What is the AI Kill Threshold?

The AI Kill Threshold is a probability (0-100%) indicating how likely your occupation will be replaced by AI. It is calculated using the Swiss Cheese Barrier Model across four independent dimensions: Learnability, Evaluation Objectivity, Risk Tolerance, and Human Presence.

How is the replacement probability calculated?

Using a Weighted Power Mean with r=-2 across four dimensions. Learnability has weight 0.30, Evaluation Objectivity 0.25, Risk Tolerance 0.20, Human Presence 0.25. An additional E+O interaction bonus applies when both knowledge is explicit and evaluation is objective.

What data sources does AIR use?

AIR aggregates from BLS OES 2023 (798 occupations employment data), O*NET (task structure labels), Eloundou et al. 2023 (LLM exposure measurement across 19,265 tasks), OpenAI GDPval (AI vs human expert blind evaluation), and the World Economic Forum Future of Jobs Report 2025.

How does AIR differ from MBTI?

MBTI measures personality. AIR measures occupational AI-replacement risk. Both use four binary dimensions producing a 4-letter code, but AIR's dimensions are grounded in automation economics literature and the output is backed by employment data from 798 real occupations.

What is the current industry-wide AI replacement rate?

As of 2026, the employment-weighted industry-wide AI replacement rate is 24.7%, calculated as Task Exposure (34.8%, Eloundou et al. 2023) times AI Win Rate vs human experts (71.3%, OpenAI GDPval).

← KI-Ersetzungsrisiko-Test machen

ESFP

The Taste Maker

“AI generates a thousand options — you know which one is right”

High Risk

Ersetzungswahrscheinlichkeit

55%–72%

Prognostizierter Jahresbereich

2029–2035

Risikostufe

High Risk

Explicit + Subjective + Flexible + Product: AI can learn, errors are tolerable, only results matter, but quality is subjective

Superkraft

Aesthetic judgment and cultural intuition machines can't learn from data

Schwäche

AI is already generating content that passes the taste test

Vier Barriere-Dimensionen

Lernbarkeit — Explizit (E)

Kann KI das Wissen und die Fähigkeiten erwerben und erlernen, die Ihr Job erfordert?

Bewertungsobjektivität — Subjektiv (S)

Gibt es bei Ihrer Arbeit eine „richtige Antwort"? Lässt sich die Qualität objektiv messen?

Fehlertoleranz — Flexibel (F)

Wenn die KI einen Fehler macht, sind die Folgen tragbar? Kann man ihrem Output vertrauen?

Personenabhängigkeit — Produkt (P)

Liegt Ihr Wert in „was Sie produzieren" oder in „wer Sie sind"?

Warum exponiert

Knowledge is learnable, errors are tolerable, and only output matters — AI can iterate endlessly at near-zero cost in these three dimensions.

Natürliche Verteidigung

Quality is subjectively judged — "good" depends on taste, context, and cultural nuance. AI struggles with the moving target of subjective standards.

Typische Berufe

Arts, Design, Entertainment, Sports & Media (27) — design/planning, Computer & Mathematical (15)

Beispielberufe-Risiko

Software Developers
AI coding assistants accelerating productivity — and also changing what developers need to know
39
Software Quality Assurance Analysts and Testers
Test generation and regression automated — exploratory and edge-case testing still requires humans
22
Computer User Support Specialists
Tier-1 troubleshooting increasingly automated by AI; complex issues still need human diagnosis
21
Data Scientists
Automated ML pipelines lower the bar — data scientists shift toward problem framing and interpretation
20
Computer Systems Analysts
Requirements translation and integration work — business context requires humans in the loop
18
Network and Computer Systems Administrators
Infrastructure monitoring increasingly automated — complex configurations and security still need humans
16

Test machen, Profil finden