Bulgaria's labor map and artificial intelligence

AI exposureWho is exposed, who is protected

Each rectangle is an occupation group; its size shows how many people in Bulgaria work in it. Red — the group will be heavily transformed by the current generation of AI; green — virtually unaffected.

The only tile with a score of 9/10 on the map is "General clerks" (39K workers, 82% women) — positions that consist almost entirely of structured document and spreadsheet processing. Below them sits a dense layer at 8/10: Economists and accountants (143K), IT professionals (77K), Bookkeeping and payroll clerks (68K), Other clerical support workers (22K). In total, 349K people stand in the 8–9/10 zone — and this is a growing layer toward which Bulgaria's higher education system and private sector are still actively steering people (see the next section).

At the bottom — all 12 groups with a score of 1 or 2/10 — are occupations where AI has little to offer: Building workers and framers (111K), Mining, construction and transport labourers (107K), Refuse workers (82K), Food and woodworking industry (81K), Childminders and care workers (65K), Agricultural labourers (58K), Cleaners and helpers (51K), Subsistence farm workers (32K). Combined: 623K people in physical and care work.

Zooming out, the most exposed are professionals (engineers, scientists, economists, IT, lawyers) — 597K workers, 20% of the entire labor force, average exposure 6.6/10 — and clerical staff — 185K, 7.9/10. The most protected: elementary occupations (cleaners, helpers, refuse workers), 290K, 1.1/10, and skilled trades (construction, metalwork, electricians), 341K, 2.5/10.

Growth 2011–2024The red above is the green here

The same tiles, recolored by the annual rate of employment change from 2011 to 2024. Green = fast-growing group, red = fast-shrinking. The overlap with the first map is striking: the reddest tiles there (most exposed) are the greenest here (fastest growing). This is the r = +0.48 correlation visualized.

But layered on the same map is a separate, older story that Bulgarian media often conflate with the AI narrative, but which is fundamentally different: Bulgaria's vanishing industrial belt. The fastest-shrinking groups are Military officers (-4.7%/yr), Agricultural labourers (-4.7%), Street vendors (-4.5%), Electrical technicians (-3.8%), Food and woodworking industry (-2.8%). None of these groups are exposed to AI (all score 1–3/10). They are disappearing for demographic and structural reasons: the old industrial structure of the 1990s continues to dismantle, and younger workers are not entering traditional labor-intensive sectors.

These two stories — "the growing ones are exposed" and "the low-exposure ones are vanishing for other reasons" — move in opposite directions, but converge on a single alarming conclusion.

In total, the shrinking groups (over 1% decline/year) are losing roughly 310K jobs in low-exposure sectors since 2011. This is the "refuge" into which people from the pressed quadrant would presumably be reallocated if AI displaces them. But it is shrinking faster than it can absorb new people, and the groups within it (cooks, salespeople, food processing, agriculture) are either stagnating or declining. The displaced economist of tomorrow cannot become a cook or electrician — those groups are no longer looking for people either.

Gender and exposureEvery third woman. Every fifth man.

Tiles recolored by gender composition — pink for predominantly female groups, blue for predominantly male, near-white for a 50/50 split. When the absolute numbers are examined, the picture becomes clearer:

Of the total 1,370K women in the Bulgarian labor force, 458K (33.5%) work in occupations with exposure of 7/10 or higher. Of the total 1,563K men, 329K (21.1%) are in the same zone.

Every third working woman in Bulgaria is in a highly AI-exposed occupation. Every fifth working man.

In the opposite direction — low exposure (3/10 or below) — 36.4% of working men versus 25.9% of working women. The gap is about 10 percentage points in favor of men. Not dramatic, but persistent. The most exposed cluster on the entire map — clerical staff (185K workers, weighted exposure 7.9/10, 69% women) — is not an anomaly but the single strongest gender signal in the entire visualization.

At the same time, the map is not symmetric. At the bottom of exposure — where AI has the least to offer — we simultaneously find 100% male "Building workers and framers" (111K), 90% female "Cleaners and helpers" (51K), and 87% female "Childminders and care workers" (65K). Physical and care work is protected regardless of gender. Meanwhile, IT professionals are 24% women and score 8/10 — not a female-dominated sector.

More precisely: men have a wider corridor toward physical labor as an escape from AI; women have two narrow corridors — care work and manual cleaning — that are already saturated. And the middle of the map (5–6/10) — 759K people, or 26% of the market — is heavily female: 33.6% of all working women vs 19.1% of all working men are in this zone. Sales workers (275K, 73% women), teachers (124K, 84% women), doctors and pharmacists (94K, 75% women). For these 460K women in the middle zone, the question is not "will AI replace me" but "will it help me or make my work pointless" — and the answer today is unknown.

Age and exposureThere is no younger generation waiting at the door.

The breakdown of the labor force by age group reveals something that needs to be stated plainly: Bulgaria has no "youth buffer" in any sector. Of 2,796K employed aged 15–64, only 99.5K (3.6%) are under 25 — and this thin layer is distributed almost evenly across the entire map. In the most AI-exposed groups (clerical staff, professionals, technicians, and managers — 1,237K combined), young workers make up 2.5%. In low-exposure physical occupations (skilled trades, machine operators, agriculture, elementary occupations), young workers are 4.4%. The gap is under two percentage points — too small to call a trajectory.

The real story is in the prime-age cohort. 954K people aged 25–54 work in the four most exposed groups — 77% of employment in that zone. These are the people to whom AI will happen over the next 15–20 years, while they still have at least that many years of working life ahead. The classic "retirement exit" is not available to them: only 252K in the same groups are over 55. Even if ten percent of that older cohort retires each year, that frees roughly 25K positions annually — approximately 2.6% of those stuck in the middle.

In Bulgaria's most exposed white-collar occupations, there are four workers aged 25–54 for every colleague over 55. The former have nowhere to go; the latter cannot lead them out.

The breakdown in the low zone offers no surprise either. Agriculture (29.8% over 55) and machine operators (26.8%) are the only sectors where the age structure signals relatively fast natural attrition of positions — but both groups are also shrinking overall (negative annual growth in the "Growth 2011–2024" section above). Managers have 24.5% over 55, which is entirely expected — one does not manage at 22 — but it also means that managerial positions freed by natural departures are just as exposed as the positions that produce them. There is no upward path that leads a professional out of the exposure zone through promotion. Promotion is movement within the zone.

Education and exposure94% of Bulgaria's professionals hold a university degree. That is no longer a protection.

Among the 596K people classified as professionals in the Bulgarian economy — engineers, doctors, teachers, lawyers, IT professionals, accountants — 558.7K (93.7%) hold a tertiary degree. The classification itself is built around that requirement, so the share is unsurprising. What is surprising is this: the group with the highest share of tertiary education in the entire economy has a weighted AI exposure of 6.6 out of 10 — the second-highest exposure on the map.

Among technicians and associate professionals (308K), the tertiary share is 57.2%. Among managers (195K) — 66.8%. Among clerical staff (184K) — 46.3%. This is not a marginal phenomenon confined to one or two sectors. All four of the most exposed occupation groups have a majority or near-majority of university graduates, and that dominance goes hand in hand with AI exposure.

When all groups are combined, the picture sharpens. Of the total 1,094K people in employment with a tertiary degree in Bulgaria, 950K — or 86.9% — work in the four groups with the highest AI exposure. A university degree in Bulgaria today is not a shield against automation. It is an entry ticket to the highest-exposure zone.

The diploma that was once a passport to the middle class is now a receipt for exposure.

The opposite end of the map is instructive. In skilled trades (341K, exposure 2.5/10), among machine operators (365K, 3.9/10), and in elementary occupations (290K, 1.1/10), fewer than 8% of workers hold a tertiary degree (2.8% for elementary occupations, 5.6% for machine operators, 7.5% for skilled trades) — compared to 93.7% for professionals. These are the occupations least affected by AI — yet they are also the ones with no upward mobility without a new degree. Bulgaria's reskilling programs and European lifelong-learning funds have historically aimed to pull people from these three sectors upward into the middle class. The problem of the next decade is the reverse: how do 950K people with tertiary education in exposed professions move laterally — toward care work, skilled manual trades, personal services — without losing the income premium their diploma brings today? No such program exists.