The first phase of workplace AI inequality was measured by adoption, but the next will be measured by opportunity. If some employees are given greater freedom to automate workflows, access valuable data and redesign how work gets done, today’s experimentation gap could quietly become the leadership gap of tomorrow.
For the past several years, much of the conversation about AI inequality has centered on access and adoption, with researchers, employers and technology companies asking which groups are using generative AI and which are being left behind. That was a reasonable place to start, but it may increasingly be the wrong question, because knowing whether someone occasionally uses an AI assistant tells us remarkably little about whether the technology is actually increasing that person’s value, influence or trajectory inside an organization.
The more consequential divide may instead be emerging between people who are allowed to use AI and people who are empowered to do important things with AI, and the distinction between those two groups could eventually matter enormously. An employee using an AI tool to summarize a meeting or polish an email technically counts as an AI user, but their experience is fundamentally different from the colleague who has been given access to company data, internal systems and important workflows and is encouraged to automate processes, build agents, redesign operations or take responsibility for work that previously required an entire team.
That difference is the beginning of what former Microsoft corporate vice president Liat Ben-Zur describes as an “AI opportunity gap,” and it deserves considerably more attention than another survey measuring how many employees have opened ChatGPT this week. If women and men are adopting AI at comparable rates but are being given different opportunities to apply it to consequential work, headline adoption statistics could suggest equality while a much more important form of inequality develops underneath them.
Experimentation Is Becoming Career Capital
One of the peculiarities of the current AI transition is that expertise is being created while the technology itself is still being invented, which means experimentation has become an unusually valuable form of professional development. The employee who gets six months to test agents against real workflows, connect models to proprietary data and learn where automation succeeds or fails is accumulating knowledge that cannot easily be replicated through a training course.
That experience compounds, because successful experimentation tends to produce more responsibility, while more responsibility creates opportunities for more sophisticated experimentation. Someone who automates a reporting process may next be asked to rethink the department’s analytics workflow, then lead an AI implementation project, then manage a larger transformation initiative, creating a career flywheel that began with something as seemingly mundane as being given permission to try.
The problem is that permission is rarely distributed as evenly as software licenses, because meaningful experimentation requires more than an account and a password. It requires access to data, systems, budgets, important assignments, managerial encouragement and, perhaps most importantly, enough institutional trust to be allowed to make mistakes while learning.
That makes AI opportunity vulnerable to many of the same informal workplace dynamics that have historically shaped advancement, even when nobody consciously intends to create an unequal system. Managers decide who gets the interesting project, leaders decide who can access sensitive systems, teams decide whose experimental failures are treated as useful learning experiences and organizations decide which employees are considered sufficiently promising to receive resources before they have demonstrated exactly what they can do with them.
When Opportunity Starts Looking Like Merit
This is where the AI opportunity gap becomes particularly dangerous, because unequal opportunity does not necessarily continue to look like unequal opportunity once enough time has passed. Eventually, it starts producing measurable differences in performance.
Imagine two equally capable employees beginning from roughly the same position, with one receiving repeated opportunities to automate workflows and experiment with increasingly sophisticated AI systems while the other primarily uses AI for routine productivity tasks. A year or two later, the first employee may genuinely be faster, more productive, more technically fluent and capable of managing a significantly broader scope of work.
Those differences are real, but they did not necessarily originate with differences in ability, ambition or intelligence. They may have originated with differences in opportunity, and by the time the organization begins making promotion decisions, the original cause can become almost invisible.
The employee with deeper AI experience now appears to be the obvious candidate for advancement because they have demonstrated greater productivity, managed more complex projects and developed skills the business desperately needs. What began as unequal access to consequential experimentation has been converted into seemingly objective evidence of merit, creating exactly the kind of inequality that is hardest for organizations to recognize because the final promotion decision can be entirely rational even if the conditions that produced it were not.
“Everyone Gets AI” Isn’t an AI Strategy
This should also force companies to reconsider what they mean when they talk about democratizing AI, because purchasing enterprise licenses and offering generic training sessions may create technological equality without creating anything resembling professional equality. Giving 10,000 employees access to the same model is meaningful only if employees also have reasonable opportunities to apply those capabilities to work that expands their skills, responsibilities and value.
Companies therefore need to start measuring AI opportunity differently, looking beyond logins, prompts and training completion rates toward the kinds of work employees are actually being allowed to perform. Who is building automated workflows, who has access to the data necessary to make those workflows valuable, who is being assigned AI-related stretch projects, whose managers encourage experimentation and who is being invited into the rooms where AI is being used to redesign the organization are much more revealing questions than simply asking who uses the technology.
For marketers, this distinction is especially important because marketing departments are becoming laboratories for applied AI, spanning creative development, research, media optimization, analytics, customer intelligence, personalization, content production and increasingly autonomous workflows. The people given meaningful responsibility across those areas today are not simply learning another marketing technology, because they are developing an emerging form of managerial literacy that could influence who is considered capable of running increasingly automated organizations tomorrow.
The Leadership Pipeline Is Being Built Right Now
Organizations tend to think about leadership pipelines over years, but AI could compress that timeline because the technology is changing the amount and complexity of work individual employees can control. Someone who becomes unusually adept at orchestrating AI systems may suddenly be capable of managing a scope of work that would previously have required substantially more seniority, larger teams or considerably greater resources.
That creates enormous opportunities for companies willing to distribute those capabilities broadly, particularly because AI could theoretically weaken some traditional barriers to advancement by giving more people access to sophisticated analytical, creative and operational capabilities. The same technology, however, could reinforce existing disparities if access to its most valuable applications flows primarily toward employees who already receive the best assignments, strongest sponsorship and greatest organizational latitude.
The solution is not to artificially equalize outcomes or turn every AI project into an exercise in demographic accounting, but companies should become much more deliberate about equalizing the opportunity to develop consequential AI experience. That means examining who receives high-impact assignments, expanding access to controlled experimentation environments, ensuring managers do not reserve AI transformation projects for their usual trusted performers and treating practical AI experience as something employees need opportunities to acquire rather than something they are simply expected to possess.
The first AI gender gap was relatively easy to see because companies could measure who was using the technology, but the next one could be considerably harder to recognize because it will emerge through assignments, permissions, experimentation and accumulated experience. By the time those differences appear in productivity numbers, promotion decisions and leadership pipelines, organizations may believe they are simply rewarding their strongest AI-era performers, without realizing they helped determine who had the opportunity to become those performers in the first place.