We Automated Everything With AI and Tripled Our Headcount

TL;DR

• 1

  AI makes expert output cheap, not finished work — Dan argues that models turn yesterday’s competence in coding, writing, and design into low-cost raw material, which creates a glut of “close but not quite right” work that still needs experts to finish.

• 2

  Every automated aggressively and still tripled headcount — At Every, where people use Claude Code, Codex, and agents constantly, the company grew from 4 people in the GPT-3 era to about 30 today, with more hiring underway.

• 3

  The further an agent gets from a human, the less valuable it becomes — Their practical experience is that AI works best in a “human-agent sandwich,” where experts set up the system, non-experts use the tools, and humans review, steer, and correct.

• 4

  Big layoffs blamed on AI may really be strategy, bloat, or bad implementation — Using the ClickUp layoffs as a prompt, Dan says struggling or poorly run companies will reorganize and cite AI, but that’s different from proving AI can replace whole functions cleanly.

• 5

  Customers still resist machine-only experiences — In research on customer service forums and Reddit, Dan found examples of companies firing support staff after AI rollouts, then scrambling to rehire because users wanted a human and the automation underperformed.

• 6

  ‘Ride the models’ is the practical survival strategy — His bottom line is simple: if you keep learning new models and folding them into your workflow, you’ll likely be fine—and may end up doing more ambitious, fulfilling work than before.