Methodology

How we calculate AI exposure

The methodology builds on Andrej Karpathy's framework for assessing digital AI exposure across professions, calibrated for the Bulgarian labor market by the CNTS team.

The 1–10 scale

• 1–3 (Low exposure): the work requires physical presence, manual skills, emotional intelligence, or high-stakes decisions.
• 4–6 (Medium exposure): a mix of routine tasks (which AI can take over) and tasks that require human judgment.
• 7–10 (High exposure): the work is predominantly with text, data, code or images — LLMs can already do a large share of the routine parts.

Data sources

• ISCO-08 occupation classification
• Employment and growth data from CNTS Labor Map, based on public statistics for Bulgaria
• Exposure calibration based on current LLM capabilities (Claude, GPT, Gemini) as of 2026

Limitations

Exposure shows how much AI can do of a profession's current tasks, not whether the profession itself will disappear. High exposure often means a shift in assigned tasks, not layoffs.

AI-generated content

The descriptions, examples and recommendations for each profession are generated by AI based on the data above. They have been reviewed for factual accuracy but may contain inaccuracies. For official career advice, consult a professional.

Last updated: April 2026