Artificial intelligence in HR management

I. Introduction

Often described as the key technology of the future, artificial intelligence (AI) holds many promises.

Artificial intelligence is a term used in a wide variety of contexts today, but it has no generally accepted definition. It is clear that AI is not a single technology, but a whole spectrum of closely linked solutions and processes that access and process large amounts of data. Its purpose is to perform tasks such as learning, evaluating and solving problems, for which a human beings would have to use their intelligence. The distinguishing feature of AI is a certain degree of adaptability and autonomy of the technical system.

Its usage covers the entire cycle of working life, from recruiting and personnel management to long-term personnel development.

However, the potential that the use of AI brings to companies also gives rise to new challenges. The law is lagging far behind the rapid advancements in AI technology and many important questions have not yet been conclusively clarified.

The following outlines the key aspects involved in using artificial intelligence in HR management.

II. Fields of application and opportunities of AI in HR

1. Recruiting

Recruiting is one of AI's main fields of application

Prior to actually advertising a job vacancy, companies can use key figure analysis to identify the recruiting channels that attract the most applicants and the portals that have generated the most hires. In addition, algorithms can use "programmatic advertising" to influence who sees the advertisement. Another conceivable option is "robot recruiting", the proactive search for suitable candidates by means of algorithms that search business networks such as XING or LinkedIn based on requirements and criteria. This could prove to be a key advantage in the "war for talent".

In a next step, "people analytics" applications can be used to make a pre-selection from the job applications received.

AI can also save companies time and money on job interviews if they use chatbots to assist them or have chatbots conduct entire interviews. Another use of AI is automated evaluation of speech, facial expressions and gestures during a job interview to analyse personality traits, such as team spirit or ability to perform under pressure. The combination of AI and Big Data, i.e. the analysis of large amounts of data from many different and variably structured sources at high speed, is of particular interest in this regard.

Online assessments often form part of comprehensive application processes; this is another possible use case for AI. To give an example, Unilever used a third-party platform in a selection process where applicants had to solve a series of "computer games" in a short period of time. These games identify and assess cognitive and emotional characteristics, especially the applicants' propensity for or aversion to taking risks. The analysis was aimed at finding the candidates whose measured characteristics best correlated with those of successful employees in the company who had a similar propensity for taking risks.

2. Personnel management

In personnel management, AI has great potential when it comes to automating routine tasks. "Virtual" HR employees in the form of chatbots can process recurring queries automatically and let employees know where to find certain information. Nestlé, for example, commissioned the development of a chatbot for its HR employees based on the SAP Cloud Platform, which processes information and HR key figures and renders them in graphical form.

Payment of wages, holiday and absence management, preparation of duty rosters, assignment of tasks and evaluation of services rendered or setting of new targets for the following period are all processes that could also be automated. The time gained from this could be spent on more meaningful "human" interaction with employees, thereby achieving greater employee retention.

3. Personnel development

AI also has the potential to revolutionise personnel development. At the individual employee level, AI, again in combination with Big Data, facilitates "matching", i.e. the best possible allocation of an employee to a position in the company.

By analysing individual interests and skills, each employee can be given access to personalised training opportunities, thereby systematically preventing excess or insufficient workloads. AI bots can systematically bring employees up to the same level of knowledge so that they can just practise what they have learned and ask specific questions in a condensed attendance training session. This can reduce seminar times by around 50 %.

AI and Big Data make it possible for companies to forecast the development of relevant variables, such as future staffing needs, and thus avoid idle times and manage expected shortages. One good example of such a process is staff planning in a call centre. This requires a forecast of how many calls the call centre will approximately receive on a given day. By using "machine learning" methods based on existing data on past calls, an intelligent IT system can provide these forecasts. Opportunities and risks can be identified at an early stage via trends and projections and necessary action can be taken proactively.

Analyses of reasons for dismissal to specifically prevent dismissals may be of particular interest to companies. It is possible even earlier for intelligent systems to use Big Data to determine the characteristics of long-term employees who are particularly loyal to a company. This enables the development of strategies for sustainable staff retention and the reduction of staff turnover.

III. Legal aspects

Despite the clear emphasis placed on its potential, there are also undeniable risks involved in using AI, which the legal system is gradually addressing.

1. Discrimination

When using AI, the so-called black box phenomenon may cause problems from the point of view of the employees concerned: The exact functioning of the AI system is usually beyond the user's, and often even the developer's, knowledge. It is not easy to work out how it has made a certain decision. In the worst case, the opportunities AI provides may even be turned on their head: If the processing is based on data that is insufficient in terms of quantity and quality ("dirty data"), the AI system itself can make discriminatory decisions and, for example, discriminate against people on the basis of their gender.

This may risk exposing the employer to claims for damages under the German General Act on Equal Treatment (AGG) due to discrimination. The German General Act on Equal Treatment (AGG) provides affected applicants or employees with aids such as a reversal of the burden of proof (section 22), so that initially only circumstantial evidence suggesting discrimination must be presented. It is also generally presumed that the employer is responsible for the discrimination by the AI (section 15 (1)). There are thus legal risks that need to be addressed at the implementation stage of the AI system.

2. Data protection

As is so often the case in connection with the introduction and use of technical facilities, using AI can also give rise to data protection issues. If an AI system is used that processes applicant or employee data, this requires authorisation.

In practice, simply obtaining the consent of the employees affected by the data processing is rarely a legally secure option for achieving data protection compliance: Such consent may be revoked informally at any time. Moreover, doubts are often raised about the voluntary nature of consent given due to the imbalance of power between employer and employee. If consent is not given voluntarily, it is invalid. In an ongoing employment relationship, the granting of consent is less of a problem, provided its refusal does not cause any disadvantages. It would be conceivable, for example, to use AI in the context of personnel development to provide employees with individual, non-binding further training offers.

Instead of individual consents, it may be advisable in practice to conclude a works agreement with the works council. The works agreement is suitable for authorising data processing under data protection law and complies with the works council's extensive participation rights in the introduction and use of technical facilities.

Authorisation by means of a comprehensive and legally secure justification of the necessity of data processing for the employment relationship is also conceivable. If it is necessary to process employees' personal data to establish, implement or terminate an employment relationship, it may be done taking into account the other requirements of the General Data Protection Regulation (GDPR). For data processing to be deemed necessary within the meaning of the GDPR, the interests of the employees concerned and of the employer must be carefully weighed against each other. It may turn out that the data can only be used for certain purposes or that there has to be restrictions on who is authorised to access them. Where possible, without compromising the AI's functionality, data should be encrypted and pseudonymised.

In recruiting, it is difficult to justify using AI to perform comprehensive personality analyses, as there will usually be other ways of making a candidate selection that require less data.

Ultimately, the other obligations arising from the GDPR must be complied with: Data subjects must be informed comprehensively about the processing of their data; data that is no longer needed must be deleted; erroneous data must be corrected and claims for information by data subjects must be met. If data collected is going to be used to further train an AI system, the employee concerned must be informed about this alternative use. Furthermore, if it is not necessary to use personal data to train the AI system, they should be anonymised before further use. They are then also no longer subject to the scope of the GDPR. Where AI applications are used that process sensitive data in an automated way, a data protection impact assessment may be required that sets out the scope, purpose and proportionality of the data processing in a comprehensive report.

3. Works constitution law

As already mentioned, the works council has extensive rights to participate in the introduction and use of technical facilities in the workplace. Before using an AI application, it must therefore be checked whether it triggers co-determination rights of the works council. Omissions during the preparatory stage can delay the subsequent use of the AI system, provoke actions for injunction by the works council and shatter the relationship between employer and works council.

The implementation of AI systems is in many cases subject to co-determination. This is true in any case if the AI system can at least theoretically be used to draw conclusions about the behaviour or performance of the individual employee. An AI system that, for example, determines a team's productivity, possibly by calculating an efficiency score, can therefore in most cases not be introduced without works council's involvement.

The rights of the works council in connection with the introduction of AI were extended in 2021. To enable works councils to exchange views on planned projects at eye level, access to experts was facilitated. The engagement of experts must be financed by the employer. It has not yet been conclusively clarified whether existing (and thus cheaper) experts in the company will be used as a matter of priority.

In addition to the right of access to experts, the works council is granted a right to be informed and consulted with regard to the introduction of AI. The works council can therefore demand its early involvement in the planning. Finally, the works council should be able to object to AI-generated selection guidelines that are used as a basis for an application procedure, for example, and insist on an adjustment. Such selection guidelines could be created in particular by the "people analytics" applications mentioned above. The reform thus affects the recruiting sector in particular and thus generally makes it necessary to involve the works council in the introduction of AI-supported recruiting software. As already mentioned, employers and works councils have to pay particular attention to non-discrimination in this context.

IV. Conclusion

It is important to note that the use of AI in human resources can present previously undiscovered opportunities. AI can help find the right person for a job or promotion. If implemented correctly, it also has particular advantages for the workforce. The legal risks here lie primarily in data protection law, but can also be found in the process of workplace co-determination.

However, with the help of expert advice, these potential problems can be safely circumvented. We will be happy to help you implement further digitalisation measures in your business and discover the opportunities that artificial intelligence holds for your business.

For more information on the influence of AI on the working world of the future, including contributions by authors from our Labour, Employment & Pensions department, see the book "Arbeitswelt und KI 2030", soon to be published in English as well.

Don't let your business fall behind on critical legal developments. Stay up to date with Law-Now, the CMS subscription service that keeps you informed on the latest legal developments. Sign up today and ensure that you never miss an important update again.