STRATEGYOther
More AI, More Humans
21MAY
Dan Shipper, CEO of media-and-AI-tools company Every, publishes "After Automation". His team grew from 4 to 30 humans while automating everything.
The strongest counter to AI displacement, written from inside the experiment. Headcount went from 4 to 30 while Every automated everything it could.
Shipper's core idea: AI commoditizes yesterday's skills, so demand for human experts goes up. Each new model just shifts what humans work on. Same job, different layer. His framer-vs-frame distinction: benchmarks measure capability inside frames humans pick.
Counters Amodei's "half of white-collar jobs go" and Griffin's "high-skill jobs automated." Read this before you cut headcount based on benchmark hype.
⚡ Why this matters
- The strongest data-backed counter to AI displacement, written by someone running the experiment.
- "Framer vs frame" gives PMs and execs a real tool for headcount decisions.
🔍 What happened
- May 21, 2026. Every CEO Dan Shipper publishes "After Automation." Viral on May 24.
- Every automated everything: Codex, Claude Code, agent employees, customer service via Fin.
- Headcount went from 4 to 30 since GPT-3 launched.
- Fin handled 65% of weekly support conversations. Closed 81 of 202 without humans.
- 95% of Shipper's email handled by AI. He still reviews every message.
💬 Smart takes
- Shipper: "AI commoditizes yesterday's expertise. That creates demand for what's different. Demand for what's different is demand for human experts."
- Shipper on benchmarks: "The score tells us how well the model operates inside a frame we supplied. It does not tell us the model has become us."
- Dario Amodei (counterpoint): AI could wipe out half of entry-level white-collar jobs.
- Ken Griffin, Citadel (counterpoint): "Extraordinarily high-skilled jobs being automated by agentic AI."
- Skeptic: Every benefits from a humans-in-the-loop business model. Sample of one. Zeno's paradox assumes humans always set the next frame; if AGI sets its own, the argument breaks.
🧭 Where this goes
- "Framer vs frame" enters mainstream AI strategy vocabulary within 60 days.
- Cursor, Anthropic, Linear, Notion, Vercel publish their own headcount-vs-automation data within 12 months.
- AI labs face pressure to release internal employment data as a credibility marker.
- The two-mode framing (agent employees vs human-agent collaboration) becomes standard procurement vocab.
🎯 Implication
- For PMs: audit your top 5 automate-able roles using framer vs frame. If framer-level work is real, redesign the role. Don't cut it.
- For execs: stop building "AI replaces N% of role X" forecasts. Start building "binding constraint migrates to Y" forecasts.