent: this will change everything over the next decade, but most people are still in the 1997 phase where the tools are real, uneven, and far from fully built out.