OpenAI’s latest Enterprise Signals report indicates that leading enterprise AI users are generating 8.3 times more output tokens per active user than typical firms, highlighting a shift from assistance to execution. The article examines how startups Basis, Clay, and Exa Labs operationalize this trend by integrating agents into core workflows such as employee onboarding, account management, and developer integrations.
- Basis automates accounting firm onboarding using Codex, reducing first-day setup time from two hours to 30 minutes through reusable skills.
- Clay employs persistent subagents for sales accounts that update deal context overnight, saving engineers roughly one hour of inbox triage daily.
- Exa Labs utilizes agents to monitor integration opportunities, create pull requests, and prepare updates, reducing handoffs across research and engineering teams.
The report outlines a framework for enterprise leaders to design trustworthy AI workflows by defining clear outcomes, establishing human review points, and making experimentation visible.