Safe From AI?
Seven questions about how you actually spend your working hours — not your job title. We score the share of your week made up of task types AI systems are already observed performing in real work settings, then hand you a verdict: safe, augmented, or automation frontline.
I run the same task-exposure analysis across whole teams, then help leaders plan reskilling, tools and the conversations managers need to have.
How this assessment works
What it measures
This tool estimates task-level exposure: the share of what you actually do in a week that current AI systems can perform or substantially accelerate. It does not predict whether you personally will lose your job.
Exposure and job loss are different things, and conflating them is the most common error in coverage of this topic.
What you put in
Seven questions about the composition of your working week. How much of your time goes to:
- Drafting and editing text
- Analysing structured data
- Rule-based decisions
- Physical or in-person work
- Novel problem-solving
- Relationships and negotiation
- Supervising others
No job titles, no company, no salary.
What you get out
A score from 0 to 100 and one of three verdicts:
Your week is dominated by tasks with low current AI penetration, typically physical, interpersonal, or high-accountability work.
Significant parts of your work can be accelerated by AI, but the tasks sit inside chains where a human remains the bottleneck. This is where most knowledge workers land.
A majority of your week consists of tasks already being performed end-to-end by AI systems in production settings.
How the score is calculated
The weights are anchored to two bodies of research. The first is the task-exposure rubric developed by Eloundou et al., which assessed every task in the O*NET occupational database for whether a large language model could reduce the time required by at least half. The second is observed exposure, a measure introduced by Anthropic economists Maxim Massenkoff and Peter McCrory in 2026, which weights real usage rather than theoretical capability and weights automated uses more heavily than augmentative ones.
We weight toward observed exposure rather than theoretical capability, because the gap between the two is the single most important finding in recent labour research. Eloundou et al. found around 94% of tasks in computing occupations were theoretically exposed. Observed usage in the same occupations ran at about 33%. Scoring against theoretical capability alone produces alarming numbers that describe a world that does not yet exist.
The weight given to each task type
score = Σ(answer × weight) ÷ Σ(answers) × 100 Your seven answers are treated as shares of one week, each share is multiplied by its weight, and the total is expressed from 0 to 100.
Where the data comes from
- O*NET, the U.S. Department of Labor occupational task taxonomy
- Eloundou et al., “GPTs are GPTs” (2023)
- Anthropic Economic Index, and Massenkoff & McCrory, “Labor market impacts of AI: A new measure and early evidence” (2026)
- Brynjolfsson et al., “Canaries in the Coal Mine?” (2025)
Limits and assumptions
This is an estimate based on seven self-reported inputs. It cannot see your employer, your sector’s regulatory constraints, your seniority, or the quality of the residual tasks in your role. Recent research argues that the expertise required by what remains may predict wage effects better than what gets automated.
Exposure measures technical feasibility and current usage, not adoption. Integration cost, workflow redesign, and approval processes sit between capability and deployment, which is why the observed figures trail the theoretical ones by 50 to 65 percentage points across every major occupational category.
Deployment moves. We revise the weights when new releases of the underlying indices land.
AI job quiz FAQ
Is this a prediction that I’ll lose my job?
No. It measures how much of your current work AI systems can already do. Whether that translates into job loss depends on your employer, your sector, and how quickly organisations restructure, factors this assessment cannot see.
Why doesn’t it ask for my job title?
Because job titles are a poor proxy for what people actually do. Two people with identical titles can have entirely different task mixes, and exposure follows tasks. Every serious study in this field measures at the task level for this reason.
What does “augmented” actually mean?
That AI can accelerate meaningful parts of your week without being able to complete the work end-to-end. Most knowledge workers land here. It usually means the composition of your role changes rather than disappears: less drafting, more reviewing and deciding.
Should I be worried if I score high?
A high score means your task mix is exposed, not that change is imminent. The clearest early signal in the research is narrow and specific: workers aged 22 to 25 entering high-exposure occupations are finding jobs around 14% less often than peers in low-exposure roles, with no equivalent effect for workers over 25.
How accurate is it?
It’s an estimate from seven inputs, calibrated against published research. It’s a useful way to think about the structure of your work, not a forecast. Treat the verdict as a prompt for a conversation, not a number to plan around.
Where to go from your verdict
Keep the reading, not just the score.
Save for Later is a personal library for the articles, tools and research you mean to come back to. Put this methodology, the four studies behind it and your next-step tools into one Save Box, and AI tags them so you can find them again when you need them.
Save for Later shares common ownership with Future of Work Insider. Read our editorial policy.