The tools are available to everyone. Membership is company-wide. Training sessions have been conducted. And yet, in offices from Wall Street to Silicon Valley, a widening divide is opening between workers and coworkers who have woven artificial intelligence into their daily work and those who have barely touched it.
The difference is not small. According to a new report from OpenAI that analyzed the usage patterns of more than one million of its business customers, employees in the 95th percentile of AI adoption are sending six times more messages to ChatGate than the average employee at the same companies. For specific tasks, the divide is even more dramatic: frontline workers send 17 times more coding-related messages than their typical peers, and among data analysts, the heaviest users use data analysis tools 16 times more often than the average.
This is not a story about access. This is a story about a new form of workplace stratification emerging in real time—one that could reshape who gets ahead, who gets left behind, and what it means to be a skilled worker in the age of artificial intelligence.
Everyone has the same tools, but not everyone is using them
Perhaps the most surprising finding in the OpenAI report is how low access explains it. ChatGPIT Enterprise is now deployed in more than 7 million workplace seats globally, nine times more than a year ago. The tools are the same for everyone. The capabilities are the same. And yet the usage varies depending on the magnitude.
Among monthly active users – those who have logged in at least once in the last 30 days – 19 percent have never tried a data analytics feature. Fourteen percent have never used reasoning abilities. Twelve percent have never used search. These aren’t obscure features hidden in submenus; Those are the core functionalities that OpenAI highlights as transformative for knowledge work.
The pattern reverses among daily users. Only 3 percent of people who use ChatGPT daily have never attempted data analysis; Only 1 percent have given up reasoning or searching. The implication is clear: the divide is not between those who have access and those who do not, but between those who have made AI a daily habit and those for whom it remains an occasional novelty.
Employees who use more are saving dramatically more time
The OpenAI report shows that AI productivity gains are not distributed equally among all users, but are concentrated among those who use the technology most intensively. Workers who perform about seven different types of tasks – data analysis, coding, image creation, translation, writing, and others – report saving five times as much time as those who use only four. Employees who save more than 10 hours per week consume eight times more AI credits than those who report saving no time at all.
This creates a mixed dynamic. Workers who experiment tend to find more uses extensively. More usage leads to greater productivity gains. Better productivity gains potentially lead to better performance reviews, more interesting assignments, and faster progress – which in turn provides more opportunities and incentives to deepen AI use further.
Seventy-five percent of workers surveyed reported that they are able to complete tasks they previously could not, including programming support, spreadsheet automation and technical troubleshooting. Workers who have adopted these capabilities are expanding the boundaries of their roles. For those who have not, borders may shrink in comparison.
The corporate AI paradox: $40 billion spent, 95 percent get no return
The individual usage differences documented by OpenAI reflect a broader pattern identified by a separate study from MIT’s Project NANDA. Despite $30 billion to $40 billion invested in generic AI initiatives, only 5 percent of organizations are seeing transformational returns. researchers call it "GenAI Divide" – a difference that separates the few organizations that succeed in transforming processes with adaptive AI systems from the majority stuck in pilots.
The MIT report found limited disruption across industries: only two of nine key sectors – technology and media – show material business change from generic AI use. Large companies are ahead in pilot volume but lagging behind in successful deployment.
The pattern is consistent across both studies. Organizations and individuals are purchasing technology. They are launching a pilot. They are attending training sessions. But most people are stuck somewhere between adoption and change.
While official AI projects are stalled, a shadow economy is thriving
The MIT study reveals a surprising disconnect: While only 40 percent of companies have purchased official LLM subscriptions, employees at more than 90 percent of companies regularly use personal AI tools for work. Almost every respondent reported using some form of LLM as part of their regular workflow.
"This ‘shadow AI’ often provides better ROI than formal initiatives and reveals what really works to bridge the divide," MIT’s project Nanda was found.
The shadow economy provides a clue to what is happening at the individual level within organizations. Employees who take the initiative – those who sign up for individual subscriptions, who experiment on their own time, who figure out how to integrate AI into their workflows without waiting for IT approval – are outperforming colleagues who wait for official guidance that may never come.
These shadow systems, largely untested, often perform better and are adopted faster than corporate tools. Worker sentiment reveals a preference for flexible, responsive tools – exactly the kind of experimentation that separates OpenAI’s marginal workers from the middle.
The biggest shortcomings appear in technical tasks that required experts
The largest relative gap between marginal and middle workers appears in coding, writing, and analysis – precisely the job categories where AI capabilities have advanced most rapidly. Marginal workers are not only doing the same work faster; They appear to be doing completely different things, expanding into technical areas that were previously inaccessible to them.
Among ChatGPT enterprise users outside of engineering, IT and research, coding-related messages have increased by 36 percent over the past six months. Someone in marketing or HR who learns to write scripts and automate workflows is going to be a markedly different employee than a coworker who hasn’t done the same — even if they have the same title and started with the same skills.
Academic research on AI and productivity paints a complex picture. Several studies cited in the OpenAI report show that AI has a "similar effect," Disproportionately helping low-performing workers close the gap with their higher-performing peers. But the parity effect may only apply within the population of workers who actually use AI regularly. A meaningful share of workers are not in that group at all. They remain light users or non-users, even as their more adventurous peers move away.
Companies are also divided, and the gap is increasing month by month
The division is not just between individual workers. It exists throughout organizations.
Frontier firms – those at the 95th percentile of adoption intensity – generate nearly twice as many AI messages per employee as the average enterprise. For messages sent via custom GPT, purpose-built tools that automate specific workflows, the difference increases to seven times.
These numbers suggest a fundamentally different operating model. In intermediary companies, AI may be a productivity tool that individual employees use at their discretion. At leading companies, AI appears to be embedded in core infrastructure: standardized workflows, persistent custom tools, systematic integration with internal data systems.
The OpenAI report notes that nearly one in four enterprises have still not enabled connectors that give AI access to company data – a fundamental step that dramatically increases the usefulness of the technology. The MIT study found that companies that purchased AI tools from specialized vendors were 67 percent successful, while those built in-house had a success rate of only one in three. For many organizations, the AI era has technically arrived but has not yet begun in practice.
Technology is no longer the problem – organizations are
For executives, the data presents an inconvenient challenge. Technology is no longer a barrier. OpenAI notes that it releases a new feature or capability approximately every three days; The models are moving faster than most organizations can handle. The bottleneck in what AI can do now is whether organizations are ready to take advantage of it.
"The dividing line is not intelligence," The MIT author writes. Enterprise AI problems relate to memory, adaptability, and learning ability. The problems arise less from regulations or model performance and more from tools that fail to learn or adapt.
According to the OpenAI report, leading companies continually invest in executive sponsorship, data preparation, workflow standardization, and deliberate change management. They create cultures where custom AI tools are created, shared, and refined across teams. They track performance and run evaluations. They make AI adoption a strategic priority rather than a personal choice.
The rest are leaving it to chance – hoping that employees will discover the tools themselves, experiment on their own time, and somehow propagate best practices without infrastructure or incentives. The six-fold difference suggests that this approach is not working.
The window to catch up is closing faster than most companies realize
With enterprise contracts locked in over the next 18 months, there is a shrinking window for vendors and adopters to cross the divide. The GenAI divide identified by the MIT report is not going to last forever. But the organizations that find their way early will be the ones that define the next era of business.
Warnings have been given in both reports. The OpenAI data comes from a company with a clear interest in promoting AI adoption. Productivity figures are self-reported by customers who are already paying for the product. The MIT study, while independent, relied on interviews and surveys rather than direct measurements. The long-term impact of this technology on employment, wages, and workplace mobility remains uncertain.
But the main findings – that access alone does not produce adoption, and that adoption varies considerably even within organizations that have made the same devices available to everyone – is consistent with how previous technologies have spread through the economy. Spreadsheets, email, and the Internet all created similar divisions before they eventually became universal. The question is how long the current gap will persist, who will benefit during the transition, and what will happen to workers who find themselves on the wrong side of it.
For now, the division is clear. Ninety percent of users said they prefer humans "mission-critical work," While AI has "Won the war for simple work." Workers who are stepping up are not doing so because they lack access to their colleagues. They’re growing because they decided to use something everyone else already had – and kept using it until they figured out what it could do.
The 6x difference is not about technology. It’s about behavior. And behavior, unlike software, can’t be deployed with a company-wide rollout.
<a href