Will AI replace the recruiter? AI in the assessment of Senior and Junior IT specialists

Introduction: 5 years of observation of the recruitment market

Пройшло вже 5 років, як один з моїх клієнтів звернувся до мене для створення проекту платформи для фрілансерів. З тих пір мені почало цікавити процеси найму: як наймають, чого наймають і чому наймають. Які метрики є, які КПІ враховуються, і взагалі як кадрова політика впливає на роботу компаній.

При написанні своєї магістерської роботи, я запропонував автоматизацію процесів, але тоді вже було зрозуміло, що не за горами той день, коли щось таке як штучний інтелект розпочне змінювати процеси у наймі, і глибоко проникне в середину багатьох компаній. Сьогодні it вакансії стають головним полем для таких технологічних змін.

Цифровізація та «океан даних»

Чим далі, тим більше ми проводимо час в інтернеті, гортаємо стрічку в телеграмі, і більше і більше часу занурюємося у світ онлайн, в тому числі під час роботи, тоді як більшість компаній в США вже інвестують мільярди у діджиталізацію процесів і цифрових систем, які призначені для збору та аналізу даних, що зберігаються з усього що ми скролимо, читаємо і лайкаєм. A technology like AI can transform this bottomless ocean of data into orderly insights. And this process is already happening now, because it was inevitable.

Співпрацюючи з компаніями в США, я частково бачив ці всі великі інструменти, завдяки яким відбувається аналіз кандидатів, побудовані на базі глибоких наукових досліджень вони ставали чимось, що важко впровадити в компанію, важко навчитися працювати і важко потім його масштабувати, і в додаток до цього це все коштувало завжди дорого. AI instead makes hiring more manageable and more transparent and unbiased. Він створює нові цифрові очі, які здатні вишукувати ті самі невловимі ознаки таланту в резюме – чи то senior фахівець, чи junior розробник – і під час співбесіди, від віртуальних інтерв’ю до пасивного збору даних та гейміфікованих оцінювань і об’єктивних нотаток, щоб потім перетворити це все в прогностичні індикатори передбачуваної ефективності вже зараз.

Реальність за ширмою ШІ-рішень

В цілому все, що відбувається зараз з ШІ, мені подобається, і те яку динаміку та роль на ринку найму ШІ починає набирати вже зараз. On the other hand, we have to be realistic.

За всіма статтями і кейсами, які люди часто публікують в лінкед іні та на інших ресурсах, стоїть реальність, і стоїть вона за ширмою сцени, на яких компанії продають свої ШІ рішення. І якщо прибрати цю ширму, то ми побачимо, що ринки талантів залишаються такими ж неефективними, як і раніше: роботодавці насилу знаходять правильну людину на вакантну роль (наприклад, коли потрібен сильний business analyst), а працівники залишаються розчарованими своїми роботами й кар’єрами. Поряд з автоматизацією, яка часто стає ще більш не гуманною, як для шукачів роботи, так і для менеджерів з найму.

Екосистема взаємного обману

Now let’s talk about the other side. Candidates want to get a job and the sooner the better. Often people who are hiring and already working do not understand this, because they already have a job and money. And the candidate at this moment is thinking about how to pay the rent and pay his bills as soon as possible, so the scheme of sending the same resume to everyone and immediately does not disappear anywhere, until tools appeared that allow you to choose your resume targeted for the vacancy.

So what do we get. An employer that uses AI to write a job, and a candidate that uses AI to target their resume to that same job, and another AI that will evaluate that resume. На додачу до того ідеально сформований супровідний лист, і заготовлені відповіді для при-скринінгових інтерв’ю, які можна знайти на будь-якому каналі про найм на ютубі, якими рекрутери діляться з вами як з кандидатами, описуючи те, що вони хочуть почути.

The result is an ecosystem in which both sides are overwhelmed, sometimes confused, occasionally overwhelmed, and mostly exhausted amid a growing crisis of confidence. Although IT work should be transparent, today this process resembles a digital maze.

About the illusion of ideality and outdated data

Nature does not like the ideal, in nature even the most ideal is imperfect, and this is what we like. During this time, AI has created new behavioral signals that are replacing traditional assessment tools. AI-driven structured interviews, text analytics, code evaluations, conversational chatbots, and more, not to mention passive internal and external candidate data collection, which can now generate massive amounts of data at a scale beyond even the wildest imaginations of traditional human-led processes. In theory, this should provide a deeper understanding of human talent and potential, creating a marketplace of pre-assessed candidates that can be automatically matched to open positions both within and across organizations.

But what really? Some interviews that were recorded a year or three ago, stored with the permission and sometimes without the permission of the candidates, take up space in the cloud. From time to time, recruiters scan these videos in the hope of finding the same candidate for a job, such as a qa engineer position, with the goal of getting hiring money with the least amount of effort. What about the candidate? The candidate is no longer at the level he was at when his interview was recorded. And we’re talking about a successful interview, but let’s be honest, even when we prepare, how many of our interviews are failures?

So candidates are mostly required to be ideal in everything, instead, companies, hoping for complete automation of the hiring process, hope that AI will identify unscrupulous candidates, and reduce the cost of hiring and the speed of processing candidates’ resumes, which can be as many as 600 for one vacancy. There is also hope that AI will be faster and more objective in evaluating candidates than a recruiter. This is a huge mistake, why we will understand this further.

Crisis of confidence: “Plastic” market

The labor market as it is now is the same as what is happening in Hollywood now: all actors are the same, all have perfect teeth, perfect smiles, perfect skin, all polished perfect plastic. On the other hand, vacancies that resemble Christmas series on Netflix, in which Christmas bows on poles hang at the same height in the frame, the plot is about nothing, and you don’t even want to continue watching it, and what you have already watched, you want to forget.

Ideality is a bluff that is sold and sold successfully. Employers know this and so do candidates. But as a result, they do not trust each other more and more, continuing to multiply the signals of mistrust. If you do a good search on the Internet, you can already find avatars that will perfectly pass the interview for you, and are also able to learn from previous interviews.

Efficiency vs. accuracy: We trade awl for soap

Chi obviously provides us with fast data processing, which should obviously lead to efficiency, allowing recruiters to process large arrays of candidates at low cost. But there is currently no convincing evidence that AI outperforms established scientifically based assessment tools in terms of accuracy and predictability with a valid system. Likewise, beyond a narrow set of massed low-skilled and highly standardized roles. It is unrealistic to completely eliminate a person from the hiring process. For most important positions, such as project manager, human judgment remains key.

Problem of lack of objective criteria

The main limitation of artificial intelligence in the hiring process is the lack of frameworks or criteria that are objective.In the process of creating a model, we mostly expect to get those candidates who are able to bring maximum value to the company’s work and, of course, make an impression on the recruiter from the first minutes of the interview. But we still have no guarantees for such identification.

We, as people, by nature, due to the circumstances of our attitude and development, have life experience, which in most cases serves as a basis for prejudice. So Chi as a tool is able to detect skills (backend developer, data analyst) or hidden indirect signals of what exactly a candidate possesses, but this does not mean that this particular candidate is able to bring value to the company. It takes a lot of data and time to process it to understand what real value a candidate can provide to the company if we are going to build artificial intelligence models capable of detecting it. AI will certainly be better at predicting the visibility of these skills, but it still doesn’t guarantee that a candidate is capable of being productive in the role they intend to fill.

Ethical issues and experience of Hoorya.eu

Our project Hoorya.eu has started to help implement AI solutions that can help with hiring, and in the last few months we have really achieved good results, but I am always skeptical even of our successes.

After conducting some interviews with recruiters who worked with our solutions, we noticed that those companies that did not receive adequate support and left the hiring decisions exclusively to AI without supervision, were even more biased towards the candidates recommended by the system. We trained our model on the data we had and on the data provided to us by the company, tracking triggers that indicated a successful hire or promotion, tried to track candidates who were in good standing with the company, and looked at those who also did not fit into the company’s framework.

But essentially by selecting candidates according to one scale, we are creating a company that does not have other opinions, and is not flexible in making decisions, taking into account how quickly models learn and how they work, we realized that if we leave everything as it is, then we are simply scaling at a high speed the prejudice against people who want to get a job. And this applies to literally everything: from discrimination, excessive filtering to privacy and non-transparency. Even when remote jobs are offered, not intervening can undermine the employer brand. Instead of strengthening the company’s brand through innovation, we will get a real strategic threat to the brand itself.

Where does AI really help?

Despite many disadvantages, if AI is applied selectively, and limited and managing the risks it carries. It can help in disciplining human judgment. Its limited use will reduce candidate bias and help structure data and consistency in their evaluation. In this case, AI acts as a structured filter before the recruiter’s subjective assessments come into play.

Shea is very good at imposing structure where people tend to improvise:

1. He will ask all candidates the same questions.

2. Eliminates jumping on questions or skipping them.

3. Will apply an objective assessment to each of them.

Such standardization is important. AI-guided interviews – if designed properly – are usually clearer, more job-relevant and more relatable than their human counterparts. The advantage of AI is not in better insight, but in a lower level of the amount of “water” that recruiters usually write in reports.

Best way forward

Business owners or decision-makers in companies bear full responsibility for the use and implementation of AI in their processes. Shea is not a panacea for all ills and it cannot replace people.Such technology is very powerful, but its value depends entirely on how it is taught and applied. In the best cases, Shi will reduce the number of unnecessary actions, ensuring consistency and impartiality. At worst, it will accelerate depersonalization, increase prejudice, and automate equally bad processes.

In most cases, the most predictive hiring algorithm will consist of a well-thought-out job analysis, structured interview, scientifically based evaluation and meaningful performance data. Shi cannot replace the human tasks that matter most: understanding motivation, ethical judgment, cultural interpretation, and the ability to build trust between candidates. Even if it is urgent work in Kyiv or Lviv, humanity should be at the core.

End Goal

So what is the ultimate goal? The goal is not to replace the recruiter and not to rebuild the hiring process, which will be exclusively based on Artificial Intelligence, thereby reducing the cost of hiring.

The goal is to reinvest the time AI saves in research and deeper conversations and testing, better evaluations, and more responsible decisions. Data entry and analysis, error tracking, and data collection to improve expertise. It is in this way that AI will help us eliminate our subjective evaluations and biases during hiring, and help us make the hiring process the most human.