Frontier enterprises now use 3.5x as much AI intelligence per worker as typical firms, up from 2x last year. The jump stems more from depth than volume; only 36% of the advantage comes from higher message counts. Workers at the frontier push AI on broader, complex tasks, feeding richer context and getting substantive outputs.

Agentic workflows mark the frontier. Codex, for instance, sees frontier firms sending 16x as many messages per worker compared to typical firms, while ChatGPT Agent and GPTs follow similar patterns. These tools let teams code, delegate multi‑step tasks, apply company context, and conduct deeper research, turning AI from a question‑answerer into a collaborator that executes complex work.

OpenAI’s new B2B Signals tracks how deeply AI diffuses across businesses, measuring token usage as a proxy for intelligence demand and flagging advanced tools that drive the frontier gap. The signals show that depth, governance, and a shift from chat to delegated agent work are the levers that move an organization from broad access to meaningful productivity gains.

Cisco’s engineering group reports a 20% reduction in build times and a 10‑15x lift in defect‑resolution throughput after treating Codex as part of the team, while Travelers Insurance plans an AI Claim Assistant that could handle 100,000 first‑notice calls annually. These cases illustrate how firms can translate deeper AI use into tangible operational improvements.