Rating model
For every profession, a language model (GPT-4 class) evaluates the typical tasks against six key skills — three hard skills (technical, measurable) and three soft skills (interpersonal, context-dependent). Each skill receives a score 0–100 expressing the probability that AI will reliably take over that sub-task in the next 5–10 years.
Scale (applies to every individual skill):
- 0–29 (low): Task requires physical presence, empathy, or unstructured context. AI faces fundamental difficulties.
- 30–54 (moderate): AI can assist, but the final judgement or responsibility stays human.
- 55–74 (high): AI already handles big parts of the task reliably. Routine portions are melting away.
- 75–100 (very high): Task is already largely automated or will be within 1–3 years.
Aggregation to overall score
The overall score for a profession is not the simple average of the six skill scores. It weights skills by estimated importance for the actual job — soft skills carry more weight in relationship-heavy professions (nursing, early-childhood education), hard skills in technical professions (software engineer, industrial mechanic). This weighting is generated by the language model in the rating prompt and is part of the internal calibration.
What the rating accounts for
- Currently available AI capabilities (LLMs, computer vision, robotics) — as of early 2026.
- Foreseeable developments over the next 5–10 years based on current trends.
- The typical job median in English-speaking labor markets — not specializations or top performers.
- Common task scope as found in standard job descriptions (US/UK/CA/AU).
What the rating does NOT account for
- Politics & regulation: Privacy, labor law, AI Acts can delay or accelerate substitution significantly.
- Economic factors: A job can be technically substitutable but economically not — and vice versa.
- Acceptance: Even when AI can perform a task, customer/patient/citizen acceptance matters.
- Individual careers: A senior practitioner in a high-risk field usually has more room than the median.
- Black swans: AGI breakthrough, fundamental robotics jumps, disruption from new business models.
Limitations — read honestly
This rating is not destiny. It's a scientific-pragmatic assessment based on visible AI capabilities today. Three important caveats:
- Language models have bias. If the model was trained on texts where artificial intelligence is associated with substitution, it tends toward substitution-optimism. We sanity-check via cross-comparison of similar professions.
- Today's capability ≠ tomorrow's capability. A score 75 today may be 95 in 3 years — or stagnate at 75 because a fundamental problem stays unsolved (e.g. world-modeling in LLMs).
- Job titles are fuzzy. An accountant at a bank does different work than one at a hospital. The rating targets the median; your concrete job may differ.
Refresh cadence
The database is fully regenerated on a regular cadence — typically every six months or after major AI shifts (new model release, robotics breakthrough). Current rating: early 2026.
What to do with the rating
- Read the skill breakdown, not just the overall score. Which of your tasks get hit first is the more important info.
- Compare with neighboring jobs. Look at how similar jobs in the same category score — relative position matters more than the absolute number.
- Bet on AI-resistant skills. Soft skills with low scores (empathy, negotiation, crisis judgement) are future-proof investments.
- Diversify. A high-risk job with strong specialization is often safer than a median job in the middle.
Open methodology
The rating prompts and the dataset are public on GitHub. If you want to reproduce or critique the rating process, you can. Pull requests with corrections or additional professions are welcome.
Disclaimer. Not career advice. The rating is a model estimate. Concrete career decisions deserve individual counseling — career coaches, mentors, or labor-market services in your country.