Odd Job Persons

← The GenAI exposure gradient

On the International Labour Organization's 2025 global study, the 7 task statements that define Odd Job Persons (ISCO-08 9622) score an average of 0.11 on a 0–1 exposure scale — more exposed than about 4% of the 427 placed occupations. Roughly 0% of its tasks fall somewhere on the exposed part of the gradient, and the typical task lands in the Not exposed band.

Exposure is task overlap, not a verdict. A high score means a generative-AI model can do part of the content of these tasks — it says nothing about whether the work is automated, whether anyone uses AI for it today, or whether jobs are lost. The gradient is scored on the international ISCO-08 system; the rest of Singulariki is U.S. O*NET/SOC, bridged below by an approximate, many-to-many crosswalk.

How its tasks split across the gradient

Each of the 7 scored tasks for this occupation, sorted into the six exposure bands — cool (human ground) to hot (almost fully assistable).

The most-exposed task

“Adjusting doors and windows;”

Scores 0.15 on the 2025 scale. The task of adjusting doors and windows involves significant manual dexterity, precision, and the ability to interact with physical materials, which are aspects that current Generative AI cannot replicate or automate. Tasks that require physical manipulation and hands-on skills, such as "Dismantling and assembling windows and side panes along with their sealing" and "Preparing windows for cleaning, including disassembling the window frame" show adjusted scores of 0.125 and 0.115 respectively, illustrating the low potential for automation when physical manual labor is involved. Generative AI can support with planning tools or providing step-by-step guides, but the actual physical execution of adjusting doors and windows needs human involvement due to the need for nuanced manual adjustments and ensuring precise alignment. The environment in a high-income country like Poland, with wide technology access, does not change the hands-on nature of the task. Given these factors, a score of 0.15 accurately reflects the limited potential for automation by Generative AI for this task, acknowledging some potential for AI assistance in preparatory or planning stages but not in execution.

Moving fastest, 2023 → 2025

“Repairing and painting interior and exterior surfaces such as walls, ceilings and fences;”

Model capability on this task changed by +0.07 in two years — the gradient is not static, it is filling in.

U.S. occupations this maps to

The American O*NET/SOC roles that crosswalk to ISCO-08 9622, biggest by employment first, via the published (approximate, many-to-many) IBS O*NET-SOC ↔ ISCO-08 correspondence. These are the closest U.S. matches — not an asserted one-to-one identity.

• Maintenance and Repair Workers, General
• Installation, Maintenance, and Repair Workers, All Other
• Geothermal Technicians
• Helpers--Installation, Maintenance, and Repair Workers
• Grounds Maintenance Workers, All Other

In context

Part of the 9 - Elementary occupations major group. Return to the full gradient to see how the whole group sits.

Odd Job Persons sit at the 4th percentile of the global GenAI exposure gradient

• Across 427 international occupations scored by the ILO, Odd Job Persons rank in the 4th percentile for GenAI task exposure — overlap with what generative AI can attempt, not a projection of displacement. (ILO / Gmyrek et al. (2025) GenAI exposure gradient)
• About 0% of this occupation's tasks fall into an exposed gradient band. (ILO / Gmyrek et al. (2025))
• Mean task exposure rose by 0.02 between the 2023 and 2025 model-capability snapshots. (ILO / Gmyrek et al. (2025), 2023→2025)
• Its most-exposed task: "Adjusting doors and windows;". (ILO / Gmyrek et al. (2025))

Source: Singulariki — "Odd Job Persons". https://singulariki.com/gradient/9622-odd-job-persons.html
Note: AI task overlap measures what today's AI can attempt, not automation, job loss, or a forecast.