Examine tasks before occupations

The ILO’s 2025 study assesses exposure to generative AI through occupational tasks. It offers a framework for understanding why exposure should not be read as a direct prediction of job losses. Source: ILO.

One role may combine drafting, fact-checking, speaking with customers and taking responsibility for decisions. Faster completion of one task does not make the others, or human input, unimportant.

How should new skills be defined?

We recommend going beyond instruction in a single tool’s interface. Problem definition, data literacy, domain knowledge and output review can transfer across jobs as tools change.

An illustrative learning project could ask participants to summarise a public document, check each claim against original sources and document mistakes. This is a design example, not an implemented GTP Global programme or a claim of measured success.

Five recommendations for preparation

Rather than promise a fixed number of new occupations or guaranteed employment, build the learning and experience needed to recognise changing tasks.

Policy recommendations

  1. Put domain knowledge first

    Understand the health, manufacturing, education or public-service context before accepting a tool’s output.

  2. Build verification habits

    Check sources, dates, calculations and quotations, and record errors or uncertainty.

  3. Protect first work experiences

    Encourage employers to offer meaningful learning tasks and mentoring for new entrants.

  4. Teach responsible use

    Address personal and organisational data, copyright and human oversight.

  5. Assess skills through projects

    Evaluate how a problem was solved, the quality of the result and the participant’s contribution rather than certificates alone.

Sources

Method note: This article reviews the public sources listed below. Its recommendations are GTP Global’s policy analysis, not a new field study or a measurement of causal effects.