AI Jobs Created and Replaced: Which Roles Are Rising and Which Are Shrinking
As of 23 July 2026, the most evidence-based answer is this: AI is creating a meaningful number of new and newly elevated job titles, but it is not wiping out most occupations outright. Instead, it is splitting the labour market. Roles that build, deploy, govern and secure AI are expanding. Roles built around repetitive clerical processing, standardised transactions and predictable information handling are the ones most clearly shrinking. [5]
This article focuses on the best-supported patterns visible in global employer surveys, official occupational projections and large labour-market datasets. There is no single universal registry of every AI-specific job title worldwide, so the most reliable picture comes from triangulating sources such as the World Economic Forum, LinkedIn’s Work Change Report, Indeed Hiring Lab, the OECD, the ECB and the U.S. Bureau of Labor Statistics. [6]
What the evidence says
The headline global figure still comes from the World Economic Forum’s Future of Jobs Report 2025. Based on survey responses from more than 1,000 large employers across 55 economies, it projects that labour-market transformation over 2025–2030 could create jobs equivalent to 14% of today’s employment, or 170 million jobs, while displacing 8%, or 92 million jobs. That yields expected net growth of 78 million jobs. The same report says technology-related roles are the fastest-growing in percentage terms, while clerical and secretarial work accounts for some of the biggest declines. [1]
LinkedIn’s January 2025 Work Change Report points in the same direction. It says more than 10% of professionals hired today have job titles that did not exist in 2000, and that Artificial Intelligence Engineer is one of the fastest-growing jobs in 15 countries, ranking number one in the Netherlands, the UK and the US. It also reports that the number of companies in the US with a “Head of AI” position has tripled over the past five years. [7]
At the same time, current evidence does not support the most dramatic claim that AI is already causing economy-wide net job destruction. The European Central Bank found in March 2026 that firms making significant use of AI were about 4% more likely to add staff, and firms investing in AI were nearly 2% more likely to hire additional workers. The ECB’s conclusion was that, so far, AI-intensive firms in Europe tend on average to hire rather than fire, even though the longer-term effect remains uncertain. [8]
That does not mean the market is calm. It means the disruption is more selective than universal. LinkedIn projects that by 2030, 70% of the skills used in most jobs will change, with AI acting as a major catalyst. The World Economic Forum similarly estimates that if the global workforce were 100 people, 59 would need training by 2030. [9]
The new AI jobs AI is creating
Build-and-deploy roles
AI engineer and ML engineer are the clearest headline roles. These jobs sit at the centre of model building, fine-tuning, application development and production integration. LinkedIn identifies Artificial Intelligence Engineer as one of the fastest-growing jobs in 15 countries, and AWS now runs a role-based Machine Learning Engineer certification explicitly designed for ML engineers and MLOps engineers, which shows how formalised this job family has become. [10]
MLOps engineer has moved from niche title to mainstream AI operations role. AWS describes its ML Engineer certification as role-based and explicitly says it is designed for ML engineers and MLOps engineers. That matters because once a company moves past prototypes, it needs versioning, evaluation, deployment, monitoring, governance and cost control. In practice, that has made MLOps one of the most durable AI-adjacent careers. [11]
NLP engineer and AI infrastructure engineer are also increasingly visible. Cisco’s 2025 AI Workforce Consortium report places NLP Engineer and AI Infrastructure Engineer among the fastest-growing ICT roles in several G7 labour markets. These are specialised jobs tied to language systems, retrieval pipelines, model serving, compute infrastructure and production performance. [12]
AI architect and AI solutions architect are not always counted separately in official labour statistics yet, but they are clearly real and growing enterprise roles. Microsoft now offers an advanced Agentic AI Business Solutions Architect certification. Microsoft describes the role as an AI-first solution architect who leads enterprise transformation by designing secure, scalable, cross-platform AI solutions and orchestrating agents, prompts, telemetry and governance. That is very close to what many employers mean when they advertise for an AI architect. [13]
AI deployment engineer is another strong example of a role created by the practical rollout of AI. OpenAI’s AI Deployment Engineer, Cyber role shows how quickly customer-facing implementation jobs are emerging. The job description centres on moving enterprises from prototype to durable deployment, evaluating AI workflows, designing safe architectures, defining human approval points and integrating AI into security-critical operations. [14]
Govern-and-secure roles
Head of AI and Chief AI Officer are executive roles that barely existed in most firms a few years ago. LinkedIn reports that the number of companies in the US with a “Head of AI” position has tripled in five years. IBM’s overview of the Chief AI Officer describes it as a relatively new executive function responsible for strategy, implementation, governance and accountability around AI. In other words, AI is not only creating technical jobs; it is also reshaping the C-suite. [15]
AI risk and governance specialist may be the most underestimated AI career of the next few years. Cisco’s AI Workforce Consortium places AI Risk & Governance Specialist at or near the top of the fastest-growing ICT roles in multiple G7 country snapshots, including Canada, the UK, France, Germany, Japan and Italy. This role exists because companies increasingly need people who can convert technical systems into something auditable, legally compliant and organisationally safe. [16]
That governance demand is also being reinforced by standards and regulation. The EU AI Act is the world’s first comprehensive legal framework on AI and applies risk-based obligations to developers and deployers. The NIST AI Risk Management Framework and the ISO/IEC 42001 AI management system standard both push organisations towards formal AI risk, controls, monitoring and governance. Whenever that happens, new professional roles tend to follow. [17]
AI safety, AI security and red-teaming specialists are another genuinely new frontier. OpenAI is currently advertising roles such as Researcher, Automated Red Teaming and Agentic Risk Analyst. Those roles focus on failure-mode discovery, jailbreaks, risk portfolios, mitigations and launch readiness. Google also advertises AI safety and security research roles focused on adversarial machine learning and trustworthy AI. These jobs barely existed as formal career tracks before frontier generative AI systems became mainstream. [18]
Business-and-integration roles
Not every AI role belongs to engineers. Cisco’s consortium identifies AI business consultant as a fast-growing ICT role in multiple G7 markets. These jobs sit between technical capability and commercial implementation. They help organisations identify use cases, redesign processes, estimate ROI and translate AI into business outcomes. [12]
There is also a growing class of AI-fluent product, operations and domain roles. LinkedIn reports that jobs listing AI literacy increased more than sixfold over the past year, even though they still account for only roughly 1 in 500 jobs. That is a crucial detail: the AI labour market is not only producing brand-new titles; it is also upgrading existing product, marketing, HR, legal, finance and operations roles with AI-specific responsibilities. [19]
The truth about prompt engineers
Yes, prompt engineer is real. But it is often overstated as a stand-alone career. LinkedIn says AI literacy skills such as prompt engineering and proficiency with tools like ChatGPT or Copilot have grown rapidly, with AI literacy skills added by members rising 177% since 2023. Yet Indeed’s data shows that GenAI mentions remained rare in many white-collar postings during early 2025. In Canada, only 0.28% of job postings on Indeed mentioned generative AI in late January 2025, and most mentions were concentrated in tech, science and maths occupations. [20]
The practical conclusion is simple. For most workers, prompt engineering is better understood as an important AI skill inside broader roles than as the single best long-term job title to chase. The stronger bets are AI engineer, AI architect, AI governance, AI deployment and domain-specific AI transformation roles. [21]
The jobs that are shrinking
The most important correction to make here is linguistic: most jobs do not suddenly “stop existing”. They usually decline, merge into other roles, or become niche remnants. That is exactly what current evidence shows. The World Economic Forum says the fastest-declining roles globally include postal service clerks, bank tellers, data entry clerks, cashiers and ticket clerks, and administrative assistants and executive secretaries. [22]
Why these roles? Because they rely heavily on structured information, repeatable workflows and standard decision rules. The OECD’s AI exposure work says current AI capabilities are closest to occupations involving routine information processing, administrative work and codifiable tasks, and furthest from work requiring contextual judgement, interpersonal understanding, complex decision-making and responsibility. That is one of the clearest single explanations of why some jobs are under more pressure than others. [4]
The clearest occupation-by-occupation forecasts come from the U.S. Bureau of Labor Statistics. Its 2024–2034 projections show that word processors and typists are projected to decline by 36.1%, telephone operators by 27.5%, switchboard operators by 26.3%, data entry keyers by 25.9%, telemarketers by 22.1%, order clerks by 17.2%, payroll and timekeeping clerks by 16.7%, and file clerks by 15.9%. These are among the strongest signs that routine white-collar processing work is steadily being automated away. [23]
The same pattern holds in customer-facing transaction roles. BLS projects tellers to decline by 13% from 2024 to 2034, cashiers by 10%, and customer service representatives by 5%. These are not purely “AI effects” in isolation, because self-service systems, mobile apps and general automation also matter. But AI is clearly accelerating the replacement of standard interactions, scripted support and repetitive transaction handling. [24]
Back-office support roles are also exposed. BLS projects bookkeeping, accounting and auditing clerks to decline by 6% over the same decade, while secretaries and administrative assistants are projected to show little or no overall change. That flat headline can hide a deeper reality: routine assistant tasks such as note-taking, formatting, scheduling, drafting and information retrieval are exactly the areas where generative AI can already raise productivity sharply. [25]
What is not disappearing wholesale are many higher-judgement professions. For example, BLS projects accountants and auditors to grow by 5%, not shrink, and paralegals and legal assistants to remain broadly flat rather than collapse. That tells you the likely direction of travel: AI may reduce demand for lower-complexity support tasks inside a profession while increasing the value of higher-level analysis, judgement, assurance and client-facing work. [26]
Why the labour market is splitting
The underlying reason is that AI automates tasks more easily than it automates full occupations. An occupation is a bundle of tasks, relationships, accountabilities and context. Current AI performs best when the work is language-heavy, pattern-based, digitally available and easy to standardise. It performs less well where the work depends on incomplete information, messy environments, negotiation, trust, physical presence or responsibility for consequences. [27]
This is also why “human skills” are becoming more valuable, not less. PwC’s 2026 AI Jobs Barometer says the skills needed for the most AI-exposed jobs are changing more than twice as fast as for the least exposed jobs. It also says that new tasks added to AI-exposed roles are 2.5 times more likely to rely on empathy, judgement and creativity. LinkedIn likewise reports that communication was the number one most in-demand skill in 2024 and that AI literacy plus human skills is becoming a key employability combination. [28]
The labour market is also becoming tougher at the bottom of the professional ladder. PwC says the most AI-exposed junior roles are seven times more likely than the least exposed junior roles to demand traditionally senior skills such as leadership and strategic thinking. In the same report, “seniorised” entry-level roles are shown growing by 35% since 2019. That does not mean entry-level work vanishes, but it does mean “easy” junior work is getting compressed. [29]
At the same time, employers are clearly hiring around AI. Indeed’s January 2026 labour market update shows the share of US job postings mentioning AI reached a high of 4.2% in December 2025. Nearly 45% of data and analytics postings contained AI-related terms, and software development, IT systems and scientific R&D postings each mentioned AI at least 20% of the time. Even marketing and HR showed noticeable growth in AI mentions during 2025. That is what a labour market looks like when a technology is spreading through existing occupations, not only creating entirely separate ones. [30]
How to stay employable in an AI-shaped market
The safest strategy is not to chase every fashionable AI label. It is to build a three-part stack: domain expertise, AI workflow fluency, and human judgement. A finance professional who learns AI-assisted analysis, controls and governance is in a much stronger position than a bookkeeper who stays confined to data entry and reconciliation. That distinction matches the BLS projection that bookkeeping clerks decline while accountants and auditors still grow. [31]
It is also smarter to learn systems, not just prompts. The enterprise AI roles with the strongest staying power are increasingly about architecture, deployment, evaluation, monitoring, security and governance. Microsoft’s AI solutions architect certification stresses secure, scalable design, telemetry, vulnerabilities and responsible AI. AWS’s ML Engineer pathway centres on engineering and MLOps. These are more durable skills than simply learning to write neat prompts. [32]
Governance literacy is becoming a differentiator far beyond legal and compliance teams. If you understand the NIST AI RMF, the EU AI Act and ISO/IEC 42001, you are already closer to the kinds of work organisations now need to operationalise AI safely. That can help non-engineers move into AI governance, implementation, policy, assurance and transformation roles. [17]
Finally, plan on continuous reskilling as normal career maintenance, not as an emergency response. LinkedIn expects 70% of the skills used in most jobs to change by 2030, and the World Economic Forum says 59 out of every 100 workers will need training by then. In other words, the competitive advantage is no longer static expertise. It is the ability to keep renewing your expertise. [9]
Key takeaways
· The strongest evidence says AI is creating more specialised roles than many people realise, especially in engineering, architecture, governance, safety and deployment. [33]
· The clearest losers are routine clerical, administrative and transactional roles, not complex judgement-heavy professions as a whole. [34]
· “Prompt engineer” is real but usually better treated as a skill layer inside broader occupations than as the single best long-term job title. [20]
· The most resilient workers will combine AI fluency with domain knowledge, communication, judgement, security awareness and governance literacy. [35]
FAQ
Is AI creating more jobs than it is destroying?
The best global employer-based estimate currently says yes overall, but not evenly. The World Economic Forum projects 170 million jobs created and 92 million displaced by 2030, for net growth of 78 million jobs. That does not mean every country, industry or worker benefits equally. [1]
Which new AI jobs look the most durable?
The most durable AI roles are usually the ones tied to enterprise implementation and control: AI engineer, MLOps engineer, AI architect, AI deployment engineer, AI risk and governance specialist, AI safety specialist and AI security or red-teaming roles. Those jobs exist because organisations need to build, run, govern and secure AI systems continuously, not just experiment with them once. [36]
Is prompt engineer still a good career bet?
As a niche specialist role, yes. As a mass-market stand-alone career, probably not. The stronger evidence suggests prompt engineering is becoming part of broader AI literacy and product, engineering, analytics and operations work rather than standing alone as the main new profession for most people. [20]
Which jobs are most exposed to AI right now?
The occupations most clearly exposed are routine information-processing and administrative roles: data entry, typists, telephone and switchboard operators, order clerks, payroll clerks, tellers, cashiers and some standardised customer service roles. OECD research and BLS forecasts both point in that direction. [37]
Which jobs look relatively safer from AI?
No job is permanently “safe”, but work that depends on contextual judgement, interpersonal trust, complex responsibility, physical presence and non-routine problem-solving looks less exposed than routine office processing. OECD’s exposure measure explicitly places occupations requiring judgement, interpersonal understanding and complex decisions further from current AI capabilities. [4]
What should a non-technical worker do first?
Start with AI fluency inside your existing domain. Learn one or two tools, one workflow you can improve, and one governance or evaluation concept. In practice that means learning how AI changes your job’s process, risks and outputs rather than trying to become an engineer overnight. Employers are increasingly looking for people who can combine AI literacy with domain and human skills. [38]
AI is creating a real layer of new work: engineers, architects, deployment specialists, governance professionals, safety researchers and AI-native business roles. At the same time, it is steadily thinning out clerical, transactional and repetitive office work. The cleanest way to think about the shift is this: AI is not mainly replacing “people” in one sweep; it is replacing repetitive tasks, redistributing responsibility and increasing the premium on technical depth, judgement and adaptability. [39]
For most workers, the opportunity is not to wait for a perfect brand-new title. It is to move towards the parts of work that AI makes more valuable: analysis, oversight, design, integration, validation, communication, client judgement and responsible decision-making. That is where the strongest evidence says the next durable careers will be built. [40]