How artificial intelligence is transforming jobs and tasks

A 2025 study by the International Labour Organization (ILO), in partnership with the Polish institute NASK, analyzed nearly 30,000 professional activities and estimated that one in four workers is in an occupation with some degree of exposure to generative AI. However, this does not mean these jobs will be replaced. In most cases, a transformation of activities is expected, with AI taking on some tasks while professionals retain others.

Which tasks have the greatest automation potential?

According to the ILO, administrative roles remain among the most exposed to generative AI, as many of their tasks involve digital information that can be processed by language systems.

The tasks with the greatest potential for automation include:

• Data entry and organization
• Document transcription and formatting
• Production of standardized texts
• Information summarization
• Document classification
• Preparation of simple reports
• Research and information consolidation

The study also identifies growing exposure in professional and technical roles, such as financial analysts, web and multimedia developers, programmers, and investment consultants. This is because AI models can now perform more specialized tasks, like writing code, analyzing documents, and handling large volumes of information.

Digital activities are more vulnerable to automation

Research shows that the more an activity relies on information already available in digital format, the greater the chance that an AI tool will be involved in its execution.

For example, a marketing professional may still be responsible for a campaign’s strategy while using AI to generate text versions, summarize research, organize data, or adapt content for different audiences.

In finance, technology can assist with data consolidation, document comparison, and preliminary analysis, but decisions involving risk, context, and responsibility still require human involvement.

Anthropic reached a similar conclusion after analyzing millions of interactions with Claude. Its initial Economic Index showed that 57% of observed usage was associated with collaboration or augmentation of human work, while 43% involved tasks performed directly by AI.

Emails, texts, and research also make the list

Another group of activities with automation potential involves professional communication. Writing a first draft of an email, summarizing a meeting, turning notes into a document, reviewing text, or gathering information for a presentation are examples of tasks that can be partially performed by generative systems.

Microsoft also identified changes in this type of work. Its 2025 Work Trend Index, based on a survey of 31,000 knowledge workers across 31 markets, highlighted the growing use of AI agents for research, data analysis, and process execution.

The difference lies in automating a task versus automating a profession. A profession typically involves dozens of different activities. Even when a significant portion of these can be performed by AI, others may require communication, judgment, responsibility, creativity, or physical interaction.

What still depends more on the professional

Automation faces more challenges when an activity requires contextual judgment, complex human interaction, or responsibility for decisions. Tools can assist in preparing, analyzing, and executing parts of these activities, but human participation remains essential.

For workers, a practical way to adapt to this change is to divide their routine into three groups:

1. Repetitive tasks, which can be automated;
2. AI-assisted tasks, where technology speeds up part of the execution;
3. Decision-making tasks, which require evaluation, context, and responsibility.

Implications for workers

The studies do not provide a fixed list of professions that will disappear. What is clearer is a change in the composition of tasks within jobs.

An accountant can use AI to organize documents and identify inconsistencies. A journalist can use it to structure information and transcribe interviews. A programmer can delegate parts of code writing and review. A human resources professional can automate screening and document production steps.

In all these cases, technology acts on specific activities. For workers, knowing the tools is only part of the adaptation. It is also important to know what to delegate to AI, how to verify its answers, and which decisions should remain under human responsibility.