Is the AI Workforce Divide Splitting Employees Into Four Groups?

As an Amazon Associate we earn from qualifying purchases.

Key Takeaways

• PwC identifies four worker cohorts, with the 56% “engine room” receiving the least AI advantage.
• Daily AI use is rising, but access to learning, trust, and perceived job security remain uneven.
• The findings point to work design and training gaps, not a simple forecast of AI-driven job losses.

What Did PwC’s 49,364-Worker Survey Find?

On September 29, 2026, PwC published one of the largest recent examinations of how artificial intelligence is changing workers’ experience of employment. Its Global Workforce Hopes and Fears Survey 2026 gathered responses from 49,364 workers across 48 countries and regions and 29 sectors during May and June 2026. The results suggest that an AI workforce divide is emerging, but it is more complicated than a simple split between people who use artificial intelligence and people who do not.

PwC found that 64% of respondents had used artificial intelligence at work during the previous 12 months, an increase of 10 percentage points from the prior survey. The proportion using generative artificial intelligence, or GenAI, every day increased from 14% to 22%. Another 59% expected their AI use at work to increase during the following 12 months.

Those figures demonstrate rapid diffusion, but adoption is only part of PwC’s argument. The firm’s more consequential finding is that workers appear to be experiencing AI very differently depending on two factors: the benefits they believe they receive from AI and the perceived market value of their skills.

PwC calls the first dimension “AI advantage.” It reflects workers’ reported gains in areas such as work quality, creativity, skills, and their value to their employer. The second dimension measures labor-market pressure, including respondents’ perceptions of demand for their roles and how difficult their skills would be for an employer to replace or acquire.

Combining these dimensions produced four worker archetypes: front-runners, AI insurgents, indispensables, and the engine room. These groups are not occupational categories. A software developer, accountant, engineer, manager, sales professional, or technician could potentially fall into different categories depending on AI use, reported benefits, organizational conditions, and perceived labor-market scarcity.

This distinction is important because the results shift the conversation away from the familiar question of whether AI will “take jobs.” The survey instead points to differences in who gets meaningful access to AI, who receives training, who is rewarded for experimentation, whose expertise remains scarce, and who feels increasingly exposed.

The survey also captures pressures that extend beyond artificial intelligence. Only 34% of workers said they could pay their bills and still have money left at the end of the month, according to PwC’s findings. Sixty-two percent said cost-of-living pressures had affected them moderately or significantly at work during the previous year. Workers are therefore experiencing AI adoption within a broader environment of financial pressure, organizational change, and uncertainty.

That context helps explain why technological adoption alone provides an incomplete picture. Two employees may work for the same organization and technically have access to the same AI platform, yet experience very different outcomes because one has time to experiment, formal training, managerial support, scarce expertise, and a clear career path, whereas the other does not.

How Do PwC’s Four Worker Groups Differ?

PwC’s segmentation places 14% of respondents among the front-runners. These workers report both a strong AI advantage and skills that they believe are in high demand. They appear to occupy the most favorable position in the emerging AI workforce divide.

More than half of front-runners, 51%, reported using GenAI daily. Nearly 80% said they had access to learning and development resources. Three-quarters said recent changes made them optimistic about their organization’s future, and 85% said they felt ready to adapt to new ways of working.

Their confidence also extends to employment. PwC reports that 85% of front-runners felt very confident about their job security. Yet that confidence may make them harder to retain rather than more loyal. Twenty-nine percent said they were very or extremely likely to change employers within the next year.

The second group, AI insurgents, represents 18% of workers. Their skills are perceived as less scarce, but they are using AI aggressively and reporting meaningful benefits from it. PwC characterizes them as ambitious and energetic. They also report particularly high weekly AI use.

This group is particularly interesting because it suggests AI can potentially alter a worker’s competitive position without immediately changing the underlying occupation. Someone whose conventional skills are relatively common may become more productive, adaptable, or valuable by becoming unusually effective at integrating AI into everyday work.

The indispensables, representing 11%, occupy the opposite position. Their skills are difficult for employers to replace, but their reported AI advantage is relatively low. These workers may include experienced specialists whose value comes from accumulated expertise, institutional knowledge, judgment, professional credentials, customer relationships, or technical skills that remain difficult to reproduce.

For employers, the indispensables pose a different challenge from AI insurgents. The question is less about creating scarce expertise and more about helping existing experts use AI without undermining the knowledge that makes them valuable.

The largest group is what PwC calls the engine room, representing 56% of respondents. These workers perform much of the organization’s routine day-to-day delivery, but PwC classifies them as having lower labor-market scarcity and low or no AI advantage.

Only 11% of this group reported daily GenAI use, compared with 51% among front-runners. PwC also reports that a majority of engine-room workers said they had never used AI.

The divide extends into development opportunities. Fewer than 40% of engine-room workers reported access to the learning and development resources they need, compared with nearly 80% of front-runners. Only about half felt confident about their job security.

Trust is also sharply different. PwC found that 75% of front-runners trusted top management and 82% trusted their immediate managers. Among the engine room, those figures fell to 33% and 40%, respectively.

The four categories therefore capture more than technical proficiency. They describe different combinations of opportunity, confidence, organizational support, market power, and technology use.

Why Is AI Access Becoming a Career Variable?

Artificial intelligence is increasingly functioning as a career variable rather than merely another office technology.

PwC reports that daily GenAI users experience the workplace differently from infrequent users. Sixty-eight percent of daily users said they were very or extremely confident about their job security. Daily users were also more likely to trust senior management, ask for promotions, and believe they could develop new skills.

Those associations do not prove that AI use causes career confidence. Workers who already possess greater autonomy, stronger skills, supportive managers, or better career prospects may be more likely to receive AI tools and use them frequently. The organizational environment may be producing both the AI adoption and the confidence.

Even with that qualification, the distribution of access matters. PwC found that overall access to learning and development resources fell to 51%, from 59% in the previous year. Workers’ confidence in their ability to develop skills, at 61%, therefore exceeds their reported access to the resources needed to do so.

The OECD’s 2026 work on AI and skills reinforces an important point: preparing workers for AI does not primarily mean turning the workforce into programmers or machine-learning engineers. The OECD estimates that fewer than 1% of workers require advanced AI-specific skills. For most workers, digital literacy, data interpretation, problem-solving, creativity, communication, and managerial capability are more relevant.

That distinction changes the scale of the training problem. An organization does not need every employee to understand model architecture. It does need people to know when an AI system is appropriate, how to formulate a useful task, how to evaluate an output, how to protect confidential information, when human review is necessary, and how the technology changes the workflow surrounding their job.

The OECD also reports that more than half of workers using AI in evidence examined by the organization had received employer-funded training, and workers receiving training were more likely to report improvements in performance and working conditions.

Training therefore appears to be part of the infrastructure of adoption rather than a separate human-resources benefit.

This issue has also appeared in broader examinations of how people are using AI. Productivity gains and displacement concerns can exist at the same time. A worker may become substantially more capable with AI yet simultaneously conclude that an employer will eventually require fewer people to perform the same volume of work.

For organizations, that creates an important management problem. Asking employees to adopt technology that they believe could weaken their bargaining position requires more than software licenses. It requires clarity about how jobs are expected to change and credible opportunities for workers to develop the skills needed in the redesigned organization.

Why Does Work Design Matter More Than Deploying AI Tools?

PwC’s survey points toward a broader distinction between using AI and reorganizing work around AI.

That distinction helps explain why widespread experimentation has not automatically produced equally widespread financial returns. PwC’s 29th Global CEO Survey, based on responses from 4,454 chief executives in 95 countries and territories, found that 56% reported neither revenue gains nor cost reductions from AI during the previous 12 months. Only 12% reported both.

The workforce findings offer one possible explanation. Organizations can purchase AI products faster than they can redesign roles, incentives, decision rights, training systems, performance measures, quality controls, and career structures.

A company might provide a generative AI assistant to thousands of employees without deciding which workflows should change. Employees then use the technology individually, producing isolated productivity improvements that may never alter the economics of the overall process.

Consider a document workflow. An AI system may reduce the time required to produce a first draft from three hours to 30 minutes. That appears to be a large productivity gain. If the organization still requires the same approvals, duplicated data entry, manual checking, handoffs, meetings, and legacy reporting steps, much of the potential gain remains trapped inside the old system.

The same principle applies to more complicated work. AI can summarize customer records, generate software, analyze documents, identify patterns, produce forecasts, draft communications, or prepare recommendations. Those outputs do not automatically determine who is responsible for validation, who makes the final decision, what happens when the model is wrong, or how saved time should be redeployed.

PwC’s 2026 Global AI Jobs Barometer provides evidence that the labor market itself is changing alongside these organizational experiments. The study analyzed more than one billion job advertisements across six continents and reported that occupations requiring AI skills were growing substantially faster than the broader job market. It also found a sizable wage premium associated with AI skills.

The Barometer distinguishes between jobs being “professionalized” by AI, where the technology increases the value of human expertise, and jobs being “democratized,” where AI makes tasks easier for less specialized workers to perform. PwC reported stronger job and wage growth among the professionalized group.

That framework helps explain why AI may increase the value of some expertise at the same time that it reduces the scarcity of other skills.

A mature AI strategy therefore requires decisions about the architecture of work. Management must determine what the technology should do, what people should continue doing, what expertise becomes more valuable, what controls are required, and how the gains from automation or augmentation affect workloads and staffing.

What Does the AI Workforce Divide Not Prove?

PwC’s four-group framework is useful, but it should not be interpreted as a deterministic map of the future workforce.

The categories are derived from survey responses rather than direct measurement of every worker’s productivity, skill scarcity, wage trajectory, or susceptibility to automation. PwC explains that its labor-market index incorporates respondents’ perceptions of demand for their roles and how difficult their skills would be to replace. Its AI advantage index is based on benefits that workers report receiving from AI.

Those are meaningful indicators, but perception and objective labor-market conditions are not identical.

A worker may underestimate the scarcity of a skill. Another may overestimate it. An employee enthusiastic about AI may report stronger benefits because the employee is already working in a supportive organization. Someone in a poorly managed organization might have access to equally capable technology but little time, permission, or incentive to experiment with it.

The survey is therefore stronger as evidence of workforce experience than as proof of causation.

The same caution applies to the relationship between AI use and job security. Daily users report greater confidence, but that does not establish that using AI causes employment security. High-performing organizations may simultaneously provide better technology, stronger training, greater managerial trust, and more secure employment.

Similarly, AI exposure should not be treated as a direct estimate of future job losses. The International Labour Organization warned in April 2026 that occupational exposure indicators show where technology could change tasks, not what will necessarily happen to employment. Adoption costs, business models, regulation, worker responses, customer demand, organizational constraints, and the creation of new tasks all affect actual outcomes.

The ILO’s broader research on generative AI and jobs estimates that one in four workers globally is employed in an occupation with some degree of GenAI exposure. Its central finding is that transformation of jobs is more likely than complete replacement in most exposed occupations.

New Space Economy has examined the same distinction in its analysis of AI and employment transition. Existing evidence shows changes in tasks and hiring patterns, but it does not establish that permanent mass unemployment is already occurring.

The PwC study also should not be interpreted as saying that every organization must convert every engine-room employee into a front-runner. Some work is less suitable for generative AI. Some roles derive value from physical execution, regulated judgment, interpersonal contact, hands-on technical expertise, or responsibilities that cannot safely be delegated to automated systems.

The important finding is unequal opportunity to adapt, not the assumption that identical AI adoption is desirable everywhere.

What Should Employers and Managers Do Differently?

The PwC results suggest that the most immediate workforce problem is not a shortage of enthusiasm for learning. It is a mismatch between workers’ willingness to adapt and the organizational conditions provided for adaptation.

Almost all respondents, 88%, told PwC they had applied new skills in their jobs during the previous year. Forty percent said they had done so to a large or very large extent. Sixty-four percent had learned new tools or technologies to at least a moderate extent.

Yet only 51% said they had access to the learning and development resources they needed.

That gap suggests employers should treat learning capacity as part of operating capacity. Training cannot simply be added to an already full workload and described as an opportunity. Workers need protected time, access to approved tools, role-specific examples, feedback, and a practical reason to apply what they learn.

This matters particularly for the engine-room majority. If these employees perform much of the organization’s recurring delivery work, leaving them behind creates an operational problem as well as an employment problem. An organization cannot become meaningfully AI-enabled if advanced tools are concentrated among a small group of specialists and enthusiasts but disconnected from the processes through which most products and services are delivered.

Training should also be differentiated by role. The OECD evidence indicates that most employees do not need advanced technical AI knowledge. A finance employee may need to understand model-assisted analysis and verification. A customer-service worker may need to understand escalation and privacy rules. A manager may need to redesign approvals and performance measures. An engineer may need to validate AI-generated technical work against formal requirements.

Policy initiatives increasingly reflect this broader definition of AI readiness. New Space Economy’s examination of the Canadian AI strategy illustrates how national AI policy is moving beyond research funding toward literacy, adoption, workforce development, infrastructure, and organizational deployment.

Managers also occupy a difficult position in PwC’s findings. Twenty-four percent of managers and 31% of senior executives said they found change difficult to manage. Managers are therefore being asked to help employees adapt at the same time that their own responsibilities, technologies, and performance expectations are changing.

Organizations that expect managers to lead AI adoption need to train them in work redesign, not merely provide demonstrations of software features.

That includes deciding what should be automated, what requires human judgment, how to assess output quality, how performance will be measured, how mistakes will be handled, and how employees can challenge an AI-supported decision without being treated as resistant to change.

How Could the Four-Speed Pattern Affect Specialist Industries?

The importance of PwC’s framework becomes clearer when it is applied to industries competing for scarce technical talent.

Consider aerospace, advanced manufacturing, cybersecurity, biotechnology, financial services, energy, telecommunications, or the commercial space sector. These industries contain both highly specialized employees and large operational workforces. AI can affect each group differently.

New Space Economy’s examination of the space economy workforce shows that a technology-intensive industry depends on much more than elite engineers. Software developers, assemblers, technicians, sales professionals, operations staff, managers, data specialists, and administrative workers all contribute to the productive system.

That workforce could reproduce PwC’s four-group pattern. An experienced systems engineer who rarely uses GenAI might resemble an indispensable. A younger software developer who aggressively integrates AI tools despite having relatively common baseline skills might resemble an AI insurgent. An experienced specialist who combines difficult-to-replace expertise with effective AI use could resemble a front-runner.

Large groups performing recurring operational work could find themselves in the engine room if AI access, training, and experimentation remain concentrated elsewhere.

The result would not necessarily be immediate job elimination. It could instead be a widening difference in mobility, pay, confidence, workload, advancement, and access to interesting work.

That possibility is consistent with recent New Space Economy coverage of how AI can create and erase work. Automation can eliminate tasks and create new checking, remediation, integration, supervision, and quality-control work at the same time.

Organizations therefore need to watch the quality of the jobs being created as carefully as the quantity of jobs being automated.

If AI removes repetitive entry-level tasks, employers also need to consider how future experts will acquire experience. Many professional careers have historically developed through lower-risk tasks that allow junior employees to observe mistakes, build judgment, and gradually take responsibility.

Automating those tasks can improve immediate productivity but weaken the apprenticeship mechanism that produces tomorrow’s scarce expertise.

The broader AI adoption evidence points toward the same tension. AI can expand what an individual can accomplish without guaranteeing that the resulting organization develops stronger human capability.

The long-term workforce question is therefore not simply who has access to an AI assistant. It is whether organizations use AI to expand employee capability, redesign work intelligently, maintain expertise pipelines, and distribute learning opportunities widely enough that technological progress does not produce an increasingly segmented workforce.

Summary

PwC’s 2026 Global Workforce Hopes and Fears Survey provides a useful framework for understanding a workforce that is beginning to experience artificial intelligence at different speeds.

The most striking finding is the size of the engine-room group. Fifty-six percent of surveyed workers fall into a category characterized by relatively low perceived skill scarcity and low or limited AI advantage. These employees perform much of the ordinary work required to keep organizations operating, yet they report less access to AI, fewer development opportunities, weaker trust in management, and lower confidence in job security than the front-runner minority.

At the opposite end are front-runners, representing 14% of respondents. They combine scarce skills with strong reported AI benefits, frequent use, extensive access to learning, and high confidence. Their strong external prospects create a retention problem for employers because nearly three in 10 say they are very or extremely likely to change employers within the following year.

Between these groups are AI insurgents, who may be increasing their value through aggressive adoption, and indispensables, whose scarce expertise remains valuable despite relatively limited AI engagement.

The framework should not be mistaken for a forecast that one group will prosper and another will disappear. PwC’s categories depend partly on self-reported experience and perceived skill scarcity. The study identifies associations rather than proving that AI use causes job security, trust, or career success.

Its strongest message concerns organizational choices.

Artificial intelligence is increasingly widespread, but access to technology is not the same as access to advantage. Workers need role-specific training, time to learn, managerial support, opportunities to experiment, clear expectations, and workflows designed to turn technological capability into useful work.

Employers face two risks if the divide widens. They may lose the AI-enabled workers with the strongest labor-market alternatives, and they may fail to develop the much larger group responsible for everyday operations.

The strategic challenge is therefore broader than increasing AI adoption. Organizations must decide whether AI becomes a capability distributed through the workforce or an advantage concentrated among workers who were already best positioned to benefit from change.

Appendix: Useful Books Available on Amazon

• Co-Intelligence: Living and Working with AI
• Human + Machine: Reimagining Work in the Age of AI
• Prediction Machines, Updated and Expanded: The Simple Economics of Artificial Intelligence
• Power and Prediction: The Disruptive Economics of Artificial Intelligence
• The Coming Wave: Technology, Power, and the Twenty-first Century’s Greatest Dilemma

Appendix: Top Questions Answered in This Article

What Are PwC’s Four AI Workforce Groups?

PwC divides surveyed workers into four archetypes based on reported AI advantage and perceived labor-market scarcity. The groups are front-runners, AI insurgents, indispensables, and the engine room. They represent different combinations of AI benefit, skill demand, access to development, and workforce position rather than four fixed occupations.

What Percentage of Workers Are in PwC’s Engine Room?

PwC places 56% of surveyed workers in the engine-room category, making it by far the largest of the four groups. These workers provide much of organizations’ day-to-day delivery but generally report lower AI advantage and lower perceived skill scarcity than front-runners and indispensables.

How Many Workers Use AI at Work?

PwC found that 64% of respondents had used AI at work during the previous 12 months. Daily use of generative AI increased from 14% to 22%. Adoption is therefore spreading rapidly, but the survey shows that frequency of use and reported benefits remain distributed unevenly across the workforce.

Who Are PwC’s Front-Runners?

Front-runners account for 14% of PwC’s workforce sample. They combine skills perceived as difficult to replace with strong reported benefits from AI. They also report high levels of learning access, adaptability, job-security confidence, and trust in management, although 29% say they are very or extremely likely to change employers within a year.

Who Are AI Insurgents?

AI insurgents represent 18% of respondents. Their existing skills are perceived as less scarce than those of front-runners or indispensables, but they use AI extensively and report meaningful advantages from it. They illustrate how effective AI adoption could potentially strengthen the position of workers whose traditional skills alone are relatively common.

Who Are the Indispensables?

Indispensables account for 11% of PwC’s sample. Their expertise is perceived as scarce and difficult for employers to replace, but they have relatively limited AI advantage. Their position highlights why organizations may need to help established experts adopt AI without damaging the specialist knowledge that makes those employees valuable.

Does PwC’s Survey Prove That AI Makes Jobs More Secure?

No. Frequent AI users report greater confidence in their employment prospects, but the survey does not establish that AI use causes job security. Better-managed organizations, scarce skills, stronger performance, greater autonomy, or better training could simultaneously contribute to both frequent AI use and higher confidence.

Does AI Exposure Mean a Job Will Be Eliminated?

No. The International Labour Organization cautions that exposure measures identify tasks and occupations that technology may affect, not employment outcomes that are certain to occur. Costs, adoption rates, regulation, business demand, new tasks, human oversight, and organizational decisions all influence whether exposed jobs expand, contract, or change.

Do Most Workers Need Advanced AI Programming Skills?

No. OECD research indicates that fewer than 1% of workers require advanced AI-specific skills. Most employees are more likely to need broader digital literacy, data interpretation, problem-solving, communication, creativity, managerial capability, and the ability to evaluate AI-assisted work in the context of their existing profession.

What Is the Main Management Lesson From PwC’s Findings?

Providing access to AI tools is insufficient by itself. Employers also need to redesign workflows, provide role-specific learning, protect time for skill development, establish quality controls, support managers, and explain how jobs are expected to change. The effectiveness of AI adoption increasingly depends on the organization surrounding the technology.

Appendix: Glossary of Key Terms

AI Workforce Divide

A pattern in which workers receive unequal benefits, opportunities, training, confidence, and career advantages from artificial intelligence. The term describes differences in how workers experience AI adoption rather than a simple division between people whose jobs are safe and people whose jobs are threatened.

Generative Artificial Intelligence

Artificial intelligence systems that generate new material such as text, images, software code, audio, analysis, or structured information in response to instructions or other inputs. In workplace settings, generative AI is increasingly used for drafting, summarization, analysis, coding, research support, and administrative tasks.

AI Advantage

PwC’s measure of how much workers report benefiting from artificial intelligence. It incorporates perceived improvements in areas such as work quality, creativity, skills, and value to an employer. It measures reported benefits rather than independently measured productivity or economic output.

Labor-Market Pressure

In PwC’s workforce segmentation, this dimension reflects workers’ perceptions of demand for their roles and how difficult their skills would be for employers to obtain or replace. It is used with AI advantage to create the four workforce archetypes.

Worker Archetype

A conceptual category used to group workers who share selected characteristics. PwC’s archetypes are analytical groupings created from survey data. They should not be interpreted as permanent employee classifications or as predictions of an individual’s future employment outcome.

Front-Runners

PwC’s category for workers who combine strong reported AI advantage with skills considered difficult to replace. Front-runners represent 14% of the surveyed workforce and report particularly high AI use, access to learning, job-security confidence, adaptability, and career mobility.

AI Insurgents

Workers who report significant benefits from AI even though their underlying skills are perceived as less scarce. PwC places 18% of respondents in this group. Their position suggests that effective AI use can potentially change how workers contribute value within an organization.

Indispensables

Workers whose expertise is perceived as scarce and difficult for employers to replace but who report relatively limited AI advantage. PwC places 11% of respondents in this category, highlighting the importance of helping established specialists integrate AI without weakening valuable domain expertise.

Engine Room

PwC’s largest workforce category, representing 56% of respondents. These workers are described as central to routine organizational delivery but have relatively low perceived skill scarcity and low or no AI advantage, accompanied by weaker learning access, trust, and employment confidence.

Upskilling

The process of developing additional capabilities that help a worker perform an existing or evolving role more effectively. In an AI context, upskilling can include digital literacy, evaluating AI outputs, data interpretation, workflow redesign, domain-specific AI use, and stronger human judgment.