The optimists who say AI will only create new jobs and everything will be fine are ignoring the specific, documented harm being done to specific categories of workers right now. The pessimists who say AI will eliminate most human employment within a decade are pattern-matching to past technological disruptions that moved much more slowly through society.
The honest answer lives in a more uncomfortable place: AI is already eliminating specific jobs, it will eliminate more, the pace of disruption is faster than most transitions in economic history, and the reassurance that new jobs will emerge is technically true at the aggregate level while being genuinely useless to the individual whose specific role no longer exists.
This post covers what the data actually shows, which jobs are genuinely at risk and which are not, what is happening in real organisations right now, and what a rational, well-informed response to this situation looks like.
What the Data Actually Shows
The
Goldman Sachs Estimate
Goldman Sachs research published in 2024 estimated that AI could automate approximately 300 million full-time jobs globally, affecting roughly 18 percent of work globally and 25 percent of work tasks in advanced economies. The report was careful to distinguish between automating tasks within jobs and eliminating jobs entirely, a distinction that most coverage of the report ignored.
The same research estimated that AI adoption could also raise global GDP by 7 percent over a ten-year period through productivity gains. The economic gains and the employment disruptions are not in conflict. Both can be true simultaneously at the aggregate level. The problem is that GDP growth and job losses are distributed across different people.
The
World Economic Forum Numbers
The WEF's Future of Jobs Report 2025 surveyed employers representing 15 million workers across 22 industry clusters and 55 economies. The headline finding: 83 million jobs expected to be displaced by AI and automation over five years, with 69 million new jobs expected to be created, a net loss of approximately 14 million roles.
The critical detail buried in the methodology: the new jobs being created require significantly different skills from the jobs being displaced. The data entry clerk whose role disappears is not automatically qualified for the AI system auditor role that emerges. The distance between the two is not just a training programme. It is often years of reorientation.
What Is
Actually Happening in Companies Right Now
The most instructive data point in 2026 is not projections. It is what is already happening. Coinbase cut 14 percent of its workforce in May 2026, citing AI-enabled productivity. Oracle has reduced headcount in customer service and documentation roles. IBM paused hiring in functions it expects to be automatable within three to five years.
These are not recession responses. These companies are profitable and growing. They are operational redesigns built around the arithmetic of AI: if one person using AI can produce what three people produced previously, the rational business decision in a competitive market is to employ fewer people for the same output.
This
pattern will expand. The question is not whether it will continue but how
quickly and across how many sectors simultaneously.
Which Jobs Are Actually at Risk
The Oxford Martin School's 2013 research on automation probability has been updated repeatedly as AI capabilities have advanced. The picture in 2026 is more nuanced than the original paper's binary high-risk/low-risk framework suggested, but the directional conclusions are consistent.
High
Displacement Risk: Routine Cognitive Work
The jobs most at risk are those involving high-volume, rule-based cognitive tasks with limited physical and relational components. These include data entry and data processing roles, basic customer service and call centre work, certain paralegal and legal research functions, routine financial analysis and report generation, administrative scheduling and coordination, basic content writing and copyediting, and entry-level coding tasks.
What these roles have in common: the core value they deliver is the transformation of information from one form to another, and AI can perform this transformation faster, at lower cost, and often with comparable accuracy. The economic case for automation is clear and the tools to accomplish it already exist.
Medium
Displacement Risk: Augmentation Rather Than Replacement
A larger category of jobs faces significant change without outright elimination. Teachers, accountants, architects, doctors, journalists, software engineers, and marketing professionals will all find that AI tools change what their jobs require of them substantially.
In these roles, AI handles an increasing share of the technical execution while human value concentrates in judgment, client relationships, ethical reasoning, and the synthesis of complex, ambiguous information that AI handles poorly. The job does not disappear but the skill mix it requires shifts significantly. People who adapt to this shift will find their value increasing. People who do not will find themselves competing with colleagues who have.
Lower
Displacement Risk: Physical, Relational, and Creative Work
The roles most protected from near-term AI displacement have one or more of three characteristics: they require complex physical dexterity in unstructured environments (plumbing, electrical work, surgical procedures, construction); they depend fundamentally on human relationships and emotional intelligence (social work, therapy, pastoral care, senior nursing); or they require genuine creative originality rather than stylistic recombination.
This last category, genuine creative originality, is more contested than the first two. AI can produce creative work that is technically accomplished and commercially viable. Whether it can produce work that is genuinely original in the way that the most significant human creative work is original remains deeply uncertain. The question has philosophical dimensions that cannot be resolved by pointing at current AI capabilities.
The Developing World Dimension That Western Analysis Consistently Misses
Most AI employment research is conducted by organisations based in wealthy economies and focuses on wealthy economy labour markets. The implications for labour markets in Africa, South Asia, and Latin America receive systematic under-attention, and they are substantially different.
Large economies in sub-Saharan Africa and South Asia have a significant portion of their middle-class employment concentrated in precisely the routine cognitive work categories most vulnerable to AI displacement: business process outsourcing, call centres, data processing, and administrative services. The BPO sector employs millions of workers across Nigeria, Kenya, India, the Philippines, and Ghana, often serving multinational clients who have every incentive to automate what they are currently paying human labour to do.
The timeline for this displacement in these markets may actually be shorter than in Western economies, because the offshoring of cognitive work to lower-cost labour markets was always a cost arbitrage strategy, and AI now offers a more extreme version of the same arbitrage without the geographic and legal complexity.
This
is not an argument for pessimism. The same AI tools that threaten certain job
categories also lower the barriers to competing globally in new ones. An
AI-literate freelancer in Nairobi or Lagos operates with essentially the same
tools as their counterpart anywhere. The question is whether the transition
happens fast enough and is supported adequately enough that the affected
workers can navigate it.
The Argument That New Jobs Will Emerge
The historical record is cited repeatedly in this debate: previous waves of technological automation, the industrial revolution, electrification, computerisation, the internet, ultimately created more jobs than they destroyed. The argument is that AI will follow the same pattern.
There are two honest responses to this argument.
The first: it is probably true at the aggregate, long-run level. The historical pattern is real and the reasons for it, new technology creating new categories of demand and new forms of value creation, apply to AI as much as to previous technologies. The AI economy is already generating new roles: prompt engineers, AI trainers, AI ethics officers, machine learning operations specialists, AI product managers.
The second: the previous waves of technological disruption moved across decades. Workers displaced by one technology had time, imperfect and painful time, to develop skills for the next phase. AI is moving faster than any previous general-purpose technology. The question of whether the transition can be managed humanely when the pace is this fast is genuinely open, and the historical analogy may be less reassuring than its proponents suggest.
The AI Vanguard Take:
The 'new jobs
will emerge' argument is used most comfortably by people whose jobs are not in
the displacement category. That does not make it wrong. It does make it
important to notice who is saying it and in what context. The aggregate
employment statistics will almost certainly improve in the long run. The
specific workers displaced in the short and medium run will not benefit from
that reassurance.
What a Rational Response Looks Like
Given the evidence, what should a working person actually do? Not in a theoretical sense but practically, right now, in 2026.
Audit
your role honestly
Look at your job and identify which parts of it involve routine information transformation versus complex judgment, relationship management, or physical skill. The parts in the first category are the most exposed. The parts in the second category are where your defensive positioning should concentrate.
Become
the person in your organisation who knows AI
In most organisations, the person who understands AI tools best is more valuable than the person who ignores them. The structural shift that IBM and Coinbase are executing is being managed, in the organisations that are managing it well, by people who understand both the business function and the AI capability well enough to redesign the workflow. Those people are not being made redundant. They are being promoted.
Invest
in skills that compound with AI rather than compete with it
The skills that increase in value in an AI economy are not the ones AI cannot currently do. They are the ones that make AI more useful. Judgment about what to ask AI to do. The ability to evaluate and edit AI output critically. Domain expertise that allows AI outputs to be contextualised correctly. Client relationship skills that turn AI-produced analysis into trusted advice. These compound. Technical skills that can be fully automated do not.
Do not
wait for the disruption to reach you before responding to it
The
workers most exposed to AI displacement are those in stable roles who have not
updated their skills in years because stability felt like safety. The pattern
is consistent across previous technological transitions and there is no reason
to expect AI to be different. The time to build adaptive capacity is before the
disruption, not during it.
Key Takeaways
•
AI is already displacing
specific job categories, particularly routine cognitive work involving
high-volume information processing. This is happening now, not in a speculative
future
•
The WEF estimates a net
loss of 14 million jobs over five years, with new jobs requiring significantly
different skills from those being lost. Aggregate numbers obscure the
individual transition challenge
•
The roles most at risk
involve rule-based cognitive tasks. The roles most protected involve complex
physical dexterity, emotional intelligence, and genuine creative originality
•
Developing economies with
significant BPO and cognitive outsourcing sectors face a displacement risk that
receives less attention in Western AI research than it deserves
•
New jobs will emerge at the
aggregate level. The pace of AI is faster than previous technological
transitions, making the transition window shorter and the individual adjustment
harder
•
The rational response:
audit your role honestly, become the AI-literate person in your organisation,
invest in skills that compound with AI rather than compete with it, and do not
wait for the disruption to arrive before preparing
Frequently Asked Questions
Will AI
take my specific job?
The honest answer requires knowing what your specific job involves. If your role is primarily high-volume, rule-based cognitive work with minimal physical or relational components, the risk is real and the timeline is shorter than most people expect. If your role involves complex judgment, genuine relationships, or physical skill in unstructured environments, the immediate risk is lower. Most roles fall somewhere between these poles, making the honest answer: some of what you do today will be automated and the value you deliver will need to shift toward the parts that cannot be.
Which
jobs will AI create?
The roles already emerging include AI trainers and evaluators, prompt engineers and AI workflow specialists, AI ethics and governance officers, machine learning operations engineers, AI product managers, and human-AI interface designers. At a broader level, any role that involves managing, evaluating, contextualising, or directing AI systems will grow in demand as AI capabilities expand. The common thread: human oversight of AI processes is a skill category with structural growth.
Should I
be learning to use AI tools right now?
Yes. Regardless of your field or role, basic AI literacy is becoming a professional baseline rather than a differentiator. The Day 7 post on the five AI tools every beginner needs is the practical starting point. The Day 2 tutorial on how to get started with AI covers the first steps. Both are available on The AI Vanguard.
Is the
situation worse in some countries than others?
Yes, in specific ways. Countries with large BPO and cognitive outsourcing sectors face concentrated displacement risk in specific industries. Countries with strong technical education infrastructure and high existing AI adoption are better positioned to capture the new roles that AI creates. Countries with limited social safety nets and high dependence on routine cognitive employment face the most difficult transitions. The global dimension of this is one of the most important and least discussed aspects of AI's employment impact.
Read This Alongside:
The post
on how to make money with AI covers twelve realistic income methods that
position readers on the right side of this transition. The beginner's
toolkit post covers the five tools that build AI literacy from scratch. Both
are on The AI Vanguard.
