The job search is often described as though it were a private problem.

Update the résumé. Search harder. Apply to more positions. Improve your interview skills. Learn another software program. Network. Follow up. Try again.

There is truth in all of this. But it also leaves something important out.

A person looking for work does not enter a neutral marketplace. They enter a system increasingly organized by software, databases, ranking systems, automated filters, recommendation engines, and artificial intelligence. Before an applicant ever speaks to another human being, a series of technological systems may already have determined what opportunities they see, whether their application is surfaced, and what information about them is considered relevant.

The question, then, is not simply whether technology can make hiring faster.

It is: Who does the technology serve?

And an even more important question follows: What would a job-search system look like if workers had a meaningful say in how it worked?

The applicant is increasingly becoming a data profile

For much of the twentieth century, applying for a job meant presenting yourself to another person.

You submitted a résumé. You spoke with a recruiter. You met a manager. You explained what you had done and what you could do.

The process was never fair. Employers have always possessed more economic power than individual applicants, and discrimination and favoritism did not begin with computers.

But digitization has changed the mechanism.

A modern applicant can become a collection of searchable fields: job titles, skills, education, location, employment history, keywords, assessments, interview scores and other signals. Software can then compare that profile against an employer’s requirements.

There is an obvious attraction here. Employers receive large numbers of applications, and software can help organize them. Job seekers can search thousands of openings that would have been difficult to discover in another era.

But efficiency for the institution does not automatically mean fairness for the worker.

The International Labour Organization’s recent work on artificial intelligence and employment makes an important distinction: AI is not simply a machine that either destroys a job or leaves it untouched. Many occupations are more likely to be transformed than eliminated, while the effects vary substantially across occupations and groups of workers.

That same distinction applies to hiring.

An automated system does not need to reject everyone to change the distribution of opportunity. It only needs to influence who gets seen first.

A faster system can still be an unequal system

Consider two people applying for the same job.

One has spent years accumulating conventional credentials and job titles that are easy for an automated system to recognize.

The other has performed substantial work through temporary employment, caregiving, community activity, informal projects, freelance assignments, interrupted employment, or jobs whose responsibilities do not fit neatly into standardized occupational categories.

A human being might recognize that the second person’s experience is valuable.

A database may not.

This is one of the central problems with automated decision-making: the system does not encounter the whole person. It encounters the representation of the person that its designers decided was useful.

The danger is not necessarily malicious software. A hiring system can reproduce inequality without anyone explicitly instructing it to discriminate.

If historical hiring decisions favored certain educational institutions, career paths, geographic locations, employment patterns, or other characteristics, systems trained or designed around those patterns can inherit their limitations.

The ILO’s recent research on AI in human resource management raises precisely these concerns, pointing to problems involving unclear objectives, biased or incomplete data, and opaque programming.

The problem therefore cannot be solved simply by saying that the machine should be “objective.”

A machine follows an objective.

The political question is: Who selected the objective?

The black box begins before employment

There is a tendency to discuss algorithmic management as something that happens after someone gets a job.

A worker is monitored by software. A platform assigns tasks. An algorithm evaluates performance. A manager receives a computerized recommendation.

But algorithmic management begins earlier.

It can begin when a worker searches for employment.

The OECD defines algorithmic management as software that automates or assists managerial functions traditionally performed by human managers, and its research finds that such systems are already widespread across the countries surveyed. The OECD also identifies concerns about accountability, the ability to understand algorithmic decisions, and protections for workers.

Recruitment is part of this larger transformation.

A job seeker may never know why one vacancy appeared prominently in a search while another did not. They may not know why one résumé was advanced and another was filtered out. They may not know whether an automated assessment affected the outcome, which data was used, or whether a human reviewed the application at all.

This creates a peculiar asymmetry.

The employer may know increasingly more about the applicant.

The applicant may know increasingly less about the process.

That is not simply a technological problem. It is a problem of power.

Transparency cannot mean another privacy policy

The standard response to technological power is often transparency.

Tell people that an algorithm is being used. Publish a privacy policy. Provide a general explanation of the system.

These are useful, but they are not enough.

Imagine telling a worker:

“We use automated systems to evaluate applicants.”

That disclosure does not answer the questions that actually matter.

What information was considered?

What information was ignored?

What characteristics influenced the ranking?

Was the system tested for disparate outcomes?

Can an applicant challenge a decision?

Can a person review the decision?

How long is applicant data retained?

Can an applicant correct inaccurate information?

What happens when an algorithm makes a mistake?

And perhaps the most important question:

Who is accountable when the system gets it wrong?

A technological system should not become a convenient place for responsibility to disappear.

If a hiring manager makes a discriminatory decision, there is at least a recognizable decision-maker. If a vendor’s software recommends that a candidate be rejected, responsibility can become much harder to locate.

The U.S. Equal Employment Opportunity Commission has explicitly noted that AI can be used in recruiting and hiring for activities such as targeted job advertising, résumé screening, recorded-video evaluation and chatbots, while existing employment-discrimination protections can still apply when such technologies are used.

The existence of technology does not eliminate the responsibility of the people and institutions deploying it.

The alternative is not to abolish technology

There is an easy mistake on the other side of this debate.

If technology can reproduce existing inequalities, perhaps the answer is to reject technology altogether.

That is unlikely to work, and it is not necessary.

Digital tools can make employment information more accessible. They can help workers find opportunities outside their immediate geographic networks. They can help employers reach candidates they would otherwise miss. They can reduce administrative work and allow recruiters to spend more time communicating with people.

The question is not whether technology belongs in employment.

It is under what conditions.

Technology can either deepen the distance between workers and institutions or reduce it.

A job platform, for example, can be designed primarily around maximizing applications and employer efficiency. Or it can be designed to help people understand opportunities, compare them, present their skills clearly, and navigate the employment process with greater knowledge.

The difference is not technological.

It is political and organizational.

What would a democratic job search require?

A more democratic employment system would begin with several relatively simple principles.

Workers should know how decisions are made

People do not need the source code of every system.

They do need meaningful explanations of the factors that affect consequential decisions about their employment.

“An algorithm decided” should never be the end of the conversation.

Workers should be able to challenge consequential decisions

An automated rejection should not necessarily be irreversible.

If an applicant believes that their experience has been misunderstood, there should be a meaningful mechanism for review.

Human review should not be treated as an expensive luxury reserved for a small percentage of applicants.

Workers should have control over their employment data

A résumé is not merely a document.

Employment records can reveal education, work history, location, skills, gaps in employment, professional relationships and other information.

Workers should have meaningful rights concerning how that information is collected, interpreted, retained and reused.

Workers should participate in the design of workplace technology

The people affected by technological systems should have a voice in determining how those systems are introduced.

This does not mean every software update requires a referendum.

It means workers should not be treated as passive objects of technological change.

There is evidence that consultation can produce better outcomes. OECD research examining worker consultation around algorithmic management found that consultation among workers, managers and worker representatives can help reconcile productivity goals with improvements in job quality.

The principle should extend beyond existing employees.

Job seekers, too, have a stake in the systems that determine access to employment.

Success should be measured by more than efficiency

The easiest metric for a hiring technology company is often speed.

How quickly can applications be processed?

How many candidates can be screened?

How much recruiter time can be saved?

These questions matter.

But they are incomplete.

A better system should also ask:

How many qualified people were missed?

Who is being filtered out?

Are applicants receiving useful information?

Can people understand the process?

Are candidates treated with dignity?

Does the technology improve the quality of matching rather than simply increasing the quantity of processing?

Efficiency is a means.

It should not become the definition of a good employment system.

From job seekers to participants

The deepest change would be conceptual.

We should stop thinking about job seekers as merely applicants moving through a funnel.

They are workers looking for access to the means by which they will support themselves and their families.

That makes employment technology fundamentally different from an ordinary consumer recommendation engine.

If a streaming service recommends the wrong movie, the consequence is minor.

If a hiring system repeatedly prevents someone from reaching a human decision-maker, the consequence can be a lost income, a missed career opportunity, a period without health insurance, accumulated debt, or the inability to support a family.

The stakes are not symmetrical.

The person looking for work has something to lose.

A different future is possible

The dominant question surrounding AI and employment is usually:

Will AI take our jobs?

It is an understandable question, but it is too narrow.

A better question is:

Who will decide what AI does to work?

The ILO’s recent research emphasizes that AI’s effects depend not only on technological capability but also on how technologies are integrated into work and how the transition is managed. Its research on social dialogue similarly examines how workers and their representatives can influence decisions concerning AI and algorithmic management.

That points toward a more useful debate.

The future of employment will not be determined by algorithms alone.

It will be determined by the institutions that deploy them, the rules that constrain them, the workers who challenge them, the employers who choose how to use them, and the public that decides what kinds of employment systems are acceptable.

The same is true of the job search.

We can build systems that treat people as profiles to be ranked.

Or we can build systems that give people more information, more agency, more transparency and a better opportunity to be understood.

Technology does not decide between those futures.

We do.

The goal should not be a job market in which machines make decisions faster.

It should be an employment system in which technology helps people make better decisions, while the people affected by those decisions retain the power to question them.

That is a much more ambitious goal.

And it is one worth building toward.

---