Use an AI Task Checker to Prepare for a Career Move

How can the checker help you prepare for a career move?

Use the free checker to prepare a task-specific question list, not to decide whether to leave. Compare its signal with your actual work, look for evidence that workflows or target roles are changing, then weigh an upgrade, an adjacent move, and a larger pivot against salary needs, location, training time, health, and family responsibilities. The aim is one bounded investigation, not a career verdict. Interpret the result with three tests: task fit, change evidence, and move feasibility. This is a practical reasoning sequence, not a validated diagnostic. Ask whether flagged tasks recur in your week and matter to its overall shape. Then ask whether a tool is entering those workflows or local target roles are changing. Finally, check whether a possible response fits your constraints and goals. The checker page presents an exposure map and cautions that exposure is not replacement. The OECD’s *The OECD AI exposure measure* describes potential occupational exposure and says actual effects depend on adoption and organizational choices. A signal can justify timely questions without showing that your employer has adopted AI, demand is falling, or you personally face job loss. Next, separate the signal from workplace evidence and compare feasible moves.

Sources: AI Proof Work: AI exposure map for 342 occupations; The OECD AI exposure measure

What does an exposure signal actually tell you?

An exposure signal tells you that a method has found some overlap between AI capabilities and the tasks or requirements assigned to an occupation. The meaning depends on what the method measured. Read it as a reason to ask a more precise question about work, not as evidence that your employer uses AI, that your role will be redesigned, that demand is declining, or that you personally face job loss. The terms mark separate steps. Capability is what a system can perform under specified conditions. Exposure, sometimes called applicability, is an estimate of how closely that capability fits tasks or requirements in a job. Observed use means a person or organization has actually used a system for work. Adoption is a broader organizational choice to put a tool into a workflow, with rules, access, and resources. Redesign means duties, handoffs, staffing, review, or accountability change around that workflow. Demand concerns how much labor employers or customers seek; displacement is a reduction or loss of work attributable to a change. A task may be technically feasible to assist while remaining unapproved, uneconomic, unreliable, or too costly to verify in a particular setting. These links require evidence at each stage, because capability alone cannot establish an employer decision or a labor-market result. The OECD’s 2026 working paper, “The OECD AI exposure measure,” illustrates one way to construct an occupational signal. It maps AI capability indicators across nine cognitive, social, and physical domains to occupational requirements, then uses an AI Capability Gap index. A lower gap means the capabilities assessed are closer to the profile required by an occupation, and therefore indicate higher potential exposure within that framework. It reports closer capability matches for routine information processing, administrative work, and codifiable tasks, and greater distance for work involving contextual judgment, interpersonal understanding, complex decisions, and responsibility. The paper also says actual effects depend on adoption, regulation, organizational change, and social choice. The International Labour Organization’s 2025 explainer, “How might generative AI impact different occupations?”, uses a different lens. Its global assessment covers 436 detailed occupations in the International Standard Classification of Occupations. Researchers assigned each task a potential automation score from zero to one, then considered both the average across tasks in an occupation and the standard deviation, or spread, of those task scores. The categories therefore distinguish a high average with relatively little variation from a more mixed bundle in which some tasks score higher and others lower. The ILO says few jobs consist of tasks fully automatable with current generative AI, and nearly all occupations include tasks requiring human input. A relatively high occupational average can hide a varied task bundle; a moderate average can still include particular tasks worth examining. Conversely, a task with high assessed potential may take little time, depend on sensitive information, require costly checking, or sit inside a process that an employer has not adopted. The ILO measure focuses on generative AI task automation potential and variation. The OECD measure maps broader AI capabilities to occupational requirements across several domains. Their labels cannot be treated as interchangeable, and their outputs should not be averaged into a more authoritative personal score. Before comparing tools, check what each one measures, which tasks or occupational requirements it includes, what systems it covers, the geography and occupation classification, and when the underlying assessment was made. Then ask what evidence would move the question from “could a system do part of this?” to “is this changing my work?” That next step needs evidence about actual use, implementation and review rules, changed duties or expectations, and labor demand in the relevant market. Until then, the exposure signal can help you decide what to investigate; it cannot tell you what career move to make.

Sources: The OECD AI exposure measure; How might generative AI impact different occupations?

Why should you check the task mix before you read the occupation label?

Because a standardized occupation is an average description, while your decision depends on which tasks you actually perform, how often they recur, how much time they consume, and what quality, context, exception handling, or accountability each requires. Start with your week, then compare it with the checker’s signal. The result can point to a task worth examining; it cannot tell you how central that task is in your role or what your employer will do with it. The distinction matters because jobs are bundles. The International Labour Organization’s “How might generative AI impact different occupations?” assessed 436 detailed occupations globally by assigning each task a potential automation score, then describing both the average score and how much task scores varied within an occupation. Its framework separates jobs with consistently higher task potential from occupations where exposure is uneven. The ILO says few jobs consist of tasks fully automatable with current generative AI and nearly all include tasks requiring human input. This is a modeled assessment of task potential, not observation of your workplace or a forecast for an individual. Still, its average-and-variation method gives a useful reason to look beneath a job title: an occupational average can hide the particular duties that matter to you. The OECD’s “The OECD AI Exposure Measure” approaches the question through capability indicators and occupational requirements across cognitive, social, and physical domains. Its capability-gap index is intended to indicate potential exposure, with a forward-looking horizon; it does not establish whether a particular employer will adopt a system or redesign a role. The ILO and OECD measures therefore should not be treated as interchangeable scores. Their scopes and methods differ. Make a short inventory from a representative week, using rough frequency or time bands rather than pretending to have precise time-use data. Include recurring outputs and less visible work: gathering records, drafting, checking exceptions, answering follow-up questions, coordinating handoffs, and signing off. Beside each task, note whether it appeared to match the checker’s signal, whether the match is uncertain, and what makes the task harder than producing a first draft. Consider sensitive inputs, unusual cases, relationships, physical presence, judgment, error consequences, and who remains accountable. Imagine an analyst whose week includes gathering records, drafting summaries, checking anomalies, and explaining exceptions. A checker might flag summary drafting, but that alone says little about the share of the week it occupies, the reliability required, or the role’s future. If summaries are occasional and most effort goes to resolving unusual cases with colleagues, the signal may identify a bounded workflow to investigate. If repetitive summary assembly recurs throughout the week, it may deserve earlier attention because a change to that process could affect more of the routine work. Neither observation establishes that the employer has adopted a tool, plans to reduce staffing, or will remove the analyst’s duties. For each task, ask what the work requires after a plausible system produces an output. Does someone need to check the source records? Which anomalies require judgment? Who explains the result to a stakeholder, and who is answerable if it is wrong? A task can be technically assisted while its review, exception handling, and accountability remain part of the job. Conversely, a task that appears simple may still be consequential if it is frequent, time-consuming, or a required step in a larger workflow. These are reasons to investigate task conditions, not labels for tasks that are safe or doomed. The most useful result of the audit is one specific mismatch or uncertainty. Turn that uncertainty into a concrete question for a manager or someone familiar with the workflow: “How is this task handled now, and what changes when the output needs correction or an exception?” Keep private or confidential work data out of any unapproved tool. A clear answer will improve the next career conversation more than treating an occupation label as a personal forecast.

Sources: How might generative AI impact different occupations?; The OECD AI exposure measure

What counts as evidence that a workflow or target market is changing?

Treat a workflow as changing when you can point to repeated, observable differences in how work is assigned, checked, measured, or staffed. A system that can draft a summary demonstrates a capability; it does not show that your employer has approved it, connected it to real data, or changed anyone’s responsibilities. An occupation exposure estimate is further removed: it describes assessed overlap between capabilities and work, not a plan for a particular organization. Start inside the workflow. Find out whether a tool is approved for the relevant information and is being used in ordinary work, rather than in a demonstration or isolated experiment. Establish what it does from beginning to end: does it produce a draft for a person to revise, suggest candidates that a worker must investigate, or complete a bounded step before another system or person takes over? Then trace the review. Who checks the output, who handles exceptions, and who remains accountable when an answer is wrong or incomplete? Ask what changed after introduction, and over what period. Has the handoff moved? Are people now expected to review a different volume, meet a different quality threshold, or explain decisions that were previously made elsewhere? Is saved time used to serve more requests, improve service, or reduce a queue, or has the organization changed output expectations or staffing? One announcement can signal an intention; sustained use and altered responsibilities are stronger evidence of implementation. Compare like with like: a revised policy, a changed review queue, or a recurring assignment is more informative than a broad claim that the organization is using AI. If possible, note when the change began and whether it persists across normal work cycles. The OECD policy brief “How Is AI Changing the Way Workers Perform Their Jobs and the Skills They Require?” offers useful context, but not a local forecast. It summarizes Lightcast online vacancies across ten OECD countries, comparing earlier pooled baselines with 2021–22, and combines vacancy skill descriptions with an occupational exposure measure. In the brief’s establishment-level analysis, more AI-exposed workplaces showed a decline of just over three percentage points in the share of vacancies requesting certain management, business, digital, and cognitive skills relative to less-exposed workplaces. The OECD described the magnitude as small, roughly equivalent to one fewer vacancy requesting at least one of those skills at an average workplace posting about twenty vacancies. For a target market, inspect a small, consistent set of current vacancies in the geography where you could realistically work, including remote roles only if they are genuinely available to you. Note which requirements recur across employers and which appear once or read like boilerplate. Compare task descriptions, required credentials, seniority, location, and salary information where disclosed; do not treat a repeated tool name by itself as proof of a new occupation or a hiring pathway. Ask a person familiar with the role what the ordinary week involves and which listed skills are actually used. The OECD working paper “Artificial Intelligence and the Changing Demand for Skills in the Labour Market” is based on the same broad vacancy-analysis program; it should not be counted as independent confirmation of the brief’s findings. Set an observation that would change your assessment before you gather more evidence. For example: “If this workflow is approved and used repeatedly, and review responsibility or output expectations have changed, I will test an upgrade that helps me verify and handle exceptions.” Or: “If several current local roles I would consider repeatedly require a skill I lack, I will investigate that gap before paying for training.” Return to the task signal, your observed work, and local evidence separately. That separation lets you respond to a real shift without treating a capability demonstration, a vacancy dataset, or an exposure label as a verdict on your career.

Sources: How Is AI Changing the Way Workers Perform Their Jobs and the Skills They Require?; Artificial Intelligence and the Changing Demand for Skills in the Labour Market; The OECD AI exposure measure

Which questions turn the result into a useful career conversation?

Ask what the flagged task looks like in your own week, whether a tool has entered that workflow, what checks and accountability remain, and what evidence would make an upgrade or move worth testing. Each question should separate what you know from what you are assuming. The International Labour Organization’s “How might generative AI impact different occupations?” assesses potential automation task by task, then distinguishes occupations by both average exposure and variation. That distinction is useful in conversation: a task signal can identify something to examine, but it cannot tell you how that work is organized in your workplace. Imagine an analyst whose week includes assembling reports, reviewing anomalies, and explaining results. This is an illustration, not a reported case. For report assembly, ask: How often does it happen? Roughly how much of the week does it take? What information goes in, what errors matter, and who verifies the output? For anomaly review, ask whether a tool is being used to surface candidates, who resolves ambiguous cases, and who remains accountable for the decision. For explaining results, ask what context, relationships, or stakeholder trust are needed beyond a draft or summary. These questions make the task concrete without assuming that any tool is approved or that a role is already changing. In a manager conversation, ask which uses are approved for the relevant data, what review rules apply, how quality and turnaround are measured, and whether training time is available. The OECD’s “How is AI changing the way workers perform their jobs and the skills they require?” summarizes online vacancies across ten countries, with observations through 2021–22. It reports shifts in advertised skill demand, including a small decline in some skill requirements at more AI-exposed workplaces; those historical vacancy patterns do not establish your employer’s plans. So ask for observable local evidence: a changed handoff, review standard, workload measure, or responsibility—not a prediction about what management might do. When speaking with someone in a possible adjacent role, ask what an ordinary week contains, which credentials are actually required for the roles they know, and which parts of your experience transfer less easily than they first appear. For your own check, ask which tasks you want to keep doing, which responsibility you want more of, and what skill gap could be tested through a small project or short course before paying for extensive training. Keep questions neutral and specific; a reassuring answer without examples is still uncertain. Do not share confidential work material to test a tool. Bring a short list to a conversation, record what is confirmed and what remains open, then choose one feasible next step.

Sources: How might generative AI impact different occupations?; How Is AI Changing the Way Workers Perform Their Jobs and the Skills They Require?

How can task overlap suggest an adjacent move without promising one?

Use task similarity to discover roles worth checking, then verify their duties, prerequisites, pay, location, and openings. Similarity is a lead for research, not evidence that you qualify or that employers are hiring. A task can travel between occupations while the conditions for doing that work do not. A checker can help you notice overlap; it cannot establish that a new title is feasible for you.

O*NET’s U.S. Job Duties Custom List makes this discovery step concrete. Its instructions say to select a current occupation, review task statements, choose duties you perform now or would like to perform in a next job, and view occupations with related tasks. O*NET says the feature draws on more than 19,000 occupation-specific task statements. The result is a set of leads organized around selected duties, not a ranking by suitability, openings, earnings, or AI resilience. The method also explains why a match is only a starting point: “closely related” tasks share a detailed work activity, while less-related tasks share a broader intermediate activity. A shared activity shows resemblance, but not how much of the new occupation’s week it occupies.

Start with the work you want to carry forward: perhaps interpreting domain-specific records, deciding which exceptions deserve attention, explaining a result to a customer, or checking quality. Compare those duties with a candidate role’s task list, then read its occupational profile. Ask what the role additionally requires: which decisions it owns, what tools or credentials it uses, who relies on its output, and what happens outside routine cases. A similar task may sit inside a different bundle of responsibility. A title that sounds distant may contain work you already do. Compare duties and requirements, not titles alone.

Imagine an experienced analyst whose checker result draws attention to recurring report preparation. A related-duty search might surface roles that also organize information or communicate findings. The analyst’s next question is whether those roles require different subject expertise, stakeholder decisions, or responsibility for recommendations. This example does not establish that a role is open or attainable; it shows how to turn similarity into a question for job descriptions and conversations.

Test transferability in separate parts. Task familiarity is one; credentials, seniority, domain knowledge, professional relationships, and local hiring rules are others. A regulated role may require a qualification even when daily duties feel familiar. A senior move may expect a record of decisions or team responsibility that a task list cannot show. Check pay independently: compare the relevant location and level with your minimum acceptable income, including any temporary reduction during training or transition. If you cannot accept a pay cut, commute, relocation, or schedule change, make that a boundary in the search rather than a detail to solve after choosing a title.

O*NET uses U.S. occupation descriptions, so readers elsewhere should use national occupation classifications and labor-market services for the equivalent comparison. In any location, check several recent postings within the geography and work arrangement you could accept. Look for recurring requirements, distinguish required credentials from preferred wording, and compare advertised duties with the official profile. Postings show advertised openings and requirements; they do not represent every job, guarantee a position remains available, or prove you would be selected. Where possible, ask someone familiar with the target work what a normal week involves and which qualifications employers require. Keep the question specific and do not share confidential workplace material.

The practical output should be a short list of candidate roles and unresolved questions, not a declaration that you should switch. For each, record familiar duties, requirements to verify, location and pay conditions, and the smallest next check: review a credential, compare postings, or arrange an informational conversation. If overlap is substantial but a skill gap seems manageable, a small project or targeted course may test interest before a larger commitment. If roles repeatedly miss your income, location, health, or family constraints, similarity alone is no reason to pursue them. The result has done its job when it makes the next investigation more precise.

Sources: Task Search; Task Search

How do you compare an upgrade, an adjacent move, and a larger pivot?

Compare all three against the same evidence and constraints. An upgrade can preserve expertise while changing how a task is done or checked. An adjacent move may reuse domain knowledge or relationships, yet require a credential or local opening. A larger pivot may fit a strong personal goal but demand more training and a longer transition. Choose by fit and evidence, not exposure rank. The OECD’s AI exposure measure maps capabilities to occupational requirements and says actual effects depend on adoption, organizational change, regulation, and social choice. It cannot decide which move suits one worker. Treat the checker as a reason to compare options, not a recommendation.

For each route, note the task change you have observed, skills or credentials to bridge, and the time and money required. Check the salary floor you must maintain, work availability within your location or remote limits, and whether training fits your health and family responsibilities. Ask whether you would choose the work without the exposure signal, and how easily you could reverse course. A route that works only if you ignore income interruption, travel, care duties, or recovery time is not yet feasible.

| Route | Evidence and skill bridge | Constraint and reversibility check | | --- | --- | --- | | Stay and upgrade | A recurring task or review step is changing; identify a workflow improvement or skill to test in the current role. | Can you learn within affordable time while protecting your salary floor? Can the experiment be adjusted? | | Adjacent move | A target role uses some current tasks or domain knowledge; verify duties, requirements, and local openings. | Does it fit your pay and location? What gap remains, and can you test the work first? | | Larger pivot | A different occupation fits your goal, and evidence points to a specific skills gap. | Can you meet entry requirements and manage training, income, and family impact? What if it proves unsuitable? | This is a decision aid, not a scoring model. A project, conversation with someone in the target role, or review of current vacancies may answer a question before formal training. A course alone does not prove a target role is accessible: verify prerequisites and hiring requirements independently.

OECD Skills Outlook 2025 offers evidence against treating exposure as an automatic instruction to exit, while not showing that everyone can adapt in place. It synthesizes studies from different settings. The report says negotiated adoption with worker consultation and training, including dedicated training time, is associated with better worker outcomes and can help steer AI toward augmentation. It also describes online consultancy-market evidence where demand weakened for less experienced workers, especially on short-term tasks complementing existing language-model capabilities, while demand for more experienced workers on complex tasks remained sustained. These findings concern different populations and work arrangements. They do not provide a universal effect to apply to one employer; they show why workplace conditions and experience matter.

If a task is changing but your role still uses judgment, domain knowledge, or accountability, first test a small upgrade, such as learning to check outputs in a recurring workflow. This is an illustrative action, not a claim that an employer offers training. If your work or goals are shifting, investigate adjacent roles that reuse experience. O*NET’s U.S. Job Duties Custom List lets users select occupation task statements and find occupations with related tasks. It generates leads for comparison; it does not establish local openings, eligibility, pay, or hiring likelihood. Outside the United States, consult national occupational sources and local postings.

A larger pivot merits consideration when it serves a clear goal and sustained evidence supports investigating it. First define the target, verify entry requirements, identify the smallest credible skills bridge, and check how salary, location, training, health, and family constraints can be met. If you cannot name a target or a specific gap, pause an expensive commitment: the next step is research, not broad retraining. Your own goal can justify a change even if the checker does not point there; exposure is not the only reason to move. Prefer the smallest reversible step addressing a demonstrated need, then escalate if workplace evidence, a feasible target, or your priorities warrant it. Reconsider when actual workflow, local requirements, or your constraint floor changes.

Sources: OECD Skills Outlook 2025; The OECD AI exposure measure; Task Search

What should you do with the result this week?

Treat the result as a prompt for one small investigation. Before opening the checker, list recurring tasks from a representative week, using rough time bands. Afterward, mark one task that seems to match the signal and one uncertainty: perhaps it is rare, or the category misses review and exception work. The checker page describes an occupation map; the brief defines its output as transparent task-level change-pressure signals. Neither establishes that your employer adopted a tool or that your position is at risk.

Choose one question that could change your next step. At work, ask whether a relevant tool is approved and used, what review remains, and who handles exceptions. If considering another role, compare its duties and requirements with current postings where you can realistically work. A posting describes advertised requirements, not guaranteed demand. Keep confidential data out of experiments; if policy permits, try a non-sensitive task with human review, .

Set a limit before acting: one conversation or a small block of learning time, with no expensive commitment until you can name the skill gap and why it matters. AI literacy may mean learning to use and check a tool in your field; it does not automatically require a degree or ML-engineering pivot. The free checker can surface questions. The paid roadmap compares staying, adjacent work, and a larger change against preferences and constraints, but guarantees neither employment nor income. End with one observable answer and a reversible next move.

Sources: AI Proof Work: AI exposure map for 342 occupations

Questions readers ask