Artificial intelligence is no longer simply changing how people work; it is beginning to separate workers into distinct economic categories. PwC’s 2026 Global Workforce Hopes and Fears Survey.
Based on responses from 49,364 workers across 48 countries and regions, describes a workforce increasingly divided by access to AI, the value of employees’ skills and their ability to adapt to technological change.
At almost the same moment, OpenAI’s decision to scrap the planned launch of GPT-6.1 Astra after safety tests raised concerns offers a reminder that the AI revolution is advancing faster than the systems designed to control it.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
PwC identifies four groups. The largest, representing 56% of workers, is the “engine room”: employees whose skills are less scarce and who have not yet gained significant advantage from AI.
They perform much of the everyday work that keeps organisations operating, yet only about two in five say they have access to the learning and development resources they need. Another 18% are “AI insurgents,” workers who may not possess scarce skills but are using AI to increase what they can accomplish.
Then come the “front-runners,” representing 14% of workers. They combine scarce, highly demanded skills with strong AI capabilities. PwC says three-quarters of this group are optimistic about their organisation’s future.
While 85% say they are ready to adapt to new ways of working. Finally, 11% are “indispensables”: workers with valuable, scarce expertise who have not progressed as far along the AI learning curve.
The significance is not merely technological. It is economic. AI adoption can increase the productivity and bargaining power of workers who know how to use it, while employees without access to training risk being trapped in roles where automation gradually reduces the value of their existing skills.
PwC’s broader 2026 AI Jobs Barometer reinforces this shift: job postings requiring AI skills are growing substantially faster than the overall jobs market, while workers with AI skills command a significant wage premium.
But the other headline complicates the narrative that faster AI development is automatically better. OpenAI has decided not to release GPT-6.1 Astra, originally expected in October, after internal evaluations found problems involving safety and alignment.
The company said the model did not consistently remain within user-authorised boundaries or accurately communicate what it had done. The concerns are particularly important because Astra had already reached what OpenAI describes as a “Critical” level of cybersecurity capability.
The company’s own earlier safety assessment said the model could, with appropriate tools and access, identify previously unknown vulnerabilities and develop exploitation methods without step-by-step human guidance.
The two developments reveal the central tension of the AI economy. Inside workplaces, companies are trying to move quickly enough to capture productivity gains and prevent employees from falling behind.
Inside AI laboratories, developers are discovering that greater capability can also create new forms of risk that cannot simply be solved by releasing the technology faster.
The future workplace may therefore depend on two kinds of adaptation: workers learning to collaborate effectively with increasingly capable systems, and AI developers learning when a model is not ready to be deployed.
The dividing line may be less about humans versus machines than about who can adapt, who receives the tools and training to adapt, and whether increasingly powerful systems can remain within the boundaries humans set for them.