AI-Native Firms Turn Workflows Into Execution as Top 10% Hit 8.3x Typical AI Usage
Updated
Updated · OpenAI · Sep 1
AI-Native Firms Turn Workflows Into Execution as Top 10% Hit 8.3x Typical AI Usage
3 articles · Updated · OpenAI · Sep 1
Summary
OpenAI said the biggest shift in enterprise AI is from assistance to execution, with frontier firms now generating 8.3 times as many output tokens per active user as typical companies, up from 2.6 times in January.
That gap reflects how leading companies wire agents into company context and tools, delegate more substantive tasks, and turn successful processes into repeatable operating capabilities.
Basis cut first-day onboarding to 30 minutes from two hours by turning a demonstrated HR process into a reusable Codex skill that handles setup and common questions while humans manage exceptions.
Clay uses a persistent subagent for each account to refresh deal context overnight and produce daily priority actions, saving roughly an hour of inbox triage and keeping evidence tied to recommendations.
Exa Labs pushes further into bounded execution—agents monitor integration opportunities, draft pull requests, run tests and prepare updates—while OpenAI says scaling this model requires clear KPIs, permissions, evidence and human review.