The AI paradox: More automation, more humans, more work

Dan Shipper, CEO and founder of Every, returns to discuss his predictions for how work will change over the next year as AI agents and coding tools like Codex and Claude Cowork become everyday work surfaces. The conversation covers the shift toward company-wide AI agents in Slack, SaaS tools running inside agent environments, the continuing need for humans to supervise AI, and which roles are best positioned for the coming shift.

01Key learnings

• Keep using the newest models and agent tools for your real work, and experiment on your own time if your employer restricts access, to stay ahead of the curve.
• Expect most company AI to start as one shared agent that everyone can talk to, likely in Slack, rather than many personal agents, since agents need a human who cares about them.
• Design products so agents can be users too: make them easy to drive via CLI and browser-friendly, because the customer's own AI tokens may power the product's usage.
• Treat SaaS as resilient rather than doomed; Shipper argues agents increase the number of software users and usage, so SaaS demand may grow.
• Don't assume AI automation removes humans; Shipper sees ongoing need for people to supervise, garden, and judge agent output, similar to managing staff.
• Product managers with strong product sense can become much more powerful now that AI handles building, so invest in judgment about what to build and why.
• Full-stack designers who can ship pull requests will be especially valuable, since AI-generated work tends to look generic without human creative direction.
• Benchmarks can overstate AI autonomy because they measure well-framed tasks; the human work of framing problems and judging results remains essential.

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