OpenAI has recently unveiled its State of Enterprise AI 2025 Report, a comprehensive study based on a survey of 9,000 workers across approximately 100 enterprises, supplemented by usage data from over 1 million business customers. While the report offers a broad overview of AI adoption, many of its headline figures, such as "message counts," can be misleading "vanity metrics." A simple message to "rewrite this paragraph" carries vastly different value than a complex request to "build a financial model," yet both contribute equally to a message count.
For founders and B2B executives seeking actionable insights, it's crucial to look beyond these superficial statistics. Here are six key metrics that truly reveal the evolving landscape of enterprise AI:
1. 320x Increase in Reasoning Token Consumption Per Organization Year-over-Year
This metric represents a significant shift: a 320-fold increase in the consumption of reasoning tokens per organization year-over-year. This is the real indicator of AI's deepening integration, far more telling than mere message counts. It signifies that enterprises are moving beyond simple AI chats, actively deploying complex, multi-step reasoning tasks in production environments. Such a dramatic surge within 12 months suggests that AI has transitioned from an experimental tool to a critical piece of business infrastructure.
2. 40-60 Minutes Saved Per Worker Per Day
The report indicates that workers are saving an average of 40 to 60 minutes daily, with data scientists and engineers reporting even higher gains of 60 to 80 minutes. While these are self-reported figures, their directional significance is undeniable. Scaled across an organization, these savings translate into substantial productivity gains. For instance, a 500-person company with 50% AI adoption could reclaim 1,000 to 1,500 hours per week. At a fully-loaded cost of $75 per hour, this equates to an annual productivity gain of $4 million to $6 million – a compelling return on investment (ROI) for CFOs.
3. 75% of Workers Gain New Capabilities
Perhaps one of the most transformative findings is that 75% of workers can now perform tasks they previously couldn't. This isn't merely about doing existing work faster; it's about unlocking entirely new capabilities. The report highlights non-technical workers engaging in activities like code review, spreadsheet automation, technical troubleshooting, and even custom tool development. This shift signifies a significant expansion of the Total Addressable Market (TAM) for "technical work," as AI redefines who is capable of performing such tasks.
4. 36% Growth in Coding Activity from Non-Engineering Functions in Six Months
Over a mere six-month period, the report notes a 36% increase in coding activity originating from non-engineering departments. This trend illustrates a powerful "bottoms-up" adoption of AI, with teams like sales operations writing Python scripts, marketing departments building automations, and finance teams developing custom analysis tools. This rapid, organic growth indicates that AI-driven transformation is not a distant, multi-year initiative but an immediate, pervasive reality, often occurring independently of formal IT sanctioning.
5. Custom GPTs See 19x Growth Year-to-Date, Comprising 20% of Enterprise Workflows
Custom GPTs are emerging as a pivotal force, experiencing a 19-fold growth year-to-date and now accounting for 20% of enterprise workflows. This trend, exemplified by organizations like BBVA running over 4,000 custom GPTs in production, highlights the rise of "internal AI apps." Enterprises are actively codifying institutional knowledge—including playbooks, processes, and tribal wisdom—into reusable AI workflows. Companies embracing this are building significant, compounding advantages, while those lagging behind risk their top talent developing similar solutions in "shadow IT" environments.
6. 25% of Enterprises Haven't Connected AI to Company Data
Despite the rapid adoption, a significant hurdle remains: 25% of enterprises, even those paying for ChatGPT Enterprise, have yet to connect their AI tools to proprietary company data. This is akin to owning a high-performance car but leaving the fuel tank empty. Further, the report reveals that 19% of monthly active users have never utilized AI for data analysis, and 14% have never engaged with reasoning models. This stark disconnect underscores that the primary challenge isn't access to AI technology, but rather the organizational capability and readiness to effectively integrate and leverage it. This represents the true frontier for AI implementation.
The Big Takeaway: Organizational Readiness, Not AI Models, is the New Enterprise Constraint
The overarching conclusion of the report, though somewhat understated, is profound:
"The primary constraints for organizations are no longer model performance or tooling, but rather organizational readiness."
With OpenAI consistently rolling out new capabilities—often every three days—the sophistication of AI models is no longer the bottleneck. Instead, the real challenge for companies lies in their ability to effectively integrate AI, train their workforce, and adapt existing workflows. This paradigm shift presents a clear opportunity for B2B founders. The market demand isn't for "more AI features," but for solutions that empower enterprises to genuinely deploy and adopt AI effectively. This includes robust integration tools, comprehensive training programs, strategic change management, and intelligent workflow redesign. These are the areas where true value is created and captured in the evolving AI landscape.
