AI in Recruitment: Who Decides If Your Application Moves Forward?

For the last two years, the biggest question around artificial intelligence and work has been whether AI will take our jobs. Maybe we were looking at the wrong part of the process.

AI may not be replacing you, but something bigger may be happening. It may simply decide whether you get the opportunity, no more and no less, to get your next job.

The story is not simply that technology is replacing you. It is changing what we expect people to be good at. When information becomes easier to find, judgment becomes more important. When a first draft takes seconds instead of hours, knowing what deserves to be in the final version becomes more valuable. The technology changes, and so does the human value around it.

But while we are discussing the skills people will need to work with AI, something else is happening on the other side, companies are increasingly using AI to decide who gets closer to that job.

From Efficiency to Making Decisions

Recruitment was an obvious place for AI to enter. It is repetitive, involves a lot of manual data, and is extremely time-consuming. AI can help recruiters write job descriptions, search for candidates, identify skills, summarize résumés, match profiles with positions, prepare interview questions, and organize information that previously required hours of manual work. Sounds great, right?

The adoption isn't just an idea; it's already happening. LinkedIn's 2025 Future of Recruiting research found that 37% of recruiting organizations were actively integrating or experimenting with generative AI, up from 27% the year before. Those using or experimenting with it reported saving around 20% of their workweek.

That is a significant productivity gain, and there is nothing inherently problematic about it. The problem starts somewhere else.

AI systems can be used to screen or filter applications, evaluate candidates, and produce rankings or recommendations. At that point, efficiency starts touching access. The problem? A candidate can be removed from consideration before having a conversation, explaining an unusual career transition, or showing something that was not obvious on a résumé.

Saving two hours is one thing. Deciding which candidates deserve those two hours is another, and we are all starting to recognize not only the difference, but the real impact this could have on the job market.

Why Recruitment AI Is High-Risk

The European Union's AI Act classifies certain AI systems used in employment as high-risk. That classification is interesting because recruitment sits alongside areas where an automated decision can have a meaningful impact on someone's life. Getting access to employment is not the same as receiving a movie recommendation or asking a chatbot to summarize an article.

Employment law has spent decades defining boundaries around how humans make hiring decisions. Companies cannot legally make employment decisions based on certain protected characteristics. Different jurisdictions have added rules around accessibility, transparency, candidate information, and discrimination.

But AI introduces a strange new question: What happens when there is no obvious human decision to analyze?

Imagine a system gives one candidate a score of 82 and another 61. The recruiter interviews the first and never sees the second. Who actually made that decision?

The recruiter could argue that the system only provided a recommendation. The technology provider could argue that the employer decided how to use it. The company could argue that the final hiring decision was ultimately made by a person.

All of those things can be true while the candidate with 61 never had a human look at the application.

This becomes even more complicated when we consider bias. Algorithms do not arrive without context. They are designed around criteria, data, and assumptions. If historical hiring patterns contain biases, systems can potentially reproduce those patterns at a scale no individual recruiter ever could.

There is also a contradiction that companies should pay attention to. Businesses increasingly say they want transferable skills, unconventional backgrounds, and people who bring different perspectives. But if an automated system becomes particularly good at identifying what a successful candidate historically looked like, it could potentially make unconventional candidates harder to see.

We could end up asking people to differentiate while building technology designed to recognize similarity.

New Tools, New Problems

The challenge is not simply whether AI should be used in recruitment, but what happens when a tool designed to create efficiency starts influencing decisions about people, including some of the most consequential ones: where you will work, how much you will earn, or whether you will even get the opportunity to be considered.

We know work still defines a significant part of our lives. It determines how we spend most of our days, our income, the opportunities we have access to, and, for many people, part of their identity. Deciding who gets access to a job has never been a minor decision.

Humans are not perfect at making those decisions. Recruiters have biases, make mistakes, and sometimes reject the right candidate for the wrong reason. But at least there is someone accountable for those decisions, and employment regulations have spent decades establishing limits around what employers can and cannot consider. When was the last time we saw a bot sued for workplace discrimination?

We have moved forward, but new tools bring new challenges. If AI helps us make decisions faster, but those decisions are wrong, biased, or based on criteria we don't fully understand, what exactly have we improved?

And speed can make the problem bigger. One recruiter can make a bad decision about one candidate. A system can potentially repeat the same bad assumption across hundreds or thousands of applications before anyone realizes something is wrong. Automation does not only scale efficiency. It can scale mistakes too.

The challenge, then, is to first understand how these tools are designed, what they evaluate, what information they use, and how much influence they have over the final decision. Transparency becomes particularly important when a candidate never knows whether a person rejected their application or whether a system prevented that application from reaching a person in the first place.

We have spent years asking whether AI will be qualified to do our jobs. Maybe the question we should be asking first is whether AI should be qualified to decide who gets the opportunity to do them.