Frontier enterprises now use 3.5x more AI per worker

OpenAI's B2B Signals research marks a shift from measuring AI adoption by access and seat count to measuring depth of use. The finding that frontier firms use 3.5x as much intelligence per worker as typical firms, up from 2x in April 2025, suggests the advantage is compounding over time. Message volume explains only 36% of the gap, indicating that richer, more complex usage—not just more messages—is what separates leaders from the pack.

The importance of advanced tools is evident in the Codex data: frontier firms send 16x as many messages per worker as typical firms. Cisco's reported results—20% faster build times, over 1,500 engineering hours saved monthly, and 10-15x higher defect-resolution throughput—offer a concrete picture of what this depth can deliver. The report also notes that AI use is broadening into specialized functions like IT, finance, and software development.

Whether the frontier gap narrows depends on whether typical firms can adopt the practices that leaders use: measuring depth, building governance for production use, treating enablement as infrastructure, and moving from chat toward delegated agentic work. The report frames education and learning as a key signal, suggesting that investing in employee skills may be central to closing the divide.