Agent orchestrators (single interfaces that spawn specialized subagents) are now able to perform full analytic workflows — research, data gathering, plotting, layout, and citation — in minutes rather than hours. That changes who does analytical labor, how outputs are audited, and what counts as authorship or provenance for public‑facing research.
— If agentic AI can routinely produce publishable analysis, it will reshape labor markets, research norms, accountability rules, and regulatory approaches to provenance and liability.
Ethan Mollick
2026.10.20
100% relevant
Ethan Mollick’s use of Claude Projects (Opus 5.5) to generate, check, and iterate the charts and text for this article — including the agent catching the author's math error — exemplifies the shift.
← Back to all ideas