McKinsey: AI May Create More Jobs Than It Eliminates

Artificial intelligence could eliminate millions of existing roles while creating even more jobs over time. That is the central tension in the McKinsey AI jobs forecast: the technology may support long-term employment growth, but the transition could disrupt workers, industries, and communities.

One report estimates that approximately 11 million U.S. workers may need to move into new careers by 2035 as AI changes the labor market Source 1. This figure describes the scale of workforce movement, not necessarily permanent unemployment.

“More jobs created than eliminated” does not mean that every displaced worker will move easily into a new position. Workers may need retraining, different employers, lower starting wages, or relocation. The outcome will depend on how businesses, educators, governments, and workers manage the transition.

What the McKinsey AI Jobs Forecast Means

AI May Destroy Jobs During the Transition

AI can reduce demand for workers who perform repetitive, predictable, and easily digitized tasks. Examples include:

• Repetitive administrative work
• Basic data processing
• Routine customer service
• Standardized research and reporting
• Simple content production
• Transaction and claims processing

Generative AI can draft documents, summarize information, classify records, prepare presentations, and answer common customer questions. Automated workflows can also connect systems that previously required manual coordination.

This does not mean that every occupation containing these tasks will disappear. In many cases, AI will first reduce the time required to complete a task. Employers may then redesign the job, assign workers more complex responsibilities, or reduce the number of employees needed for the same output.

Workers may experience several outcomes:

1. A position may be eliminated.
2. A job may be redesigned around AI tools.
3. A worker may move to another occupation.
4. An existing role may become more productive and valuable.

These outcomes can occur simultaneously within the same company.

The 11 Million-Worker Transition by 2035

A report estimates that approximately 11 million U.S. workers could need to transition into new careers by 2035 Source 3.

The estimate represents workforce movement. It should not be interpreted as a prediction that 11 million people will remain permanently unemployed. Some workers may find new roles quickly, while others may require substantial training or accept temporary income losses before securing comparable employment.

A transition can involve changes in occupation, industry, technical skills, daily responsibilities, employer, or geographic location. Even when total employment rises, individual workers can face financial and social costs, including interrupted income, reduced benefits, weaker professional networks, and uncertainty for their families.

This distinction separates net employment growth from individual job security. An economy can gain jobs while many workers lose their previous positions, and employment can rise while wage growth remains uneven.

Why Job Growth Can Coexist With Job Losses

Labor-market statistics measure the total number of jobs. Workers experience how those jobs are distributed.

Suppose AI eliminates 100 routine roles, while businesses using the technology expand and create 130 new positions. The economy records a net gain of 30 jobs. However, the 100 displaced workers still need new opportunities, and the new positions may require different skills.

The distribution of gains matters as much as the total number. Workers with advanced education, strong digital skills, or access to training may benefit first. Workers in declining industries, isolated communities, or lower-income households may face greater adjustment costs.

How AI Could Create More Jobs Than It Eliminates

New AI-Specific Roles

AI adoption creates demand for professionals who build, deploy, manage, and evaluate intelligent systems. Potential growth roles include:

• AI product managers
• Machine learning engineers
• Data scientists
• AI implementation specialists
• Model evaluators
• AI compliance professionals
• Cybersecurity specialists
• Human-AI interaction designers

Some roles require advanced technical education. Others combine technical knowledge with experience in a specific industry. Healthcare organizations, banks, and law firms may need professionals who understand both AI systems and their sectors.

AI knowledge will become useful beyond technology companies. Healthcare professionals may use diagnostic systems, teachers may design AI-supported lessons, financial analysts may assess automated forecasts, and operations managers may use predictive tools to improve staffing and supply chains.

AI-Enabled Expansion

AI may help organizations serve more customers rather than simply reduce headcount. Lower operating costs can make previously expensive or slow services more accessible.

Potential areas of expansion include:

• Healthcare administration and patient support
• Personalized education
• Scientific research
• Software development
• Small-business services
• Logistics and supply-chain management
• Creative production
• Financial and insurance services

If AI reduces the cost or time required to complete work, organizations may increase output. Higher output can generate demand for complementary workers, suppliers, sales teams, managers, and service providers.

The employment effect depends on business decisions. Companies may use productivity gains to reduce staffing, increase wages, lower prices, expand services, or increase profits.

Complementary Jobs Around AI Systems

AI systems require human support. Organizations need workers to prepare data, monitor performance, test security, correct errors, train users, and create operating policies.

Complementary work may include data preparation, quality control, system monitoring, security testing, error correction, user training, policy development, and vendor management.

Many of these roles will not include “AI” in their titles. A compliance officer, customer-support manager, or software tester may perform AI-related responsibilities without becoming a machine learning engineer.

Human judgment remains especially important when decisions involve safety, ethics, legal responsibility, emotional intelligence, or complex stakeholder needs. AI can identify patterns and generate recommendations, but people remain accountable for many high-stakes outcomes.

Jobs Most Exposed to AI Automation

Routine Knowledge Work

Generative AI affects white-collar work as well as manual labor. Office roles often contain tasks involving text, data, classification, and standardized decisions.

High-exposure activities include drafting standard documents, summarizing information, classifying data, preparing basic presentations, scheduling, repetitive analysis, and routine reporting.

Exposure does not guarantee full replacement. AI may first change the composition of a job, reducing routine output and increasing the importance of interpretation, communication, and exception handling.

Administrative and Customer-Service Roles

Data-entry positions, call-center jobs, transaction-support roles, and basic claims-processing work face exposure because many interactions follow predictable patterns.

Chatbots and automated workflows can handle common questions, retrieve information, complete forms, and route requests. Human workers will still be needed for escalated cases, sensitive conversations, complex troubleshooting, relationship management, complaints, and policy exceptions.

Creative and Professional Work

AI also affects creative and professional work. Richard Florida has discussed a reported 189,000-job wipeout while emphasizing that AI can both enhance workers’ capabilities and eliminate certain jobs Source 5.

That figure should not be treated as a universal forecast without additional verification. It illustrates a broader issue: AI can increase the output of creative professionals while reducing demand for standardized services.

Potentially affected work includes graphic design, copywriting, basic video production, market research, entry-level legal analysis, and routine financial reporting. Original judgment, brand knowledge, client relationships, cultural context, and accountability can distinguish high-value work from standardized output.

Jobs That Could Grow in an AI-Driven Economy

Roles Requiring Human Judgment

AI has difficulty with ambiguous, high-stakes, or socially complex situations. Strategic decision-making, negotiation, leadership, empathy, ethical reasoning, crisis management, and interpersonal communication are likely to remain valuable.

These capabilities do not make a job immune to change. Professionals may still need to use AI, interpret its output, and explain decisions to others.

Skilled Technical and Hands-On Work

Jobs involving physical environments, unpredictable conditions, and on-site problem-solving can be difficult to automate fully. Potentially resilient roles include electricians, equipment technicians, healthcare workers, construction specialists, maintenance professionals, and skilled tradespeople.

AI can improve these jobs through predictive maintenance, diagnostic tools, automated documentation, training simulations, and route optimization. Workers still need to inspect equipment, manage safety risks, complete repairs, and provide physical or emotional support.

AI Governance, Safety, and Oversight

AI adoption creates demand for risk managers, auditors, safety researchers, compliance officers, privacy specialists, and cybersecurity teams.

Some AI research fellows have warned that private laboratories may operate models without standard safeguards, making behavior and risks difficult to assess. This claim should be treated as a warning about transparency and oversight, not as a verified description of every laboratory Source 9.

Organizations need workers who can evaluate model reliability, bias, privacy risks, security vulnerabilities, regulatory compliance, and operational failures.

Why Reskilling Matters

Workers Need Transferable Skills

Workers should develop capabilities that remain useful across tools and employers. Valuable skills include data literacy, digital collaboration, prompt design, critical evaluation of AI output, communication, project management, and industry-specific expertise.

AI literacy means understanding how to identify useful applications, verify accuracy, protect sensitive data, and recognize bias. The strongest combination is technical fluency, industry knowledge, and human-centered judgment.

Employers Must Redesign Jobs

Businesses can approach AI through replacement-focused automation or augmentation-focused job redesign. Responsible implementation involves:

1. Identifying repetitive tasks.
2. Automating low-value activities.
3. Assigning workers higher-value responsibilities.
4. Measuring productivity and quality.
5. Providing training before implementation.

Poorly managed adoption can reduce morale, increase surveillance, create skills shortages, and remove institutional knowledge. Employers should involve workers in redesign decisions because employees often understand process weaknesses and customer needs better than outside vendors.

Governments and Educators Have a Role

Public policy can reduce transition costs through affordable adult education, short-term credentials, career-transition assistance, wage insurance, apprenticeships, and employer training incentives.

Education programs should reflect actual labor-market demand. Partnerships among employers, universities, community colleges, workforce agencies, and technology providers can create practical pathways. Preparation should begin before displacement occurs.

Risks Behind the “More Jobs” Forecast

Job Quality May Decline

A larger number of jobs does not guarantee better employment. New roles may offer lower wages, limited benefits, temporary contracts, or greater performance monitoring.

Companies and policymakers should evaluate wage growth, stability, benefits, autonomy, advancement, and working conditions alongside job counts.

Benefits May Be Unequally Distributed

Workers with advanced education or strong digital skills may gain faster. Entry-level workers, administrative employees, people in declining industries, and communities dependent on one major employer may face greater adjustment costs.

Access to reliable technology, training, and supportive employers will influence who benefits.

Forecasts Are Not Guarantees

The McKinsey AI jobs projection is a scenario-based forecast, not a fixed outcome. Results could change with AI adoption rates, regulation, economic growth, consumer demand, investment, public trust, education capacity, and reskilling success.

The forecast should serve as a planning signal rather than a promise.

How Workers Can Prepare

Audit Current Tasks

Workers should divide their jobs into routine, analytical, relationship-based, physical, and decision-making tasks. This exercise can identify where AI may assist or automate work and where human judgment remains central.

Learn to Use AI Productively

Practical applications include research assistance, drafting and editing, data analysis, meeting summaries, workflow automation, and brainstorming. Workers must verify AI output and understand its accuracy limits, privacy requirements, bias, security risks, and appropriate use cases.

Build a Record of Adaptability

Workers should document productivity improvements, new certifications, AI-related projects, process improvements, and cross-functional experience. Evidence of results is more valuable than a general claim of AI familiarity.

Conclusion

McKinsey expects AI to create more jobs than it eliminates over time. At the same time, approximately 11 million U.S. workers may need to move into new careers by 2035. Those claims are not contradictory.

AI will automate tasks across manual, administrative, creative, and professional work. New jobs will emerge around AI development, expanded services, technical support, governance, safety, and human oversight.

Reskilling and institutional support will determine who benefits. Employers must redesign jobs instead of treating AI only as a cost-cutting tool. Governments and educators must expand practical training, while workers build AI literacy and strengthen judgment, communication, and industry expertise.

AI is not simply a job killer or a job creator. It is a force that reorganizes work. The quality of that reorganization will depend on decisions made before disruption reaches its peak.

Frequently Asked Questions

Will AI create more jobs than it destroys?

McKinsey’s forecast says AI will ultimately create more jobs than it eliminates. The transition may still cause significant disruption, including the need for approximately 11 million U.S. workers to move into new careers by 2035.

How many jobs could AI eliminate?

Available source summaries cite approximately 11 million jobs destroyed or displaced during the transition. This figure should not be interpreted as a permanent net loss because new roles may emerge as businesses adopt AI.

Which jobs are most vulnerable to AI?

Jobs with repetitive, predictable, and easily digitized tasks face the greatest exposure. Examples include data entry, basic customer service, routine administration, standardized reporting, and some entry-level creative and analytical work.

Which jobs are likely to grow because of AI?

Potential growth areas include AI development, cybersecurity, data science, AI governance, system monitoring, healthcare, education, skilled technical work, and roles combining industry expertise with AI tools.

What should workers learn to remain competitive?

Workers should develop AI literacy, data skills, critical thinking, communication, project management, and deep industry knowledge. The strongest position combines technical fluency with judgment, creativity, and interpersonal skills.

Does AI exposure mean a job will disappear?

No. AI exposure means that some tasks within a job could be automated or assisted. Many occupations may be redesigned rather than eliminated, allowing workers to focus on higher-value responsibilities.