Essay

Material

3 min reading · 445 words

Work Is More Than Exposure

A task can be exposed to automation without a worker being replaced, helped, or protected.

Predictions about AI and work often begin with a task list. Can a system draft an email, summarize a file, translate a paragraph, write a line of code, rank a candidate, or answer a question? Those are useful questions. They are not yet a description of a job. Jobs include authority, coordination, apprenticeship, responsibility, pace, relationships, discretion, error recovery, and the ability to refuse unsafe or unreasonable demands.

A system may reduce one task while intensifying five others. It may make a person faster while increasing surveillance. It may turn a skilled practice into a sequence of prompts and reviews. It may make a junior worker seem redundant even though junior workers need real tasks in order to learn. It may make a manager responsible for a decision they are no longer given time or information to understand. Exposure is a technical category; work is a social arrangement.

Augmentation needs a distribution question

The phrase 'augmentation' sounds benign because it suggests assistance. But assistance is not evenly distributed. Who chooses the tool? Who receives the saved time? Who bears the new liability? Who owns the output? Who is evaluated against a new speed baseline? Who can say no? The answer may differ across occupations, organizations, countries, and contractual positions. A credible account of benefit has to include those differences.

The International Labour Organization's analysis is useful precisely because it avoids a single replacement headline. It treats task exposure, job quality, and gendered distribution as distinct concerns. That structure should be preserved when people talk about a product launch. A demonstration of capability is not a labor policy.

The human review myth

Organizations often promise a human in the loop. The phrase can mean almost anything. Is the person trained? Do they have enough time? Can they see the input, the model's uncertainty, and the consequences of overriding it? Can they meaningfully disagree? Are they blamed if they defer and blamed if they do not? A nominal reviewer without authority can function as a liability shield rather than a safeguard.

The operator is not optional extends this concern beyond an individual employee. A responsible organization must design jobs, escalation paths, performance measures, and remedies that make human judgment possible. It cannot simply insert a person at the end of a pipeline and call the system accountable.

This is another place where an enchanted story about AI can make people disappear. The question is not whether a machine has a career. The question is what kind of working life a company or public institution is creating for the people around it.

Continue the route

Related work.

Search the field archive

Type a word or browse the suggested paths below.

Open the full libraryTrace an argumentRead source notes

Local reading list

Nothing saved yet. Use “Save +” on any work; your list stays in this browser only.