AI Adoption Deepens as Frontier Firms Scale Codex Workflows

OpenAI’s newly introduced B2B Signals report sheds light on how “frontier” enterprises—those in the top 5% of AI adoption—are extending their competitive edge by embedding artificial intelligence deeper into workflows and leveraging advanced tools like Codex at unprecedented scale. According to the research, these firms now use 3.5 times more AI-generated intelligence per worker compared to typical firms, up from 2x a year ago.

The findings emphasize that AI’s value for these leaders lies not just in broad application but in depth and complexity. Frontier firms aren’t merely answering questions with AI—they’re delegating complex, multi-step tasks to agents. For example, these organizations are sending 16x more Codex-powered messages per worker than their peers, a clear indicator of how they’re embracing agentic workflows to redesign how work gets done.

From Access to Depth: The Next Phase of AI

The report outlines a paradigm shift in AI adoption. While the early days of enterprise AI focused on tools and seat deployment, the competitive advantage now lies in how deeply AI is integrated into processes. The research uses tokens generated as a proxy to measure this depth, indicating how much work AI performs for employees. Typical firms use AI for simple interactions, but frontier firms are pushing the envelope, tasking AI with richer, more substantive outputs.

Cisco, a leader among these frontier firms, provides a compelling example. By treating Codex as an integral part of its engineering team, Cisco reduced software build times by 20%, saved over 1,500 engineering hours per month, and increased defect-resolution throughput by 10-15x. This approach highlights how advanced AI tools, when used effectively, can drive transformative results in production workflows.

Agentic Workflows Signal Maturity

One of the most striking trends in the report is the rise of agentic workflows. These workflows involve delegating meaningful work to AI agents capable of operating across files, codebases, and long-horizon tasks. Tools like Codex, ChatGPT Agent, and GPTs are enabling frontier firms to shift from basic chat-based assistance to more robust, context-aware systems.

Travelers Insurance exemplifies this trend with its AI Claim Assistant, built using OpenAI technology. The tool automates first notice of loss calls, policy queries, and claim creation, handling an estimated 100,000 calls annually. This level of integration demonstrates how AI is moving beyond support roles into core business operations.

Specialized AI Use Cases Gain Traction

While writing and communication remain the most widespread applications of AI, the report points to increasing specialization across functions. IT and Security teams are using AI for procedural guidance, Software Development teams leverage it for coding tasks, and Finance departments rely on it for analysis and calculations. This diversification indicates that AI is becoming a fundamental enabler of productivity tailored to each department’s unique needs.

What Sets Leaders Apart

OpenAI’s research highlights several actions that help firms transition toward frontier-level AI adoption. These include measuring depth of use, building governance frameworks for production workflows, scaling successful practices, and moving beyond general chat tools to fully delegated AI agents. Education and enablement also play a critical role, as leading firms actively invest in training employees to maximize AI’s potential.

The B2B Signals report aims to provide ongoing insights into enterprise AI adoption. Future updates will track how these trends evolve as businesses continue to refine their strategies for embedding AI into their operations.

Image source: Shutterstock