A rational conversation on where AI is actually going

Benedict Evans, a longtime tech analyst (formerly at a16z and an equity researcher) who now works independently, discusses how AI compares to past platform shifts like the internet and mobile. The conversation covers adoption spread, the consulting and professional-services boom around AI labs, job impacts and Jevons-style demand, where value in the AI stack will accrue, and the growing anti-AI backlash.

01Key learnings

• Treat AI as roughly as big as the internet or mobile, and recognize we are early, like 1997, with most applications not yet built or proven.
• Expect adoption to be very uneven; survey data shows many people are not using AI at all, so don't assume the people in your bubble represent everyone.
• Ask whether a job is a task that can be automated or a deeper outcome; consultants and professionals are hired for judgment, politics, and change management, not just deliverables.
• Use history of price elasticity: cheaper work often expands demand rather than shrinking headcount, as with accountants and spreadsheets.
• Be skeptical that model labs will keep pricing power; if models commoditize, value may shift further up the stack to distribution and applications.
• Distribution is becoming more decisive when products are commodities; incumbents with default placement can win even with adequate products.
• Don't dismiss AI or sit on the sidelines; dive in, learn what it can actually do, and build fluency to stay employable.
• Be wary of sweeping automation-exposure scores for whole professions; you cannot decompose a senior role into automatable percentages, and outcomes are hard to predict.

02Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.