AI Job Exposure Score (Research-Based, Not a Prediction)
This tool is an aggregator of published research (Eloundou et al. 2023, O*NET, Goldman Sachs 2023), NOT a prediction about you specifically, and not career advice. You estimate what share of your work falls into four task types — routine-cognitive, creative-cognitive, interpersonal, and physical — and the calculator returns a transparent, deterministic exposure score using disclosed weights. It tells you how the research literature tends to rate a task mix like the one you describe; it does not know your job, your employer, or your future.
Research-based, not a prediction. This is an aggregator of published research (Eloundou et al. 2023, O*NET, Goldman Sachs 2023), not a prediction about you specifically, and not career advice.
Estimate your task mix
What share of your work falls into each task type? The four shares must total 100%.
AI exposure score (0–100)
53.00
Moderate exposure — a research-based rating of this task mix, not a forecast about your job.
Per-category breakdown
Contribution by task type
How it works
You split your work into four task types whose shares must add up to 100%: routine-cognitive (rule-based analysis, writing, structured information processing), creative-cognitive (novel, non-routine problem solving), interpersonal (managing, negotiating, caring, persuading), and physical (manual execution). The score is a simple weighted sum: each share is multiplied by that task type's exposure weight, and the total is scaled to 0–100. The four per-category contributions always add up to the headline score.
The weights are routine-cognitive 0.8, creative-cognitive 0.5, interpersonal 0.3, and physical 0.1. Their ordering is grounded in the research: Eloundou et al. 2023 ("GPTs are GPTs") find large language models are most exposed to routine language and analytic tasks and effectively unexposed to tasks needing physical execution (their "E0" category), and the O*NET Work Activities taxonomy provides the structure for separating cognitive, interpersonal, and physical work. So exposure falls as work becomes less routine, more human-facing, and more physical.
The magnitude of the weights is a disclosed calibration, not a number copied verbatim from any single study. The values were chosen so that a representative task mix reproduces the studies' headline findings — roughly 19% of workers have at least half their tasks exposed (Eloundou et al. 2023), and about two-thirds of occupations face some AI automation exposure (Goldman Sachs 2023). Because the weighting is fixed and published here, the same task mix always yields the same score, and you can see exactly which task types drive it.
Frequently asked questions
Is this a prediction about my job?+
No. This tool is an aggregator of published research (Eloundou et al. 2023, O*NET, Goldman Sachs 2023), NOT a prediction about you specifically, and not career advice. It reports how the research literature tends to rate a task mix like the one you enter, using fixed, disclosed weights. It has no information about your actual role, your employer, your skills, or what happens next in your career. Treat the number as a way to think about task exposure in general, not as a forecast about your individual job.
Where do the weights come from?+
The ordering of the weights (routine-cognitive highest, then creative-cognitive, interpersonal, and physical lowest) is grounded in Eloundou et al. 2023 ("GPTs are GPTs"), which finds LLM exposure is highest for routine language and analytic tasks and near-zero for physical tasks, and in the O*NET Work Activities taxonomy, which structures cognitive, interpersonal, and physical activity. The specific magnitudes (0.8 / 0.5 / 0.3 / 0.1) are a disclosed calibration chosen so a representative task mix reproduces the studies' headline figures — they are not lifted verbatim from any single study.
What are the limitations?+
Several. First, the task-share inputs are your own subjective self-estimate, and people are not always accurate about how their time actually breaks down. Second, exposure is not automation and not job loss: a task being exposed means AI could assist or accelerate it, not that it will be done without a human or that your role disappears. Third, the four-category model is a coarse simplification of real work, and the weights are a calibration rather than a measured law. Use the result to reflect on which parts of your work are most AI-exposed, not as a definitive verdict.