An occupation-level estimate gives you a broad map: it compares a standardized bundle of duties with a model of AI capability. A task-level checker can bring the question closer to your own work by separating activities such as drafting, checking exceptions, and explaining a recommendation. That extra detail helps you choose what to inspect or test first. It still cannot tell you whether your employer has adopted a tool, whether demand for your role will change, or whether your job is at risk. Use the occupation view for context, the task view to form a practical question, and workplace and labor-market evidence before making a large career move.
What can task-level detail show inside one occupation?
A job title bundles activities that may have very different relationships to AI. Consider a knowledge worker who prepares a routine summary, checks unusual cases against local rules, and explains a recommendation to a client. An occupation-level signal may combine all three. A task-level view can prompt separate questions: could a system draft the summary, what evidence must a person check, and who remains responsible for the recommendation? This is an illustration of how to inspect a role, not a claim that a particular employer uses AI for these tasks.
The ILO’s 2025 global GenAI index shows why task descriptions matter even when the published result is occupational. Its researchers gathered worker input on 2,861 tasks drawn from a Polish classification, compared it with expert discussion, and used the task repository to estimate scores for tasks in an international occupational classification. The resulting gradients describe occupational exposure; the ILO says most occupations still contain tasks requiring human input, so transformation is the more likely pattern in its analysis. A task list can therefore reveal a mixed bundle that a single label hides, but it does not observe your actual week or workplace.
The OECD’s 2026 measure answers a different broad question: how closely AI capabilities across nine cognitive, social, and physical domains match occupational requirements over a five-to-ten-year horizon. Its method matters when reading the result. OECD analysts manually rated a selected sample of occupations against O*NET requirements and reconciled ratings through a consensus process; language models then helped extend those mappings across the wider occupational space, with the paper reporting validation of the model ratings. The resulting capability-gap index weights domain gaps by their importance to an occupation. It finds closer matches around routine information processing and codifiable work, while gaps remain around contextual judgment, interpersonal understanding, complex decisions, and responsibility. This is a structured capability mapping, not direct observation of workers or employer outcomes, and its horizon and method differ from the ILO index. Their scores are not readings on a shared scale.
Sources: Generative AI and Jobs: A Refined Global Index of Occupational Exposure; The OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations
What does the occupation-level view still do well?
Occupation-level estimates offer a portable baseline. They let a reader compare broad role families using a common classification, and they may include tasks the reader forgot or described too narrowly in a checker. A self-entered list has the opposite strength: it can reflect the particular mix of duties a worker actually handles, if the list is complete and specific. In short, occupational data travel better across roles; task analysis gets closer to an individual workflow but depends more on the quality of the input.
Keep exposure separate from labor demand. The U.S. Bureau of Labor Statistics’ Occupational Outlook Handbook organizes profiles around typical duties, preparation, pay, and projected employment change. Those are useful broad context for comparing a possible move, but its outlook is not an estimate of AI-caused job loss. Likewise, an AI exposure index describes a capability match or potential task change, not current adoption. In its 2026 brief on exposure indicators, the ILO cautions that these indicators rely on static task descriptions and do not account for adoption constraints or economic feasibility. That is a limit on what an exposure estimate can establish; it is not evidence that a particular employer will adopt a tool or that employment, wages, or demand will move in a specified direction.
This also explains why two exposure views can disagree without either being a personal verdict. They may use different task taxonomies, definitions of capability, time horizons, and assumptions about how a role is performed. The OECD explicitly notes that real effects depend on adoption, regulation, organizational change, and social choices. Treat a mismatch as a prompt to check those assumptions, rather than averaging the numbers into a new score.
Sources: The OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations; Occupation Finder: Occupational Outlook Handbook; Workers’ exposure to AI: What indicators tell us – and what they don’t
What should I do when the two views do not match?
Use a two-pass check before considering a career exit. First, use an occupational estimate to set broad context. Then write down the recurring tasks in your own role and roughly how often they occur. Mark which involve repeatable digital inputs and outputs, and which depend on unusual cases, confidential context, client trust, physical work, or a decision you must explain and own. The ILO’s 2025 research and the OECD’s capability measure both support looking beneath a job label; neither can verify that your description captures your specific position.
Next, check what has actually changed around you. Is a tool approved and used in the workflow, or is it only mentioned in a presentation or job advertisement? Who checks its output, handles exceptions, and bears responsibility? Local hiring and pay are a separate research step: compare recent vacancies and reliable wage data for the roles and locations you could realistically consider. The sources discussed here do not establish those local conditions. The ILO’s 2026 brief cautions that exposure indicators do not account for adoption constraints or economic feasibility, so exposure alone cannot show whether a firm will automate a task. A checker cannot fill that gap by making its result more personal-looking.
The verdict is practical: occupational estimates are better for broad screening; task-level detail is better for choosing what to investigate or improve in your current workflow. Neither is sufficient for a major career decision. If your task list points to a bounded opportunity, start with a reversible experiment, such as testing a permitted tool on a low-stakes draft and measuring review time and error types. If the evidence points toward a larger change, compare adjacent roles and training requirements against your real constraints before committing. The free checker at /ai-job-risk-checker can structure a task-level reflection; its change-pressure signal is not a validated probability of displacement. If you want to compare a stay-and-redesign option with adjacent or larger-change paths, the paid roadmap at /career-roadmap can organize those scenarios around your preferences and constraints, without guaranteeing a job or income. A useful next conversation with your manager is: “Which task in my role are we considering changing, what will remain my responsibility, and how will we judge whether the new workflow works?”
Sources: Generative AI and Jobs: A Refined Global Index of Occupational Exposure; Workers’ exposure to AI: What indicators tell us – and what they don’t; Artificial intelligence; Occupational Employment Projections Data
Questions readers ask
Does a high task-level change-pressure result mean my job will disappear?
No. It identifies tasks that may merit closer examination under the checker’s stated method. It does not measure employer adoption, labor demand, or a personal probability of job loss.
Why might my task result differ from an occupation estimate?
The tools may use different task descriptions, taxonomies, capability assumptions, and time horizons. Your own list may also omit or overstate duties. Check the methods and inputs before interpreting a mismatch.
Which result should guide my next step?
Use occupation evidence for broad context and task detail to choose what to inspect in your workflow. Before a large move, also check actual workplace adoption, local labor evidence, and your salary, location, time, health, and family constraints.
Sources and notes
- Generative AI and Jobs: A Refined Global Index of Occupational Exposure
The ILO describes a 2025 index based on worker input about 2,861 tasks, comparison with expert input, and AI-assisted prediction of scores for tasks in ISCO-08 documentation. It reports four exposure gradients and concludes that transformation is the most likely impact because most occupations include tasks requiring human input.
- The OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations
The 2026 OECD paper maps nine AI capability domains to occupational requirements in O*NET and constructs a capability-gap index weighted by domain importance. It manually rates selected occupations, establishes a consensus reference, benchmarks and validates language-model ratings used to extend mappings. It describes a 5-to-10-year horizon, findings about relative capability gaps, and cautions that actual impacts depend on adoption, regulation, organizational change and social choice.
- Occupation Finder: Occupational Outlook Handbook
The BLS Occupation Finder provides U.S. occupation-level profile fields including entry education, on-the-job training, projected new jobs and growth, and 2025 median pay; the page does not attribute projected changes to AI or establish local hiring conditions.
- Workers’ exposure to AI: What indicators tell us – and what they don’t
The ILO says exposure indicators rely on static task lists, omit adoption constraints such as economic conditions and institutional barriers, and indicate technological susceptibility rather than labor-market outcomes such as employment, wages or demand.
- Artificial intelligence
The ILO states that automating tasks need not produce redundancies: automation or augmentation depends on the task's centrality, integration into work processes, and management's choices about retaining people to perform or oversee tasks.
- Occupational Employment Projections Data
The BLS provides U.S. employment projections, occupational separations and detailed-occupation data on employment, change, openings, education, training and wages. These are broader labor-statistics sources to consult; they do not establish conditions in a particular locality or AI-caused employment change.
Apply this to your own work
See the whole job market at once.
Explore which occupations AI may reshape, then turn the signal into a practical response.
Explore the job map