Where AI Breaks Hiring: Why Candidates Don't Pass the Recruiter Filter

Igor Nikitin, founder of WMT AI and WMT Group, believes that the problem begins not when AI selects personnel, but when it is unconditionally trusted with decisions about people.

“AI can analyze a resume perfectly, but it is not always able to see a strong specialist behind it,” he says.

The expert gives a clear example: the candidate successfully solved complex problems for years, but in another industry or for positions with a non-standard name. His competence fully meets the request of the employer, but there are no necessary keywords in the resume.

“If the system is set to formally coincide with the requirements of the vacancy, such a person risks not even getting an interview,” Nikitin said.

At the same time, it does not devalue the potential of the technology: AI quickly parses hundreds of resumes, compares experience with requirements, conducts initial interviews and structures data, freeing the recruiter from routine for the sake of meaningful evaluation.

At WMT AI, we see the digital recruiter not as a replacement for professional HR, but as a tool that helps scale its expertise. The goal of technology is to expand search capabilities, not narrow them down to a set of keywords. After all, a good recruiter evaluates not only a person’s past experience, but also his potential, motivation, ability to solve specific tasks and matching the team. Therefore, I am convinced that AI should help find and evaluate candidates, but the final decision on hiring should be left to the person.

Aliyev Samir, the head of advertising agency Craft Star, sees the picture in two ways: AI is both an assistant and a problem.

“Assistant because it saves time. Weeds out hundreds of resumes, frees the recruiter from the routine. The problem begins where the algorithm is given the final decision, he says.

The algorithm is far from perfect and does not always understand the context.

“AI does not understand context. He sees the keywords, not the person. The resume says “improvement of operational processes”, in the vacancy – “optimization of operations”. It's different things for a car, and a good specialist goes into the basket. He does not see career logic, does not see achievements behind dry formulations, does not see potential, says Samir.

Moreover, algorithms also translate bias.

“Studies show that language models overestimate candidates with certain names and underestimate older applicants.” Formally, discrimination is prohibited, and the "three to five-year experience" filter cuts off people over fifty automatically. There is no intention, the result is the same, the expert draws attention.

Polina Shebarshova, HRD of Zerocoder University, answers the question succinctly: “Definitely an assistant, but not in everything.”.

Initial selection of responses, she would not trust the entire AI: “the algorithm focuses on formal coincidences and may not notice a strong person who otherwise described his experience.”.

The Shebarshovaya neural network itself helps where you need to quickly and accurately structure data: analyze the results of communication with the candidate, prepare job descriptions, formulate tasks. It is especially helpful when you need a versatile specialist or a professional in a narrow niche, where it is important to accurately describe who you are looking for.

In the HR environment, she recalls a revealing incident: “the manager responded to his own vacancy and received an automatic rejection. The filter that worked against the person who created the vacancy is not an isolated exception, but a symptom.”.

Her personal rule: to view the resume using a neural network, and cover letters and texts of responses always read yourself.

“It is more honest with the applicants, and it leads more accurately to the choice of the “right” candidate. The decision to hire and whether a person will fit into the cultural code of the company must be made by the person. AI is needed for the initial analysis of typical steps that are incorporated in the assessment of a candidate for a particular position, the expert explains.

There are a lot of vacancies, there are a lot of applicants - but there is no employee. Why?

Igor Nikitin calls this a paradox: companies complain about a shortage of specialists, candidates – for silence in response to resumes. It’s not always the lack of suitable people.

“Often employers create too narrow a search box for themselves. For example, they want to find a specialist with a rare combination of competencies, many years of experience and high motivation, but they offer conditions that do not meet these expectations.

The second typical problem is the formal approach to the text of the vacancy: the company lists dozens of requirements, of which not all are necessary for work. As a result, the recruiter, and then the AI, are looking for a perfect match with the job description, not a person capable of solving business problems.

“So I would not start by looking for new tools, but by reviewing the job profile itself. What tasks will the employee have to solve? What skills are really required? And what can be taught already in the process of work? – suggests Nikitin.

And only after that, connect artificial intelligence: let it analyze the requirements, expand the search at the expense of candidates with related experience and quickly process the flow of resumes.

The main mistake is to try using AI to quickly find the ideal candidate from an initially incorrectly composed portrait. The technology will speed up the process, but will not correct the mistake in setting the task, the expert warns.

Samir Aliyev formulates the diagnosis harsher: “The selection is broken.”.

“Companies receive three hundred responses, more than half irrelevant. The recruiter is drowning in noise and does not reach strong candidates. The strong are drowning, too, because they didn’t fit the keywords, he describes.

A language barrier is added: “a person writes about their results, the system looks for matches with the pattern.” It does not go beyond the threshold.” A mailing in one click aggravates the picture: the applicant with the help of AI sends a resume for fifty vacancies, without reading, and the employer receives a stream that is impossible to understand.

Polina Shebarshova is sure that hiring is partly a lottery. The resume cannot show human behavior, stress resistance and a number of professional qualities important in the team. With live communication, the complex is evaluated: both soft skills, and understanding of hard skills, and their symbiosis. The human factor is also involved.

“I can be fascinated by the job seeker’s story, the way he talks, the way he answers uncomfortable questions. But the hiring manager will be disappointed with the communication. Such discrepancies happen more often than we would like. Because of this, the speed of hiring and closing a vacancy can be delayed, she admits.

The expert lists the patterns that give the “test” vacancy: it hangs for more than a month with an unchanged description – there is probably no urgent need; there are no clear tasks, a salary fork and an understandable leader – a reason to be wary; the behavior of the company after the response – long pauses, postponements of stages, unwillingness to name the terms. “A real position, under which there is a budget and a head, usually closes faster,” says Shebarshova.

You can check the reality of the vacancy yourself: check it with the company’s website, read reviews of candidates who have been interviewed, and at the first contact directly ask whether the position is open now, who is the head and in what time it is planned to leave the employee. “A normal employer answers such questions calmly,” the expert reassures.

Conclusion: All three agreed on the main thing – AI is good as an input assistant, but not as an arbiter. The last word about a person must still be said by a person.