Special report · AI and work

12 min readResearch updated January 2026

AI will not simplytake human jobs.It will redraw the boundary between tasks, skills and value.

The real story is not a contest between people and machines. It is a fast, uneven redesign of work—where routine output becomes cheap, judgement becomes more visible and the ability to learn becomes a core economic skill.

Enter the story Explore the data

40%

Global jobs exposed

IMF

1 in 4

Jobs with some GenAI exposure

ILO

+78M

Net jobs forecast by 2030

WEF

New unit of workHuman × AIDirection · execution · verification · trust

Automation

Routine tasks

Faster · cheaper · scalable

Augmentation

Human capability

Reach · quality · speed

Scarcity shift

Judgement

Context · taste · responsibility

Scroll to continue↓

At 9:03 a.m., Maya opens her laptop. Overnight, the company has added an AI assistant to the tools she already uses. Nothing about her job title has changed. Everything about the rhythm of her work has.

The assistant summarizes yesterday's customer calls, drafts a market brief and turns a rough voice note into a presentation outline. By lunch, Maya has completed work that once occupied most of her morning.

But the saved time does not remove the need for her. It moves the difficult part of the job. She now has to decide which customer signals matter, which claims are trustworthy, what the team should do next and where the machine is confidently wrong.

This is the central tension of AI at work: the technology can make output abundant while making judgement, context, trust and accountability more valuable.

“
The future of work is less about humans versus AI—and more about which humans learn to direct, question and combine AI with real-world responsibility.

Before

Time spent producing

Drafting, searching, formatting

During

Time spent verifying

Checking, correcting, governing

After

Time spent deciding

Prioritizing, persuading, owning outcomes

Chapter 02 · Deconstruct the job

A job is a bundle ofvery different tasks.

AI rarely encounters an occupation as one indivisible object. It encounters emails, analyses, images, forms, conversations and decisions. Some can be automated. Some become faster. Some remain stubbornly human.

Explore illustrative task profiles

Service task profile

Customer support specialist

76/100

AI exposure

68/100

Human + AI upside

Routine questions can be resolved instantly, while emotionally complex or high-stakes cases become more valuable human work.

Likely to automate

−Classifying tickets

−Drafting standard replies

−Summarizing conversation history

Likely to amplify

+Real-time knowledge retrieval

+Suggested next-best actions

+Multilingual communication

Human advantage

✦Empathy

✦De-escalation

✦Commercial judgement

Educational task-mix illustration, not an official probability of job loss. Actual outcomes depend on workflow design, regulation, adoption, cost and demand.

Chapter 03 · The global picture

The headline is disruption.The outcome is still open.

The World Economic Forum projects large-scale job creation and displacement by 2030. These are employer expectations across many forces—not an AI-only forecast—but technology is a central driver.

WEF Future of Jobs 2025 · outlook to 2030

A labour market in motion

Projected job movement

Millions of roles, 2025–2030

Scale / 180M

Created

new roles

170M

Displaced

existing roles

92M

Net change

more roles

78M

170M

Roles projected to be created

Equivalent to 14% of current employment

92M

Roles projected to be displaced

Across structural labour-market change

+78M

Projected net increase

Creation minus displacement

Growth is not only technical

Care, education, delivery, construction and farming are among roles expected to grow strongly in absolute terms.

Clerical work is under pressure

Cashiers, administrative assistants and other routine information-processing roles face sharper decline.

Skills are the bottleneck

Nearly 40% of core skills are expected to change by 2030, making transition capacity as important as technology access.

Exposure is not extinction

One in four jobs may change.Only a fraction sit at the highest exposure.

The International Labour Organization's 2025 task-level index finds that one in four workers are in an occupation with some exposure to generative AI. Yet only 3.3% of global employment is in the highest exposure category.

The ILO's core conclusion is crucial: because most occupations still contain tasks requiring human input, transformation is more likely than full automation.

Exposure measures what technology could touch. It does not measure whether firms will adopt it, whether customers will accept it, or whether a worker will lose a job.

25%Some GenAI exposure

21.7%

Some occupational exposure

A meaningful share of tasks may change, but the occupation is more likely to be redesigned than removed as a whole.

The fragile first rung

AI can remove the practice work that once created experts.

Entry-level employees often learn through research, drafting, formatting, basic analysis and repeated exposure to real cases. These are exactly the tasks generative AI can perform quickly.

A January 2026 IMF analysis found employment in AI-vulnerable occupations was 3.6% lower after five years in regions with high demand for AI skills than in regions with lower demand. The IMF also notes that entry-level roles have higher exposure.

Traditional learning ladder

04Independent judgementOwn complex outcomes

03Pattern recognitionSee many real cases

02Guided practiceDraft, review, correct

01Routine productionResearch, format, summarize

When level 01 disappears, organizations must intentionally redesign apprenticeships—not assume expertise will still emerge.

Chapter 04 · The human edge

When output becomes abundant,discernment becomes scarce.

The strongest careers will combine AI fluency with capabilities that become more—not less—important when machines produce fast, persuasive answers.

JDGJudgementChoose under uncertainty

Selected capability

Judgement

Choose under uncertainty

Recognize incomplete information, weigh trade-offs and decide when the model should not be followed.

Durable skill stack

Domain depthAI fluencyCritical thinkingCommunicationEthics

Interactive role lab

What part of your workis exposed—and what part grows?

Adjust the task mix below. This educational model does not predict job loss; it helps reveal where automation and augmentation may appear inside a role.

Task profile

Describe a typical week

Routine and repeatableRules, templates, predictable inputs and outputs60%Digital and information-basedText, data, code, documents, images and screens75%Judgement and ambiguityTrade-offs, exceptions, incomplete information and risk45%Human interaction and trustCare, persuasion, leadership, conflict and relationships35%

Illustrative result

Automation exposure

Augmentation upside

Your role is likely to be reshaped more than removed. The opportunity is to automate routine output while expanding context, relationship and decision work.

01Automate the repeatable—

02Strengthen verification✓

03Move toward judgement and trust✓

This score is a transparent educational heuristic based only on your slider inputs. It is not a labour-market forecast or career assessment.

Chapter 05 · Turn insight into action

The transition is manageablewhen it is designed.

For individuals

Build a career around leverage, not resistance.

The goal is not to outrun every model. It is to become the person who can frame the problem, direct the system, verify the output and own the result.

Map your tasks

List recurring tasks, then mark which are routine, digital, judgement-heavy or relationship-heavy.

Automate one workflow

Choose a low-risk task and learn the full loop: prompt, context, output, review and correction.

Deepen a human advantage

Pair AI fluency with domain expertise, communication, negotiation, care, leadership or creative direction.

The final frame

AI does not decide the future of work.People decide how AI enters work.

Technology sets new possibilities. Institutions, leaders and workers decide whether those possibilities become better jobs, fewer entry points, wider inequality or a more capable workforce.

The most valuable question is no longer “Will AI take my job?” It is “Which parts of my work should machines do—and what will I become responsible for next?”

Revisit your role profile Read the source notes