A high-pressure signal on one task is a reason to investigate that workflow, not evidence that your role is about to disappear. First check how often the task occurs, what inputs and exceptions it involves, how costly errors are, and how much human review remains. Then look for actual changes in your workplace: approved tool use, reduced task volume, new sign-off duties, or changed staffing. If the rest of your role still matters and you can test safely, redesign the task before making an expensive career move. Prepare an adjacent or larger change when several important responsibilities are shifting or the role no longer fits your practical needs.
What does one high-pressure task flag actually tell you?
A checker flag is a hypothesis about a task, not a verdict on your job. The ILO’s 2025 global index combined occupational task data, expert validation and AI-assisted scoring to estimate where generative AI may affect work. It estimated potential exposure across occupations and explicitly cautioned that the measure does not record actual job losses. Its central unit is work that could change, not an individual worker’s chance of being dismissed. [ILO–NASK index summary](https://www.ilo.org/resource/news/one-four-jobs-risk-being-transformed-genai-new-ilo%E2%80%93nask-global-index-shows)
Audit the flagged activity at workbench level. Write down the exact output it produces, how often it happens, what information it depends on, which cases are unusual, and what a mistake would cost. Add the time needed to check an output and who remains responsible for it. A task that produces a routine first draft may be technically exposed, while missing context, correcting errors, explaining a decision, or obtaining approval still requires accountable human work.
Consider an example: a customer-support representative sees “summarize incoming requests” flagged. That label leaves crucial questions open. Are requests standardized? Does a summary miss urgency or prior promises? Who checks it before a response goes out? Does the employer permit a tool to process customer data? The useful next step is to map this real workflow and its exceptions. Do not put confidential material into an unapproved tool just to see what it can do.
Sources: One in four jobs at risk of being transformed by GenAI, new ILO–NASK Global Index shows
Does an exposed task mean the whole role is unstable?
No. Task exposure, hiring demand, employer adoption and displacement are separate signals. The U.S. Bureau of Labor Statistics discussed occupations potentially susceptible to AI in its 2023–33 projections, yet projected employment growth for software developers and personal financial advisors during that period. Those forecasts are not proof that AI caused growth or that any incumbent is secure; they show why an exposure label cannot substitute for an outlook measure. The figures are U.S. projections, not a forecast for another country or a particular employer. [BLS employment projections](https://www.bls.gov/opub/ted/2025/ai-impacts-in-bls-employment-projections.htm)
Nor does an occupational projection tell you whether your own workload is being redesigned. Compare the checker’s task-level signal with observable local evidence: has the task’s share of your week fallen, has a manager introduced an approved workflow, have review duties increased, or have responsibilities moved elsewhere? A tool can improve one step while leaving domain decisions, exception handling, customer communication and sign-off in the job. It can also change the role enough to matter even if the occupation remains in demand.
The OECD brief analyzes Lightcast online vacancies across ten countries. It ranks occupations by AI exposure using the overlap between their tasks and AI capabilities, then compares the skills requested in vacancies over the studied decade: pooled 2012–13 base years for English-speaking countries and 2018–19 for European countries, against pooled 2021–22 end years. In highly exposed occupations, management, business and digital skills remained commonly requested, while skill demand shifted across the period; the brief also reports small declines in some skills at workplaces more exposed to AI. These are vacancy patterns, not a forecast for one worker or proof that AI alone caused every change. Exposure can coexist with continuing skill demand and workplace change. [OECD vacancy analysis](https://www.oecd.org/content/dam/oecd/en/publications/reports/2024/11/how-is-ai-changing-the-way-workers-perform-their-jobs-and-the-skills-they-require_842aa075/8dc62c72-en.pdf)
Sources: AI impacts in BLS employment projections; How is AI changing the way workers perform their jobs and the skills they require?
When should you redesign the task before changing careers?
Start with task redesign when the broader role still uses valued experience, the activity can be tested safely, and a human can check results at an acceptable cost. A bounded experiment might compare a small set of routine cases with and without an employer-approved tool. Record time, corrections, exceptions and the decisions that still require your knowledge. Agree on what would count as useful before the trial; faster output alone does not prove better service or a safer role.
One field study helps explain why the result can vary by worker. Researchers studying 5,172 customer-support agents reported an average 15% increase in issues resolved per hour after a conversational assistant was introduced, with larger gains among novice and lower-skilled agents. The most experienced and highest-skilled workers saw small speed gains alongside small declines in quality. This was one firm, one tool and a productivity measure. It does not establish job counts or predict another workplace. It does show why an exposed activity may augment some workers’ performance while changing the value of experience differently across a team. [authors’ accessible working paper](https://arxiv.org/abs/2304.11771)
If tools are prohibited or the data cannot safely be shared, the experiment can still be a manual workflow audit: identify repetitive steps, handoffs, review time and failure points, then ask what capability a proposed tool would need to demonstrate. The learning step should match the gap. For using an approved tool in your current field, practice evaluation, data handling and verification on safe examples. A general certificate is not automatically better than a small project that demonstrates those skills in your workflow.
What evidence would justify an adjacent or larger move?
Consider a broader move when multiple core responsibilities are being removed or redesigned, adoption is concrete, accountability or required skills are shifting, and the remaining role no longer fits your needs. OECD workplace case studies documented varied changes, including AI complementing work, automating parts of tasks and creating new tasks. The report also says its cases are not a representative sample, so they illustrate possible mechanisms rather than their frequency across the economy. [OECD workplace case studies](https://www.oecd.org/content/dam/oecd/en/publications/reports/2023/03/the-impact-of-ai-on-the-workplace-evidence-from-oecd-case-studies-of-ai-implementation_b4c2c6ee/2247ce58-en.pdf)
Compare three paths. Leaving or retraining immediately may make sense when the role is already ending or cannot meet your salary, location, health or caregiving constraints, but it carries the largest cost before you know what transfers. Staying and redesigning is usually the more reversible first step when valuable responsibilities remain and safe experimentation is possible. An adjacent move can preserve domain knowledge while shifting toward work such as exception review, client communication, quality control or process coordination, but check actual openings and entry requirements where you live before investing in training.
Use evidence rather than the score to set a review point. For example: after mapping one recurring task and discussing it with a manager or trusted work contact, revisit it after a month or after a documented workflow change. Escalate your preparation if core task volume repeatedly shrinks, responsibility is removed, local opportunities weaken, or review duties become unsustainable. Take salary floor, training time, tuition, location, health and family demands into the comparison; a theoretically attractive pivot may not be workable now.
The verdict is conditional: when only one task is flagged and the rest of the role still has value, audit and test that task first. A checker cannot tell you whether your employer has adopted a tool or whether your job is secure. The free task checker at /ai-job-risk-checker offers task-level change-pressure signals; the paid roadmap at /career-roadmap compares a stay-and-redesign path, adjacent pivots and a larger-change scenario against your experience and constraints. Neither predicts redundancy or guarantees employment or income.
Sources: One in four jobs at risk of being transformed by GenAI, new ILO–NASK Global Index shows; The impact of AI on the workplace: Evidence from OECD case studies of AI implementation
Questions readers ask
Should I quit if the checker gives one of my tasks a high-pressure flag?
Not on that flag alone. Confirm how the task works in your setting, check for actual employer adoption and changes to responsibilities, then test a safe redesign if valuable work remains. Prepare a broader move when several core duties are changing or the role no longer fits your constraints.
Sources and notes
- One in four jobs at risk of being transformed by GenAI, new ILO–NASK Global Index shows
Defines the 2025 global task-based exposure index and states that exposure is potential, not actual job losses.
- AI impacts in BLS employment projections
Shows projected U.S. 2023–33 growth in some occupations BLS identifies as potentially susceptible to AI.
- How is AI changing the way workers perform their jobs and the skills they require?
Analyzes Lightcast online vacancies across ten countries, ranks occupational AI exposure by task-capability overlap, and compares skill demands from pooled country-specific base years (2012–13 or 2018–19) to 2021–22; reports vacancy patterns, not individual outcomes.
- Generative AI at Work (accessible 2024 version)
Reports a field study of 5,172 customer-support agents and a 15% average productivity increase; novice and lower-skilled agents gained more, while the most experienced and highest-skilled workers had small speed gains and small quality declines. It does not estimate aggregate job effects.
- The impact of AI on the workplace: Evidence from OECD case studies of AI implementation
Documents varied workplace task-change mechanisms and says the case studies are not a representative sample.
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