Oyster‘s 2026 Global AI Talent Report points to a mismatch at the center of workplace AI adoption: the skills employees are expected to use are changing faster than the jobs employers formally describe. AI requirements appeared in 15% of roles filled through Oyster’s platform in the first half of 2026, up from 5% in 2022. Yet new expectations often arrive without a revised title, a pay adjustment, or even a clear explanation to the employee.
The research, announced October 14, combines an analysis of more than 15,000 job descriptions with a survey of more than 1,000 HR and people-management professionals. Its most consequential finding is about the ordinary jobs being rewritten around AI, rather than a sudden surge in positions with AI in their names.
For employers, this raises a practical question: if AI fluency matters enough to require it, how should that capability be defined, evaluated and recognized? The report suggests many organizations have begun asking for the skill before establishing a consistent answer.
AI requirements are spreading into familiar jobs
Oyster’s platform analysis counts a role as requiring AI skills when its description explicitly names AI tools, AI skills or AI-assisted workflows as a requirement. That share rose from approximately 5% in 2022 to 6% in 2024, 10% in 2025 and 15% in the first half of 2026.
The comparison measures the proportion of jobs with an explicit requirement, not the absolute number of AI jobs or the percentage of employees using AI. It also captures only roles filled through Oyster, whose customers already have an interest in global employment. Those distinctions matter when interpreting the pace of change.
Engineering remains a prominent source of demand: 23% of engineering roles in the platform data from 2025 through June 2026 required AI skills. But the same period also shows requirements in 17% of marketing coordinator roles, 15% of recruiter roles and 10% of operations coordinator roles.
These jobs illustrate how AI adoption can alter a familiar occupation without creating a new one. A recruiter may be expected to use AI in sourcing or administration; a marketing coordinator may work with AI-assisted content. The important capability is not simply access to a tool, but knowing when its output is useful, when it needs checking and where responsibility stays with a person.
The gap between expectations, communication and pay
Among surveyed respondents whose companies added AI requirements to existing roles in the previous year, 62% said neither titles nor compensation changed. The report identifies a base of 601 respondents for that finding. It should not be read as saying 62% of all employers introduced new AI duties without changing pay.
A separate finding concerns communication. Among respondents whose companies added AI requirements to traditionally nontechnical roles, 46% said the new expectations were not formally communicated. That result has a different base of 646 respondents.
Together, the results describe a management gap. Employees can find themselves judged against expectations that have become part of a manager’s thinking but have not yet become part of an explicit role description. That makes it harder to distinguish a useful new skill from an unspoken increase in workload.
Hiring processes show a related inconsistency. Among respondents whose companies ask candidates what they have personally built with AI, 53% said the answer does not influence hiring or compensation decisions. Asking a question and having a reliable way to interpret the answer are different steps.
The report does not establish that every new AI requirement warrants a raise, or that an unchanged title necessarily means an employee has been treated unfairly. It does show why employers need a clearer connection between the capabilities they request and the decisions they make. Otherwise, AI fluency risks becoming a vague screening phrase rather than a defined part of the job.
Training comes first, but global hiring is part of the response
When AI skills are missing, 53% of surveyed HR professionals said their companies prefer to train existing employees. Another 31% favor combining training with external hiring, while 14% primarily prefer hiring externally.
That preference makes organizational sense: current employees already understand the business, its customers and its constraints. But training has to make room for practice and review. Adding an AI course while leaving expectations, time allocation and evaluation untouched may not resolve the problems identified elsewhere in the research.
For organizations looking outside their workforce, geography becomes another variable. Among respondents whose companies prioritize external AI hiring, 34% had hired internationally because they could not find AI-ready candidates domestically. The corresponding share among those prioritizing internal training was 18%. This is an association between hiring strategies, not evidence that international hiring itself solves a skills shortage.
Oyster’s platform data also shows variation in the share of roles requiring AI skills across hiring markets: 21% in Switzerland, 20% in Poland, 19% in Portugal and 17% in Romania. These are rates within the platform’s country-specific hiring activity. They do not rank countries by the total supply or quality of AI talent, and differences in the mix of roles can affect the comparison.
Among the 211 surveyed respondents whose companies had hired internationally to address a domestic AI talent gap, 59% said they primarily staff AI roles with full-time employees. Another 34% use a mix of employees and contractors, and 7% primarily rely on contractors. The pattern suggests that, for this group, international recruitment is often part of building a continuing team.
Employment technology still needs accountable judgment
Oyster has a commercial interest in cross-border hiring. Its employer-of-record service provides an employment framework for hiring abroad without customers establishing their own entity in each market. That business context is relevant to the report’s emphasis on international recruitment and employment infrastructure.
The company’s integrations platform connects employment workflows with applicant-tracking, HR information and accounting systems. Its documented functions include syncing candidate and employee information, time-off data and approved expenses. An API also supports custom workflows. These capabilities concern moving and managing employment information; they should not be mistaken for evidence that software can independently determine whether an AI-skilled candidate is qualified.
The report’s classification findings reinforce that distinction. Among respondents whose organizations primarily use contractors or freelancers for AI roles, 31% lacked confidence in correct classification under local labor law or had never considered the issue. The comparison was 20% among those primarily staffing AI roles with full-time employees. Uncertainty is not proof of misclassification, but it is a sign that finding talent and establishing the right working relationship are separate challenges.
For HR leaders, the useful next step is to make AI expectations concrete: specify the tasks involved, the approved tools, the standard for checking outputs and the person accountable for the result. Hiring assessments can then test relevant work, while development plans can address identifiable gaps.
What the evidence can and cannot tell employers
The survey was fielded through Prolific in August 2026 among manager-level or more senior professionals responsible for hiring or workforce decisions in the United States, United Kingdom, Canada and Europe. It records respondents’ accounts of company practices, rather than independently audited employment outcomes. The platform analysis covers January 2022 through June 30, 2026; most role and country comparisons use January 2025 through June 2026.
The research therefore provides a view into how a particular group of globally minded employers is adapting. It does not measure the entire labor market, prove that AI caused compensation changes or establish a productivity return from any specific training strategy.
Its strongest message is nonetheless clear. AI can change what a job requires before an organization changes how it describes that job. Closing that gap means giving employees explicit expectations, assessing skills against real work and deciding how new contributions will be recognized. Better tools can support those decisions. Employers still have to make them.