OpenAI's new B2B Signals data shows frontier firms pulling away

Firms at the 95th percentile of AI usage now consume 3.5 times as much "intelligence" per worker as typical firms, up from 2x in April 2025, according to new OpenAI research. OpenAI attributes the widening gap not to sheer activity but to depth: message volume explains only 36% of the frontier advantage, while the majority comes from richer, more complex use of AI.

The finding comes from B2B Signals, a business extension of OpenAI's Signals product that the company introduced today. It provides a recurring measure of how AI is diffusing across businesses, drawing on privacy-preserving, aggregated usage signals from enterprise use of OpenAI products. OpenAI says the data shows how deeply AI is used inside firms, which tools and tasks correlate with frontier adoption, and where business use cases are broadening across industries, products, and functions. All analyses are based on de-identified, aggregated enterprise usage data; message content was classified using automated systems, and no OpenAI employee reviewed individual customer data.

The compounding frontier advantage

For many enterprises, the first phase of AI adoption was about access: how many seats were deployed and whether employees were experimenting. That still matters, OpenAI writes, "but access is no longer the differentiator." The new differentiator is how much work employees actually delegate to AI.

OpenAI uses tokens generated as a proxy for intelligence demanded. Tokens do not directly measure business value, the company notes, but they capture how much work employees ask AI to do, making them a useful proxy for depth of use.

The distinction matters for how enterprises benchmark themselves. "Typical firms are using AI to answer questions; frontier firms are using it to help execute complex work," the report states. Workers at the frontier ask AI to take on more complex tasks, provide richer context, and generate more substantive outputs. For leaders, OpenAI argues, the question is shifting from how many people have access or how often they use AI to where AI is deepening workflows and changing how teams operate.

The stakes are significant. If depth of use — not seat counts — drives business value, then conventional adoption metrics understate how quickly the leading edge is compounding its advantage over the rest of the market.

Agentic workflows mark the frontier

The frontier is also moving toward delegation, and the gap is widest in advanced and agentic tools. Codex shows the largest divergence: frontier firms send 16 times as many Codex messages per worker as typical firms. ChatGPT Agent, Apps in ChatGPT, Deep Research, and GPTs show similar directional patterns, suggesting frontier firms are better at adopting tools that help workers code, delegate multi-step tasks, apply company context, and conduct more complex research.

OpenAI frames this as a structural shift. As AI systems become more capable of using tools, working across files and codebases, and completing longer-horizon tasks, enterprises will need to adapt to delegating meaningful work to AI agents. The firms moving first are building the operating muscle to use AI not just as a faster interface but as a way to redesign work from the ground up.

Cisco offers the report's most concrete production numbers. The company uses Codex to speed up complex software work across a large enterprise engineering organization. In production workflows, Codex helped reduce build times by about 20%, save more than 1,500 engineering hours per month, and increase defect-resolution throughput by 10-15x. As Cisco's team put it, the biggest gains came when they treated Codex as "part of the team."

Broad use, but increasingly specialized

AI is also moving into production workflows across the business. Companies are deploying API use cases across in-app assistants, coding and developer tools, and customer support — places where AI becomes part of products, services, and internal systems rather than a standalone tool.

Usage patterns are splitting along functional lines. AI use remains broadest in writing and communication, but function-specific usage is growing: IT and Security teams concentrate queries heavily in how-to and procedural guidance, Software Development and Data Science teams show high coding usage, and Finance teams use AI for analysis and calculation. OpenAI reads this as AI moving beyond general productivity into work tied to each function's core responsibilities.

Notably, OpenAI says there is no single AI adoption leaderboard. Some industries lead in broad ChatGPT adoption, others in Codex use, API intensity, or message intensity. That means organizations have multiple entry points: scale access, deepen usage, adopt agentic tools, or build AI directly into products and systems.

Travelers Insurance illustrates the product-embedded path. Its AI Claim Assistant, built with OpenAI, guides customers through first notice of loss, answers policy questions, gathers the information needed to start a claim, and creates claims directly inside Travelers' systems. Travelers expects the assistant to handle approximately 100,000 first notice of loss calls in its first year.

What leaders do differently

OpenAI cautions that the gap between frontier and typical firms should not be read as a fixed divide. Many organizations are still early in moving from broad access to deeper, more integrated AI use. The value of the frontier data, the company argues, is that it shows which practices help firms build momentum over time.

One of the clearest signals is education and learning, where the task-level frontier advantage is largest. That suggests leading firms use AI not only to complete work but to help employees build the skills, habits, and confidence needed to use AI well.

OpenAI lays out a playbook for closing the gap: measure depth of use, build governance that enables production use, treat enablement as core infrastructure, identify frontier teams and scale their impact, and move beyond chat toward delegated work with agents.

The company positions B2B Signals as an answer to a growing leadership problem: enterprise AI is evolving quickly, and leaders need clear data to understand what translates adoption into business value. This first release focuses on depth of use, agentic workflows, and emerging patterns across industries and functions. Future updates will track progress on these measures and adapt the signals as enterprise AI evolves — giving enterprises, and OpenAI's own sales pitch, a recurring benchmark for who is actually converting intelligence into output.