The corporate conversation around artificial intelligence has officially outgrown the basic seat-deployment phase. For the last few years, executive boards focused almost entirely on accessibility: purchasing enterprise software seats, spinning up internal sandboxes, and tracking how many employees logged in each week. But on May 6, 2026, OpenAI launched B2B Signals, a comprehensive, privacy-preserving analytical suite tracking aggregated enterprise usage data, and the findings confirm a quiet, massive divergence in the marketplace.
The data proves that simple access is no longer a competitive differentiator. Instead, a deep gulf has opened up between typical organizations and “frontier firms,” those operating in the 95th percentile of AI utilization. These leading companies aren’t just sending more messages; they are consuming 3.5x more computational intelligence per worker than the average firm, creating an operational advantage that changes how day-to-day work gets done.
This systemic acceleration relies entirely on trust, a challenge explored in How ChatGPT Learns While Protecting Your Privacy. The development of frontier artificial intelligence relies on a massive structural contradiction: for an AI model to become highly capable at complex real-world tasks like parsing code, analyzing data, or executing multi-step research, it has to ingest a massive footprint of human knowledge. Yet, the more data a system absorbs, the higher the risk of swallowing private, individual information. Frontier firms are pulling ahead precisely because they have figured out how to balance this tension, giving workers the confidence to feed deep context into the system while maintaining rigid boundaries around proprietary data.
Volume vs. Depth: How Elite Workers Use the Technology
When looking at why frontier firms are running away from the competition, the instinctive assumption is that their workers are simply typing faster or querying the system more often. The data completely refutes this. Raw message volume accounts for a mere 36% of the gap between frontier firms and typical enterprises. The remaining 64% of the divide comes down to absolute depth.
By tracking generated tokens as a proxy for the total volume of intelligence demanded, the research reveals a fundamental behavioral split. Average companies use AI as an advanced search engine or an autocomplete tool mostly to answer quick questions or clean up copy. Frontier firms use AI to execute structural labor. They provide hyper-detailed contextual frameworks, upload entire system files, and expect the machine to return highly substantive, production-ready outputs. Typical firms use AI to answer questions; frontier firms use it to execute complex work.
The Rise of Agentic Delegation
The clearest marker of a frontier organization is the shift from chat-based assistance to pure delegation. This pattern becomes obvious when analyzing how elite teams interact with advanced developer and research tools.
For instance, frontier firms generate a staggering 16x more Codex messages per worker than typical companies. They are no longer using AI just to generate individual lines of code; they are using it to automate end-to-end integration testing and optimize system architecture. Elite teams also show a massive concentration of activity inside specialized agents, deep research tools, and custom internal configurations, routinely offloading long-horizon tasks like comprehensive market intelligence mapping or multi-file data reconciliation to autonomous software systems.
We are already seeing this operational muscle pay off at scale. Cisco integrated Codex directly into its engineering pipelines, treating the agent as a core team member rather than an external utility tool. The result was a 20% reduction in software build times, over 1,500 engineering hours saved per month, and a massive 10-to-15x increase in defect-resolution throughput. Similarly, Travelers Insurance deployed an AI Claim Assistant capable of guiding customers through the initial notice of loss and spinning up claims directly inside internal databases, a workflow projected to handle 100,000 calls in its first year alone.
Functional Specialization Across the Enterprise
The B2B Signals report emphasizes that there is no longer a single, generic AI leaderboard by industry. Instead, different sectors are developing hyper-specialized habits tied directly to their core corporate responsibilities.
Information technology and corporate security teams concentrate their queries heavily on procedural guidance and infrastructure patching to accelerate threat response. Software development and data science teams focus their utilization on heavy code generation and agentic testing to slash build times and automate quality assurance cycles. Meanwhile, finance and operations teams utilize the technology for deep analytics and dynamic calculation, parsing complex ledgers and forecasting market variables.
Playbook: How to Move Your Organization to the Frontier
The divide between elite AI adopters and typical firms is not a permanent historical sentence. If your enterprise is currently stuck using AI as a basic writing assistant, you can consciously build the operational muscle required to catch up.
First, stop measuring clicks and start measuring token depth. Move away from tracking vanity metrics like daily active users and start auditing the complexity of the prompts your teams are submitting. Look for departments that are uploading complete datasets, using advanced tools, and demanding substantive, multi-step execution.
Second, treat enablement as core infrastructure. The single largest task-level gap between frontier firms and everyone else is in education and learning. Elite firms use AI to teach their employees how to build better habits, run complex prompts, and verify agentic outputs. Don’t just hand people an API key; build a continuous internal training program.
Third, identify frontier outliers and scale them. Look closely at your internal departments. You likely have a small, isolated team of developers, analysts, or marketers who are quietly utilizing agentic tools at a 16x rate. Study their workflows, map out their prompt frameworks, and institutionalize their habits across the rest of the enterprise.
Where is the biggest bottleneck in your organization keeping your teams stuck in simple chat interactions instead of delegating complete projects to autonomous agents?
