TaskExposed maps AI exposure at the task level — not the job title level. Because the question isn't “will AI replace accountants?” — it's “which of the twelve things an accountant does every day is AI already doing?”
Occupation-level averages hide the truth. A lawyer who spends 80 % of their time on document review faces a different future than one who spends 80 % in court. We go one level deeper.
We refuse to use fear or false reassurance. Our scores are calibrated estimates with explicit confidence intervals — not marketing copy dressed up as data.
Every formula, threshold, and weighting decision is documented in the Methodology page. If you disagree with our approach, we want to hear why.
We build for the person trying to figure out their next career move — not for the executive trying to justify headcount reductions.
Every week brought a new study: “47 % of jobs at risk.” “AI will replace coders by 2025.” “Your profession is safe.” None of it helped real people make real decisions about their careers.
The underlying academic work — O*NET task mappings, GPT capability rubrics, BLS projections — was solid. But it lived in PDFs and paywalled journals, not in a form a nurse or paralegal could actually use.
TaskExposed is the interface layer between that research and the 150 million workers in the US trying to understand what the next five years look like.
Research begins — reviewing 40+ academic papers on AI and labour markets
Scoring model v1 built and backtested against 2016–2023 employment data
Beta launch with 50 occupations; 12,000 community votes in first month
Expanded to 200+ occupations; Anthropic Economic Index integration added
Task-level granularity released; Compare feature launched
900+ occupations, model v2.3, 140k monthly active users
We don't generate opinions. We synthesize the best available datasets and academic research into a single, navigable interface.
Occupational Information Network maintained by the US Department of Labor. Provides task-level descriptions and time-allocation weights for 900+ occupations.
Real-world AI usage data from millions of Claude interactions, showing which task categories workers are actually automating today.
Eloundou et al. (2023) — the foundational academic paper mapping GPT-4 capabilities to O*NET task rubrics across the full occupation spectrum.
Bureau of Labor Statistics data for employment counts, median wages, and 10-year outlook projections used in the Salary and Growth sections.
Task-based framework for modelling automation's effect on wages and employment — the theoretical backbone of our resilience scoring.
"The Impact of Artificial Intelligence on the Labor Market" — patent-text matching method for assessing AI exposure at the task level.
Built the exposure scoring model, sourced and cleaned O*NET task data, and calibrated capability scores against frontier model benchmarks.
Designed the visual language, built the web application, and obsessed over how to communicate probabilistic data without causing panic or complacency.
Economists and AI policy researchers who pressure-test our methodology and keep us honest about what the data can and cannot say.
We publish our scoring model openly and welcome scrutiny. If you're a researcher, journalist, or policy maker working on AI labour impact, we'd love to talk.
TaskExposed data has been quoted in the New York Times, Financial Times, Wired, and MIT Technology Review. For media enquiries and data licensing, use the press contact below.
press@taskexposed.com