AI and the 2027 Labor Market: Jobs and Skills

AI and the 2027 Labor Market: Jobs Growing, Changing and Declining

The most defensible forecast for the 2027 labor market is not that artificial intelligence will eliminate work, nor that it will automatically create better jobs. AI is changing the task mix inside occupations while demographic change, digital access, energy investment, security risk, economic conditions and policy also affect hiring. By 2027, the strongest opportunities are likely to appear where AI investment creates demand for data, software, cybersecurity, infrastructure and governance; where growing sectors need managers and analysts; and where human presence, licensure, trust or physical delivery limits full automation.

The roles under greatest pressure are those dominated by standardized digital production, routine clerical processing and low-context information transfer. Even there, “decline” does not mean every job disappears. It means fewer entry points, a higher output expectation per worker, redesigned teams or a shift toward exceptions and accountability.

This article provides a practical map for employees, managers and learners. It combines global research with current U.S. occupational projections, then converts the evidence into a task audit and a 2027 preparation plan.

What the latest evidence actually says

The World Economic Forum's Future of Jobs Report 2025 surveyed more than 1,000 employers representing over 14 million workers across 55 economies. For 2025–2030, respondents' plans and expectations imply 170 million roles created and 92 million displaced by major structural trends, a net increase of 78 million. The report estimates that 22% of today's formal jobs will be affected by creation or displacement during the period. These are modeled employer expectations, not a guaranteed world total or a timetable for an individual company.

The AI component is two-sided. The WEF jobs outlook estimates that AI and information-processing technologies may create about 11 million jobs and displace about nine million, while robotics and autonomous systems are expected to produce a larger net displacement. Fast-growing role families include big-data specialists, fintech engineers, AI and machine-learning specialists, software developers and security-related professionals. Large absolute growth is also expected in care, education, delivery, construction and other areas not reducible to AI occupations.

The ILO–NASK refined global index reaches a complementary conclusion. It estimates that 25% of global employment is in occupations with some exposure to generative AI, rising to 34% in high-income countries. Clerical occupations remain the most exposed, but exposure has increased in some professional work, including financial analysis, programming and advisory activities. The study emphasizes transformation over full replacement because occupations contain tasks that differ in automation potential and because actual adoption depends on technology, organization, cost and institutions.

The OECD analysis of AI and changing skill demand adds an important correction to the idea that everyone must become an AI engineer. Most workers in highly exposed occupations will not need specialized AI-development skills. Management, business-process, social, cognitive and digital capabilities remain relevant. The practical competition is therefore not simply human versus machine. It is between operating models that combine people and AI with different levels of quality, speed, control and trust.

A four-zone map for 2027

Classify work by how AI changes the economic logic of the role.

Zone	What changes	Typical response	Career question
Create	Adoption generates new systems, risks and services	Build, secure, govern and integrate	Can I enter the new value chain?
Expand	Demand grows for reasons larger than AI; AI raises capacity	Combine domain skill with digital leverage	Can I help a growing sector scale?
Transform	The occupation remains, but routine tasks shrink	Move toward diagnosis, exceptions and ownership	Which higher-value tasks will remain scarce?
Contract	Standardized output needs fewer workers or fewer entry roles	Redeploy, specialize or change pathway	What transition can I start before demand weakens?

An occupation can occupy more than one zone. Software development may create new AI-product work, transform ordinary coding and contract some basic implementation tasks. Accounting may expand with business complexity, transform through automated transaction work and create demand for controls around AI-generated analysis. Use the map at task level, not as a label attached forever to a job title.

Jobs likely to be created or accelerated

AI product, data and model operations

Organizations need people who can turn models into reliable services. That includes AI and machine-learning engineers, data engineers, data scientists, evaluation specialists, model-operations roles, product managers and domain experts who define acceptable performance. Titles will vary. The enduring work is connecting a business problem to data, system behavior, tests, monitoring and accountable decisions.

The opportunity is broader than training models. Many employers will buy or configure existing systems. They will still need people to design retrieval, integrate tools, manage permissions, construct evaluation sets, investigate failures and measure whether the system improves an outcome. A candidate who can show an evaluated workflow may be more useful than one who merely lists several model names.

Cybersecurity, identity and AI assurance

More autonomous systems create more credentials, tool calls, data flows and potential attack paths. The current U.S. Bureau of Labor Statistics 2025–2035 projections project information-security analyst employment to grow 21%, with about 14,100 openings per year and a May 2025 median wage of $129,180. That is a ten-year projection rather than a 2027 promise, but it supports the direction of demand.

New work will include securing AI applications, testing access controls, monitoring model and agent behavior, responding to data leakage, evaluating vendors and connecting AI incidents to existing security and risk processes. Not every role will carry “AI security” in its title. Security analysts, architects, auditors, privacy professionals and technology-risk managers may absorb the responsibilities.

AI governance, compliance and control design

As AI enters consequential processes, organizations must decide which uses are permitted, how systems are documented, who approves deployment, what evidence is retained and how people can challenge outcomes. Roles may appear as AI governance lead, responsible-AI manager, model-risk specialist, AI assurance analyst, policy manager or product-risk partner.

This field rewards cross-domain fluency. A purely abstract policy writer may miss technical failure modes; a purely technical evaluator may miss law, operating authority and customer harm. Useful preparation combines process mapping, risk assessment, evidence design, data governance, stakeholder communication and enough technical understanding to test claims.

Integration, automation and agent operations

The move from chat interfaces to systems that call tools and complete multistep work creates demand for workflow architects, automation engineers, enterprise application specialists, solution consultants and operations leaders who can redesign processes. The work includes defining inputs, permissions, exception paths, service levels and human approval.

Some organizations will create dedicated “agent operations” roles; others will assign the work to product, IT, operations or transformation teams. Do not wait for one standardized title. Search for responsibilities such as workflow orchestration, intelligent automation, AI integration, evaluation, process mining, tool governance and human-in-the-loop design.

Data-center, energy and infrastructure work

AI services depend on physical infrastructure: facilities, networks, electricity, cooling, semiconductors and supply chains. Opportunities can span electrical and power engineering, data-center operations, network architecture, construction management, procurement, capacity planning and sustainability. These roles illustrate why AI labor effects are not limited to software companies.

Infrastructure work is also exposed to cycles and geography. Before training for a specialized path, inspect current projects, employer concentration, local licensing, travel expectations and the transferability of the capability to other facilities or energy systems.

Jobs likely to grow for broader reasons and use AI as leverage

Healthcare operations and management

The current BLS release projects medical and health-services managers to grow 24.2% from 2025 to 2035, adding about 155,100 jobs, with 62,300 openings per year and May 2025 median pay of $123,860. Demographics, service demand and organizational complexity are major drivers. AI may assist scheduling, documentation, analysis and planning, but regulated care, resource allocation and accountable management remain human-intensive.

The career opportunity is not “healthcare plus a chatbot.” It is the ability to improve access, quality, capacity, economics and compliance using reliable systems. Domain entry requirements matter: many roles expect healthcare experience or knowledge of clinical operations.

Management analysis and operational improvement

BLS lists management analysts among occupations expected to add the most jobs in 2025–2035: about 109,200, with a May 2025 median wage of $101,860. AI can accelerate research and documentation, but organizations still need problem framing, stakeholder interviews, process diagnosis, financial reasoning and implementation.

The risk is that generic slide production becomes commoditized. The opportunity is to become evidence-driven: map the process, validate data, calculate economics, design a controlled experiment and help people adopt the change.

Computer and information-systems management

BLS projects about 108,100 additional computer and information-systems manager jobs over 2025–2035 and reports May 2025 median pay of $175,140. These roles combine technical direction, investment choices, talent, vendors, security and operating accountability. AI increases the number of strategic choices rather than removing the need to make them.

Entry usually requires substantial related experience. An early-career learner should not treat the title as an immediate destination. Build the pathway through systems analysis, engineering, security, product, data or technology operations while accumulating decision evidence.

Skilled technical and physical work

Electricians appear among occupations with the largest projected job gains in the current BLS list, with about 75,900 additional jobs by 2035. Healthcare practitioners, construction roles and many service occupations also grow. Physical context, licensing, safety and interpersonal trust make complete remote automation difficult.

AI can still change these jobs through diagnostics, scheduling, training, documentation and augmented-reality support. Resilience comes from combining certified physical competence with digital tools, not assuming physical work is untouched.

Jobs most likely to transform

Software development

Code generation lowers the cost of producing a first draft. It does not eliminate architecture, requirements, security, integration, testing, operations or product judgment. Junior work may change sharply if teams expect one person to supervise more generated output. Entry-level candidates will need to demonstrate debugging, systems reasoning, test design and the ability to explain trade-offs, not only produce code quickly.

Accounting and finance

Transaction classification, reconciliation drafts, variance summaries and document review can be automated. Accountants, financial managers and analysts remain responsible for definitions, evidence, controls, interpretation and decisions. The current BLS most-new-jobs list still projects about 79,400 additional accountant and auditor jobs and 84,900 financial-manager jobs between 2025 and 2035. Growth and automation can coexist.

The transition risk is concentrated in routine production and some entry tasks. Build evidence in controls, scenario analysis, business partnering, systems, regulation and communicating uncertainty.

Marketing, sales and customer service

AI can create content, personalize messages, summarize accounts and answer routine questions. That raises volume and makes generic output less valuable. People remain important for positioning, research design, brand judgment, relationship building, negotiation, escalation and complex customer outcomes.

Customer-service roles dominated by scripted answers face pressure. Roles that diagnose unusual cases, retain high-value customers or improve the service system can become more important. Measure your work by resolution, retention, margin or customer risk—not by content volume.

Legal, research and advisory work

Search, extraction, drafting and comparison become faster. Responsibility for source validity, privilege, professional standards, contextual interpretation and advice remains. The ILO index's increased exposure estimates for some professional occupations are a warning against complacency, not proof that the occupations disappear.

Professionals should build verified workflows, keep sources visible and learn to communicate where the system is uncertain. Clients may pay less for production that appears automatic and more for trusted judgment in consequential cases.

Project and middle-management coordination

Agents can prepare updates, create tasks and summarize risks. Managers who mainly transfer information may face a narrower role. Managers who set priorities, resolve conflict, allocate capacity, coach people and own outcomes remain necessary.

The response is to redesign meetings and reporting before they are redesigned for you. Automate status collection where safe, then spend the recovered time on decisions, dependencies and improvement.

Roles facing the greatest contraction pressure

The WEF report identifies clerical and secretarial roles among the largest expected declines, including cashiers and ticket clerks, administrative assistants, printing workers and some accounting clerks. The ILO likewise finds clerical occupations most exposed to generative AI.

Pressure is strongest when five conditions occur together:

inputs are already digital and standardized;
outputs follow repeatable rules;
exceptions are rare or inexpensive;
errors can be detected automatically;
customers accept a lower-touch experience.

Data entry, routine scheduling, template document production, basic transcription, first-line scripted support and simple content adaptation often meet several conditions. However, local regulation, language, accessibility, trust and exception complexity can slow adoption.

Do not respond by hiding automation or trying to preserve inefficient work. Map the role's adjacent paths. An administrative professional may move toward operations coordination, executive support, project control, customer success, procurement administration or knowledge management. The transition requires evidence: process improvement, stakeholder handling, tools, data quality and decisions—not a renamed résumé.

The TASK-4 audit for your own job

List the 15 to 25 activities that actually consume your month. Classify each:

Transfer: safe to automate or delegate after validation.
Augment: keep human ownership but use AI for speed or coverage.
Strengthen: invest because judgment, trust or domain depth creates value.
Kick-start: a new task made possible by AI, such as monitoring a larger evidence set.

Then score each activity on frequency, time, standardization, error cost, data sensitivity, exception rate and decision authority. Start automation experiments with high-frequency, standardized, low-consequence work. Do not begin with a rare process where failure is catastrophic.

For every Transfer or Augment candidate, define:

approved inputs and data rights;
expected output and quality test;
prohibited actions;
escalation conditions;
named human approver;
monitoring and retained evidence.

The audit should end with a future role statement: “By December 2027, I will spend less time on X and more time owning Y outcome, using Z evidence.” That statement is testable and more useful than “I will become AI-proof.”

A 2027 preparation plan

In the next 30 days

Map your tasks and choose one low-risk AI workflow.
Read five current vacancies in your target role family and code recurring outcomes, not just tools.
Identify one adjacent domain that employers repeatedly connect to the work.
Build a baseline: time, quality, cost and stakeholder satisfaction.

In the next 90 days

Produce one verified before-and-after workflow case.
Learn one durable method: process mapping, statistics, financial modeling, security control design, customer research or experiment design.
Interview three people doing the target work.
Create an artifact that another person can inspect.

In six months

Own a result that crosses team boundaries.
Demonstrate how you handle exceptions and failures.
Build relationships in two adjacent role families.
Recheck current vacancies and official data; discard assumptions that did not survive.

By 2027

Present a portfolio of decisions, systems and outcomes.
Explain how AI changed your workflow without pretending it made the decision.
Show domain depth and cross-functional range.
Maintain at least one transition option before your current role is under acute pressure.

Questions to ask before choosing training

Does the target role have current employer demand, not only media attention?
Is demand concentrated in a few companies or regions?
What education, licensure or experience gates apply?
Which responsibilities are growing inside the role?
Can the program produce a portfolio artifact employers can evaluate?
Does it teach evaluation, security and accountability as well as tool use?
What is the smallest work sample you can complete before paying for a long program?

Avoid training that promises a new title after a short course without acknowledging entry requirements. Use education to close a defined capability gap inside a realistic pathway.

Limits of the forecast

The global sources aggregate employer expectations and occupational exposure. BLS projections cover the United States over ten years, not the precise number of jobs in 2027. Adoption speed depends on cost, regulation, capital, organizational redesign and public acceptance. A role can grow nationally while shrinking in a particular employer or region. Median pay is not an offer, and occupational categories hide variation.

The correct use of this evidence is scenario planning. Build capabilities that perform across several plausible futures: AI workflow design, domain expertise, data and financial reasoning, security awareness, communication, customer understanding and responsible decision ownership.

Final conclusion

The 2027 labor market is likely to reward people who can build and secure AI-enabled systems, help growing sectors scale, redesign workflows and remain accountable for outcomes. It is likely to put pressure on roles dominated by standardized digital production and low-context coordination. Most professionals will experience transformation before replacement.

Do not bet your career on a single prediction. Audit your tasks, move toward scarce decisions, create inspectable evidence and build a transition before you need one. The most valuable question is not which job title survives. It is which combination of domain depth, technology leverage and human responsibility keeps creating value as the title changes.

Continue building cross-functional executive capability

Prepare to lead work that spans strategy, finance, operations, AI and people

The 100% online, self-paced Advanced Executive Program integrates six management domains and a capstone so that professionals can connect technology change to business decisions. It leads to an Advanced Certificate for Executives and is professional education—not an MBA or academic degree.

Explore the Advanced Executive Program in Management & Business Administration →