Samenvatting
AI can make HR faster and generate insights that were previously unavailable, but it does not automatically lead to better, fairer or more objective decisions. Its impact depends on the data used, the assumptions built into the system and the purpose it serves. As AI takes a more active role in recruitment, performance management and employee monitoring, HR must critically assess vendor claims, protect privacy and human judgement, and decide which parts of people management should remain genuinely human.
AI tools promise HR faster, more objective decisions and are taking on an increasingly active role in all kinds of processes. For Prof. Aizhan Tursunbayeva, that raises a fundamental question: who actually becomes more powerful through AI in HR, and for what purpose?
What exactly does AI-augmented HRM mean?
‘In the past, HR technology was mainly about automating tasks and standardising processes. Augmentation goes a step further. It means using technology to extend human capabilities and generate new insights. AI systems can, for example, analyse cover letters to infer personality traits, or analyse facial expressions and eye movements during a video interview.
Those analyses generate information about candidates and employees that HR simply did not have before. But who benefits from that knowledge? It can be used by HR professionals, managers or other parties making decisions about people. So we should not only ask what AI can do, but also whose capabilities it expands and for what purpose.’
How is that different from the HR technology organisations were already using?
‘With AI-augmented HRM, it is not only the type of data we collect that changes, but especially what technology can do with it. In the past, people entered data and received, for example, a report in return. Now a system can identify patterns itself, make predictions and communicate directly with employees.
A chatbot, for example, can not only retrieve information but also answer questions and guide people through a process. Other systems try to infer from different signals how employees are feeling or when they should be nudged to do something. Technology therefore takes on a more active role.
At the same time, we should be careful with the label AI. In the research I am currently conducting with some colleagues, we found that some products were marketed as AI while they were in fact not AI-based technologies. When we contacted vendors directly, some even told us that there was actually no AI behind their product.’
AI promises to make HR faster and better. What changes in practice when more and more decisions are made together with AI systems?
‘Faster is easier to achieve than better. Speed can be measured, but what is a “better” decision? Is it cheaper, fairer or more socially acceptable? Even becoming faster requires work first: data need to be cleaned, systems need to be connected and AI output needs to be checked.
AI does not remove human judgement either. Someone decides which data a system uses, which filters it contains and according to which logic it works. If there is already a bias in a process, for example, AI can amplify that bias.
At the same time, technology can take over tasks that used to belong to HR or managers. Look at Uber: performance management is largely based on the stars customers give, without a manager or HR professional being involved in the traditional way. And we are seeing that change not only in platform work, but also in traditional organisations.
What I find particularly interesting is what that does to the relational side of HR. I hear examples of people applying to very large, traditional organisations and going through almost the entire recruitment process without speaking to another person until they sign the contract. Recruitment was traditionally also a form of employer branding: the organisation is not only assessing you, it is also presenting itself to you and trying to build a relationship.
If a large part of that interaction is automated, that relationship changes. So the question is not only whether AI makes HR faster, but also what role HR itself still plays when decisions and interactions increasingly take place through technology.’
One major promise is that AI can make recruitment more objective. Why is that idea so attractive and can AI actually deliver on it?
‘Recruiters can have conscious and unconscious biases. Their judgement can be influenced by personal preferences or by characteristics of a candidate that have little to do with their competence. That makes the promise of AI attractive: if a system can ignore those characteristics, in principle it could focus more on competencies. AI vendors sometimes compare this to blind auditions in orchestras. By placing a curtain between the musician and the jury, the jury cannot see who is playing and can focus more on the performance. The suggestion is that AI can do something similar. However, that only works if the system itself has not been built with bias. Much AI is trained on historical data. If those data contain gender bias, for example, the system can simply reproduce it. AI therefore does not automatically make a process objective, and there is a risk that biases are applied on a much larger scale. A recruiter might speak to eight or ten candidates a day. An AI system can analyse thousands. If there is bias in the system, many more people can therefore be affected by it.’
You mentioned HR technology that can analyse people’s facial expressions and writing, or measure their performance in ever greater detail. Where is the line between supporting people and controlling them?
‘That line mainly depends on what the technology is being used for. Some professional football players, for example, use helmets with sensors that measure their performance in great detail. The purpose is very clear: to improve performance. They are also a small group of very highly paid professionals. You can ask whether that affects their willingness to accept this kind of measurement.
In an ordinary organisation, the context is very different. Technology can be used to continuously monitor large groups of employees, while those employees are financially dependent on their employer. That makes the question of whether someone is truly consenting much more difficult. You may formally give permission, but how voluntary is that consent if refusing could have consequences for your job?
And people also adapt to such systems. If an organisation monitors whether someone is moving their mouse, for example, employees look for ways to get around that measurement. There are videos circulating of people placing their computer mouse on a robot vacuum cleaner so that it keeps moving. That shows quite clearly that monitoring can also provoke behaviour that you probably never intended to measure.’
Does this increasing measurability of work create a risk that employees have less room to experiment and be creative?
‘Absolutely. In academia, we can spend weeks or months reading before we start writing. Knowledge work simply requires time to think and reflect, and you are not constantly faced with the same problem that can easily be automated.
In a working environment where speed and measurable performance become increasingly important, that time to think can quickly come under pressure. A certain amount of stress can potentially improve performance, but constant stress has negative effects.
The same applies to observation. Research in organisational behaviour has shown for a long time that people behave differently when they know they are being observed. If everything constantly has to be visible and measurable, you risk overlooking the value of reflection and creativity.’
You studied how candidates respond when AI and digital data are used in recruitment. What did you find?
‘We conducted an experiment with candidates in Italy and Spain to examine whether their perception of an employer changed when AI was used in recruitment. Interestingly, the use of AI by itself did not make that much difference. What mattered much more was the type of data being used. We distinguished between professional data, such as information on LinkedIn, and more personal data, such as photographs or information that might reveal someone’s political or religious beliefs.
So it was not so much the fact that AI was being used, but rather which data the organisation was analysing with it that affected both how candidates viewed the organisation and their willingness to continue with the application. Some candidates perceived a company using AI as more innovative and technologically advanced, but still decided not to proceed with their application. Being perceived as innovative therefore does not automatically create trust.
We also saw differences depending on people’s trust in technology. Candidates with greater trust in technology were generally more willing to continue, while people with less trust were more likely not to do so. Another interesting finding was that engineering students showed more reservations than business students. At first sight, that seems counterintuitive, because you might expect engineers to feel more comfortable with technology. One possible explanation is precisely that they understand better what these systems are capable of and may therefore also be more aware of how personal data can be used or misused.’
Many of these systems are bought from external vendors. How can HR evaluate those products properly?
‘HR professionals are entering an area in which they need at least some technological understanding, while that was not necessarily part of traditional HR education.
We studied how vendors market their AI products for HR and found three broad types of arguments. The first is pragmatic: the technology will make you faster, more efficient or improve your decisions. The second is moral: decisions will become more objective, fairer or more inclusive. The third presents technology almost as something self-evident or inevitable: this is simply the next calculator.
HR needs to learn to recognise those different claims and ask what actually matters to the organisation. If you focus only on efficiency, you may lose sight of the human dimension. But focusing on fairness generically without considering what it actually means in your context and how it should be operationalised in practice is not enough either. The challenge is to keep those different objectives in balance instead of simply adopting the vendor’s definition of what “better” means.’
What happens to the application process when both candidates and recruiters start using AI?
‘We already see candidates pasting a job description into generative AI and asking it to rewrite their cover letter so that it matches the job description. Some recruiters try to detect that by placing a hidden instruction in the vacancy – for example in white text that people cannot see but AI tools can – saying that the word “parrot handling” or “hyppopotamus” must appear in the letter. Someone who accepts the AI output without checking it may then unintentionally reveal that the output was not reviewed critically.
There are also tools that allow people to apply automatically to large numbers of vacancies, while employers use AI to screen those applications. You can end up with a situation in which AI writes an application, another AI assesses it and human beings are increasingly less directly involved.
That brings us to a fundamental question: which parts of recruitment do we actually still want to keep human? Technology is making more and more things possible, but we have by no means decided yet what we consider acceptable in all of those situations.’
Five tips for HR professionals
1) Decide what you want AI to improve.
Do not automatically assume that faster also means better. Decide beforehand whether your main priority is efficiency, fairness, employee experience, better decisions or a combination of these.
2) Know what data your tools use.
Make sure you understand which data are being collected, what conclusions are being inferred from them and whether employees or candidates know how their data are being used.
3) Preserve human judgement.
Do not blindly outsource decisions to AI. Check the output critically, especially when decisions affect someone’s career.
4) Question vendor claims.
Ask what is really behind promises of greater objectivity, fairness or efficiency. HR does not need to become a team of engineers, but it does need enough technological understanding to ask the right questions.
5) Look beyond the official AI systems.
Think not only about the tools your organisation purchases, but also about how HR professionals can use tools such as ChatGPT on their own initiative. Pay particular attention to privacy, sensitive employee data and transparency about the use of AI.
