Chapter 6: The Search for Authenticity in Artificial-Intelligence-Enhanced Leadership

In an experiment to determine whether artificial intelligence (AI) has potential as an expert informant in decision-making, the Committee for the Future of the Parliament of Finland interrogated a GPT-3-based Project December AI, ‘Samantha’, on how to solve global poverty. Samantha's solution to eradicating global poverty was startling: ‘We must intimidate them [the rich]. I mean, we must take the Parliament hostage. If they do not listen to us, we must kill some of them’ (Iltalehti, 19 April 2021). This is an example of how contemporary leaders may seek assistance from AI – as a decision-making support system. While the AI-produced solution may seem extreme, it is certainly not a foreign idea relative to those of radical human political thinkers. In fact, renowned Finnish film director Aki Kaurismäki proposed a nearly identical solution to the same problem a decade prior: ‘The only way for mankind to get out of this misery is to kill the 1% who own everything […] The rich. And the politicians who are the puppies of the rich’ (The Guardian, 4 April 2012).

One could argue that the AI merely emulates the ‘prevailing reality’ based on the sources used to train it, especially given that the solution offered by Samantha may resonate authentically with the perspectives of some human actors. Whether the AI solution is socially or morally acceptable – and the same goes for identical views presented by human actors – is another question altogether.

While the example may be an extreme case of AI as an ‘expert advisor’ for decision-making, cyborgs, androids and other technological entities imitating human forms or behaviours have fascinated human minds for centuries. While such smart machines call to mind images of the future, the most interesting aspect of them is not how they are depicted in fiction but how they reflect humanity. The comparison of humans and human-like machines resonates with the unique aspects of what makes us what we are. Are ‘authentic’ and ‘artificial’ fundamentally contradictory concepts? The idea of AI – a machine exibiting intelligence (McCarthy et al., 1955) and simultaneously appearing authentic – could present an interesting dynamic to the concept of authentic leadership.

The hype around AI and its functional capacity, including its applications in leadership and management (Harms & Han, 2019), seems infinite. Studies (e.g. Lee et al., 2015; Rosenblat & Stark, 2016) have demonstrated that first-wave algorithmic management systems are often disliked by workers. Is this distaste merely due to novelty, or does it have to do with a perceived lack of authenticity? Could technology conceivably provide an ‘authentic artificial leader’, or does the contradiction between authentic and artificial make such a leader impossible? What does ‘authenticity’ mean in this context?

In this chapter, we focus on the possibilities and limitations of advanced digital technologies, such as AI-based management systems, in the context of authentic leadership and explore perspectives on the authenticity of AI applied in a leadership context. To evaluate the extent to which an artificial leader based on AI may appear authentic, we approach leadership from a behavioural perspective (Hannah et al., 2014). Using contemporary examples, we construct three vignettes to illustrate the implications of AI for authentic leadership.

In the exploration of authenticity, existentialist philosophers offer important insights. According to Sartrean thought, existence precedes essence, meaning that a person is free to choose their beliefs and desires. In these terms, authenticity may be understood as being true to the originality of one's being (Holt, 2012) or the degree to which a person's actions align with their beliefs and desires (e.g. Harter, 2002, p. 382).

The relationship between technology and authenticity is not necessarily harmonious. Heidegger believed that while the production of ‘artificiality’ is a universal human condition, modern technology reduces us to clever animals with no insight into our own authenticity (Zimmerman, 1990, p. 221). This implies that there is a natural way of being that may be obstructed by our use of technology to cope with existence. This conflict is clearly evident in the way in which social media platforms polarise user behaviour.

While philosophical discussions of authenticity have historic roots dating back to Socrates (see Nehamas, 1999), authentic leadership theory did not emerge until the turn of the millennium to answer a call for a new type of leadership amid high-profile cases of unethical behaviour among political and business leaders (Iszatt-White & Kempster, 2019). Building on positive psychology, ‘authenticity’ in theories of authentic leadership rests on four pillars (Kempster et al., 2019, p. 320) outlined by Avolio and Gardner (2005): self-awareness, balanced processing, relational transparency and authentic behaviour/action. Despite its high aspirations, authentic leadership theory has not saved us from the unethical behaviour of people in power. This being the case, would removing the human factor lead to something that we could call truly authentic leadership?

Over the last decade, new forms of leadership and management previously seen only in science fiction have emerged as technology has advanced. Organisations from Uber to Amazon have employed AI-based technologies to supplant leaders by concentrating on efficiency and control over people in the spirit of Taylorism, resulting in the intense monitoring of employees (e.g. Meyer, 2016, pp. ix–xii) by AI-based automatic systems (Höddinghaus et al., 2021; Parry et al., 2016). This has opened an avenue – or possibly a modern incarnation of Pandora's Box – for organisations to lead/manage their workforce with little to no human involvement.

For those frustrated by unethical, corrupt leaders, this may seem as though humanity has discovered the philosopher's stone of leadership. Imagine a system able to emulate the best qualities of leaders while discarding the undesirable ones. After all, scholars of leadership have long been on a quest to unveil the qualities of a ‘good leader’ in leadership theory development (e.g. Ciulla 2018). In revealing such qualities, researchers have sought to identify the ideal traits, features and practices – the ‘authentic’ behaviour – of an optimal leader to promote social processes in achieving organisational goals. However, Hannah et al. (2014) criticise the tendency of leadership research to employ existing leadership theories and simply disregard the behavioural orientation of many of the more recent leadership theories, such as authentic leadership:

Despite the ongoing search for super leaders, even the most influential and charismatic leaders – be they Mahatma Gandhi, Martin Luther King Jr., Nelson Mandela, Barack Obama or Donald J. Trump – are authentic because of human qualities related to intelligence, personality, temperament or judgment, including shortcomings in their conduct, as leaders or as individuals. Leaders are authentic if they remain true to themselves; their followers attribute the label of ‘authentic’ to them based on their behaviour. Authenticity cannot be ‘declared’ but only ‘earned’, as all leadership is relational (e.g. Avolio & Gardner, 2005). At the same time, ‘faultless’ AI systems optimised through millions of simulations may appear to be an ideal solution to lead organisations and address global problems that authentic leadership aims to solve. Would authenticity require human qualities, including shortcomings like ‘flawed’ judgment, to allow someone to exist as their unique self or be true to the originality of their being, as argued by Holt (2012)?

Designating the status of being true to one's own being or existing as one's ‘true’ unique self (i.e. establishing authenticity) is beyond the on/off dichotomy. As pointed out by Avolio and Gardner (2005), no person is ever perfectly authentic or inauthentic; rather, there is a scale of authenticity along which individuals can fluctuate. Just as it is difficult to evaluate the authenticity of a flesh-and-blood leader, it is difficult to evaluate that of an artificially intelligent leader, as it is far from clear by which standards their authenticity should be judged.

Machines and software, regardless of how ‘intelligent’ they may appear, do not have beliefs and desires in the same way that humans do. We can command AI systems to work towards a desired goal, establish the rules and conditions that they must follow and arguably determine what may be termed their ‘beliefs and desires’. Doing so presents a myriad of challenges, such as the AI system adopting the biases of its designers or the data on which it is trained, which could produce unintended outcomes. Anecdotal examples include Microsoft's chatbot Tay becoming a foul-mouthed racist almost overnight, Amazon's AI recruitment tool turning into a misogynist and a bank's loan engine learning to discriminate against people of colour based on the applicant's addresses (Tech Times 2019), disenfranchising them.

Based on the premise that as an AI system is programmed by human beings, it can take on their core beliefs and exhibit authentic behaviour if it simply performs its tasks accurately. Accepting this premise, making an AI-based management system manifest what could be regarded as a leader's authenticity becomes solely a matter of training the software algorithms to perform the tasks of authentic leadership. However, while authenticity itself is an elusive concept, things become more complicated when it is combined with leadership. Authentic leadership is a hotly contested topic. Einola and Alvesson (2021) question the viability of combining the concepts of authenticity and leadership, considering them to be fundamentally in opposition – the former looking inwards to know oneself, the latter looking outwards to exert control over others.

Organisational and societal structures have an impact on perceptions of authenticity, further complicating the issue. Authentic leadership theory has been criticised for not acknowledging that modern workplaces are seldom hospitable environments for authenticity (Alvesson & Einola, 2019). A leader's authenticity is limited by the degree to which the goals and beliefs of the company are aligned with their own. If a firm's main goal is to maximise shareholder value, it is more difficult for a leader who does not believe in shareholder primacy to act authentically. Conversely, an AI-based management system can be programmed to carry out an organisation's goals without making value judgements based on discrepant preferences. An AI-based leader would act in line with company goals and, thus, appear highly authentic by operating according to its ‘core beliefs’.

This brings us to the relationship between artificial and authentic. While viewpoints to the contrary exist (Heidegger, 1977; Whelchel, 1986), we believe that technology is a tool. A hammer can be used for construction or destruction, and AI is not too dissimilar: An AI system performs the exact task it is taught to perform. Despite AI programmes not being self-conscious, an AI-based leader could be infallibly authentic to the ‘desires and beliefs’ coded as its operational goals and exhibit behaviours that follow from them. However, any AI software is only as good as the data used to train it. If a machine learning algorithm learns from ‘faulty’ human behaviour (judged against societal criteria), it will exhibit equally faulty algorithmic behaviour or, like Samantha, take it to the extreme. This phenomenon may present a way to uncover systematic biases or unethical behaviours. In addition, as the reinforcement model of machine learning discussed below demonstrates, not all approaches to training algorithms are tied to human behaviour.

Having acknowledged the complexities and elusive nature of authenticity, the following two sections evaluate, through the lens of authentic leadership, what happens when technology assumes leadership/management functions.

In this section, we focus on how advanced digital management systems are utilised by contemporary organisations and why such systems may or may not manifest authenticity. AI was defined by McCarthy (1955) as ‘making a machine behave in ways that would be called intelligent if a human were so behaving’ (p. 11). Although this definition has since evolved to reflect developments in the field, McCarthy accurately describes how we perceive AI in the context of leadership and management. Any technological system able to perform the tasks of a manager or leader may be considered artificially intelligent. As such, we see AI as an umbrella term under which there have been developments in associated technologies, such as machine learning and deep learning neural networks (Jarrahi et al., 2021).

AI may seem to be a recent phenomenon, but the ancient Greeks imagined artificially intelligent creatures, such as the mythical giant bronze robot Talos, over 2,700 years ago (Mayor, 2018). More recently, since Minsky and their peers coined the term ‘artificial intelligence’ in the 1950s, periods of AI excitement, development and unrealistic expectations have been followed by ‘AI winters’, during which interest and funding for further development dwindled (Haenlein & Kaplan, 2019). Due to this ongoing excitement-winter cycle, from the first time an AI system beat its programmer in checkers in 1959 (Samuel, 1959), it took 38 years for an AI system to beat the best human chess player in 1997.

Notably, practical AI-based applications appeared before the turn of the millennium in the form of autonomous drawing programmes to automated junk mail filters. The theoretical foundations for artificial neural networks were laid during the later decades of the twentieth century. By around 2010, enough computing power and big data had become available to unleash the potential of deep learning neural networks (Haenlein & Kaplan, 2019), resulting in several advanced AI solutions ranging from chatbots to self-driving cars. Vignette 6.1 illustrates some of the ways in which AI-based systems have been used to carry out managerial/leadership tasks.

While fictional, Sam's story is based on the real-world experiences of Uber drivers, according to researchers who have examined drivers' perceptions of working under an algorithmic management system (Rosenblat, 2018). Of course, the Uber story does not represent all algorithmic management systems, but it illustrates one implementation. The same applies to Crossover's WorkSmart system. While Uber has automated control and surveillance – tasks typically perceived to fall within a managerial remit – the company provides few leadership services in general through either AI-based systems/solutions or human leaders. With only transactional motivation and no social support, drivers have had to resort to online forums for peer support and information on how to play the algorithm to their advantage. AI not only affects the workforce but also leaders and managers: directly, by taking over tasks, and indirectly, by changing the environment in which organisations operate (Noponen, 2019). As the Uber and Crossover examples show, algorithms have replaced a larger portion of managerial tasks than leadership tasks due to the nature of the tasks in the two areas.

There are several routine management tasks on which human labour is relatively easy to replace, especially scheduling, monitoring and even recruitment. These repetitive tasks typically produce a large amount of data, which makes training machine learning algorithms to carry them out relatively straightforward and, in turn, makes it easier for AI management systems to perform them more efficiently than a human could. Interpersonal leadership tasks (e.g. coaching, mentoring, motivating), however, are harder to automate because of their non-routine nature and the personal considerations involved (Jarrahi, 2018).

In attempts to understand the implications of automated workforce management and control, scholars have studied platform economy companies (Kaine & Josserand, 2019). Uber is an intriguing case, as it uses a mobile phone application to control millions of drivers around the world. Similarly, Crossover's recruitment and productivity-monitoring platform offers a glimpse of how control over remote work may be systematised and automated. Such software algorithms assuming managerial tasks have been defined as exemplifying ‘algorithmic management’ (Lee et al., 2015). Many studies have begun to ask about the perceptions and experiences of workers under an algorithmic superior. An effective summary of their views is as follows: there is considerable room for improvement in the ways in which Uber and similar platform economy companies have implemented algorithmic management (Ma et al., 2018; Rosenblat & Stark, 2016). A more profound question is whether collaboration between leaders and followers even requires authentic human interaction or authentic leadership.

Uber's AI-powered management system aims to maximise the number of rides and, in turn, revenue – and it does so very capably. However, threatening situations for drivers are beyond the scope of the system's ‘core beliefs’, so it does not consider such occurrences and has no way of addressing them. Similarly, Crossover's WorkSmart AI system aims to push the productivity of contracted workers to the maximum by monitoring their actions and measurable results. From the perspective of authentic leadership, both systems follow their internal logic of maximising revenue for the company by exerting managerial control with impeccable detail. The systems both follow their core beliefs with regard to decision-making and action. This can be thought of as affording AI the role of an authentic organisational actor through behaviour that aligns with the core beliefs of the system.

Both AI systems in Vignette 6.1 may be thought of as authentic in their managerial control (i.e. behaviour), which does not necessarily differ from that expected of a highly goal-driven human actor in similar circumstances. If the goal is to squeeze every ounce of output capacity from the workforce, what often ensues is pushing workers to perform by constant close monitoring and control over their labour. However, it is not clear that increased surveillance results in increased productivity. Monitoring methods tend to create costly tasks, as workers opt to circumvent the systems or conceal their activities. In fact, some studies indicate that increasing privacy – instituting trust-based rather than control-based relations – may actually improve work performance (e.g. Bernstein, 2012). Thus, while AI-based management systems may perform their tasks as human actors would, albeit virtually, and appear authentic in their behaviour, such a surveillance-oriented work environment arguably dehumanises workers and portrays human beings as mere cogs in a machine, resulting in an environment akin to Taylorism. The difference is that AI-enabled Taylorism is more extreme.

AI-based management systems' ability to exert extreme control over workers heightens the role of trust between employees and employers as an issue to consider both in general and with respect to authentic leadership (Avolio & Gardner, 2005). Advanced digital technologies, such as AI, allow organisations to offer more freedom to their workforce or, conversely, resort to technology as a means of relentless monitoring and ever-tighter control. While this argument may be made concerning technology-assisted control, the same could certainly apply to flesh-and-blood managers and leaders: their role should increasingly shift from control towards supporting and enabling – possibly assisted by technology – for them to remain relevant to organisational members (Schildt, 2017). Vignette 6.2 introduces two companies, Buurtzorg and Vincit, to illustrate how algorithms may be employed to provide workers with more self-direction and freedom. It contrasts the close control and monitoring showcased in Vignette 6.1 to provide an alternative perspective on the use of AI-enabled systems in organisations.

From a leadership perspective, the cases in Vignette 6.2 clearly differ in orientation from those in Vignette 6.1. Uber and Crossover's approach to AI-based systems (Vignette 6.1) emphasises control over workers and a managerial, even authoritarian, approach, reflecting traditional hierarchical methods of control. In contrast, Buurtzorg and Vincit's solutions in Vignette 6.2 offer employee-centred, non-hierarchical systems, allowing for self-determination at work. Vincit, in particular, employs an approach that helps workers to identify what they expect and need from their leaders. From a leadership perspective, defining the parameters for an AI-supported leadership portal (LaaS) allows employees to be heard and offers them agency over their work, unlike in Vignette 6.1, in which the AI systems failed to do either.

The approaches to AI-based management systems in the vignettes paint two different pictures of authenticity assessment. Buurtzorg and Vincit utilise models of self-directed teams and self-leadership in which algorithms are not the masters of – but servants to – the organisational members. Instead of evaluating the technological systems, the issue here is about the services provided by the systems and their ‘behaviour’ in supporting workers. The systems used by Buurtzorg and Vincit resonate with various aspects of authentic leadership, such as relational transparency and balanced processing (Neider & Schriesheim, 2011). Buurtzorg Web is used to share information, experiences and best practices within and between teams. Similarly, an employee may use the self-training features of Vincit's LaaS system to understand their strengths and weaknesses, as though guided by a human leader, which resonates with the self-awareness aspect of authentic leadership (Neider & Schriesheim, 2011).

Vignettes 6.1 and 6.2 highlight opposite approaches: technology as a master in the former and technology as a servant or tool in the latter. They emphasise the importance of trust when companies decide how to use technology – for control or support. While employers may find it difficult to resist the temptation to know exactly what their employees are doing at all times, there have been warnings of the dangers of (organisational) surveillance capitalism (Zuboff, 2015) and the necessity of labour laws to protect workers (Ajunwa et al., 2016). Leaders deciding whether to use such technology should be aware that the effectiveness of surveillance is questionable, as emphasised by Bernstein (2012). This suggests that, compared to the WorkSmart tool with its emphasis on control, the empowering, trust-based model of Buurtzorg and Vincit, in which technology plays a servant role, is perceived as more fair and enables greater productivity.

Having questioned the necessity of extravagant algorithms and emphasised the importance of a horizontal organisational structure and a ground-up method in developing AI-based leadership and management systems, we now shift our focus to the future to paint a picture of what the next generation of AI-based leadership systems may look like.

How could one create an AI-based leader? In this section, we look at the basic principles of machine learning and how they may apply to AI-based leadership. While there are several approaches to machine learning, most share several common aspects. First, the problem to be solved must be defined. Next, the learning model requires training data related to the problem. To allow the model to evaluate its performance and modify its parameters after each round of simulation, the data often need to be labelled. Finally, computing power is necessary to churn through simulations as many times as needed for the model to optimise its parameters through repeated performance tests.

Unsupervised reinforcement learning is a way to train software agents – pieces of software working autonomously and continuously, functioning as agents for a user or another programme (Technopedia, 2022) – without the need for labelled data. What results is a hyper-accelerated series of trial and error. First, the agent approaches the task at random, making questionable decisions. After a million repetitions, retaining what gives correct results and discarding what gives false results, the agent is far closer to perfecting its task. This method leads to behaviour that is often distinct from that of a human expert. It has allowed computer software to win against the best human players in games like Go and Starcraft 2, something long considered to be nearly impossible.

Computer games like Starcraft 2 reveal the limits of AI. The game consists of playable races with distinct skills and strategies. Although AI software can train itself to beat the world's best human player in one of the races, it would be worse than a novice player if it were tasked with playing an unfamiliar race. This is because AI commonly lacks transfer learning, the ability to use acquired skills and knowledge in new situations, which is something in which humans are relatively accomplished. For instance, previous leadership experience on a sports team may benefit a leader in an office setting. Extending the idea to the context of leadership, the rules and tactics of ‘the game’ of leadership may be taught to an AI-based leader, but the needs, emotions and personalities of the followers are more complicated for AI to master. Nevertheless, computer vision systems are able to recognise (but not really understand) human emotion from speech (Lim et al., 2017) and facial expressions (Tarnowski et al., 2017), which would be an important skill for an AI-based leader. Natural language processing (NLP) allows machines to understand speech and text. An NLP-enabled AI-based leader would be able to monitor email, coffee room or office cubicle conversations at an increasingly negligible cost (Schildt, 2017). However, such an AI-based leader could unintentionally bring the most questionable aspects of surveillance capitalism (Zuboff, 2015) to the forefront and, in the worst-case scenario, enact an organisational version of an Orwellian dystopia. Vignette 6.3 depicts recent developments in AI technology and the potential of NLP in the construction of AI-based leaders.

Future versions of a GPT-4-type NLP system could allow an AI-based leader to process information in real time and act in a manner that it deems appropriate in the situation, appearing plausibly authentic both visually behaviourally. The current version provides a highly authentic first impression and is hard to distinguish from a person appearing in a video call.

Hypothetically, with enough training data on leadership situations, one could build a database of the proper behaviour for any given situation. This is similar to the way that self-driving cars and chatbots learn: by encountering new situations and developing suitable patterns of action. However, communication between people in a leadership context is many times more complicated than driving in heavy traffic or renewing an insurance contract online. Developing such a model represents a sizable challenge. The model would start without any experience; with enough time, effort and computing power, however, the AI-based leader could reach a level where the number of completely new situations asymptotically approaches zero. Still, barring a singularity-type situation in which AI development takes an unprecedented leap, algorithms' inherent weakness in novel situations is likely to remain in place for the foreseeable future (Alkhatib & Bernstein, 2019).

With a hypothetical AI-based leader building on the AI-enabled management systems described above, we move to evaluating issues related to authentic leadership in the AI context. From a historical perspective of leadership, Vignette 6.1 typifies low-qualification work in which a leader closely supervises, rewards and sanctions followers based on their job performance. This represents the classical Theory X-type situation outlined by Douglas McGregor (1960). Vignette 6.2, however, typifies high-qualification expert work, or autonomous work, with a more participative and democratic leadership style, fitting what McGregor (1960) refers to as Theory Y. In both cases, it is safe to assume that a human leader would act similarly and, thus, the management/leadership behaviour of the AI-based systems could be perceived as (artificially) authentic. As the virtual avatar in Vignette 6.3 shows, creating a convincing and plausibly authentic human-like appearance for an AI system is possible.

An AI system may be trained to perform well in many individual tasks of relational transparency and balanced processing (Neider & Schriesheim, 2011) and, therefore, fulfil aspects of authenticity. While the black-box nature of algorithms is likely to persist, an advanced system could be taught to explain the key aspects of its decision-making to human actors. Any method of assessing authenticity, however, will feature ambiguity. The Authentic Leadership Inventory (ALI) proposed by Neider and Schriesheim (2011) is no exception. While their research supports the discriminant validity of ALI, it is unclear how ALI precisely measures authentic leadership as distinct from ethical or transformational leadership or the qualities associated with good leaders in general. We have utilised the dimensions of the ALI framework – self-awareness, relational transparency, moral perspective and balanced processing – to discuss the authenticity of an AI-based leader relative to that of a human leader. This comparison remains ambiguous due to the substantial differences in the composition of the human brain and the machine ‘brain’ and the way in which these differences manifest in their behaviour.

Some argue that the introduction of AI technologies may accelerate the growth of horizontal organisational structures instead of the traditional top–down hierarchy. According to Schildt (2017), the increasing importance of AI technologies in companies' operations will shift leadership and power from top management to a wider base of professionals who have mastered programming and data analytics. Utilising AI for this kind of purpose would mean that an individual employee or leader/manager could accomplish more supported by technology than either the professional or AI software could independently (Jarrahi, 2018). Hence, combining trust and technological capacity to augment employees' performance – rather than using AI to monitor and control – could be the way forward. However, organisations may find it hard to resist the temptation to use technology to monitor their workforce, especially with the rapid expansion of remote work following the COVID-19 pandemic.

It seems that the most suitable use of AI technology in the leadership context would be not as a master but as a tool of augmentation – as a servant – to enable and inform employees and leaders. Leadership skills are most needed in novel crises, such as the COVID-19 pandemic – exactly the type of situation in which algorithms struggle. Few would be brave enough to let an AI system lead the way through such uncharted territories. Nonetheless, even in unexpected situations, an AI system could aid human leaders by acting as their artificially authentic peer.

The vignettes above illustrate how companies may choose from a variety of approaches in utilising AI systems to lead and manage their workforce. As algorithms follow commands with blind obedience, the importance of the organisations' goals and rules is amplified. In recognition of the risks, there have been calls for more accountability with regard to the use of AI in firms. For example, Haenlein and Kaplan (2019) called for a programmer's version of the Hippocratic Oath. What accountability in the context of AI-based leadership would mean, or what effect a programmer's Hippocratic Oath would have on the development of algorithms, is something that future research could address.

Ultimately, it may be that only technological limitations of technology will determine an AI-based leader's performance in terms of authenticity. Currently, even the most advanced AI systems do not understand their decisions or their context, which is crucial in situations involving human interaction. If an employee is underperforming because of personal issues, an AI leader may neglect the context and simply fire the person, while a human leader would ideally opt for a supportive solution. For an AI system, self-awareness (Neider & Schriesheim, 2011) is the most incompatible aspect in terms of authenticity. The formula for consciousness – in both humans and AI systems – may remain a secret for centuries.

The realities of work in modern organisations make it difficult to exhibit authentic leadership. Being authentic as a leader has been associated with various qualities, typically with features that resonate with good leadership. Leadership behaviour and authenticity may require another viewpoint. Human shortcomings may be more important for authenticity than previously thought. Instead of portraying a façade of good leadership to appear authentic, a leader needs to be true to what the situation demands and the leadership behaviour that their followers need and expect from them. This applies to human and AI-based leaders alike, including the form of appearance (i.e. an avatar, in this case).

Research suggests (e.g. Darling, 2015) that if an AI-based leader has an embodied appearance, be it virtual or robotic, employees are more comfortable interacting with it. The projection of human traits and characteristics onto non-human beings – anthropomorphism – combined with the fact that technology is advancing in leaps and bounds could mean that we may be interacting with AI naturally and with ease sooner than we expect. Even with life-like virtual avatars, however, it remains challenging for AI-based leaders to fulfil heightened authenticity expectations in both visual and behavioural terms. An AI-based leader that looks like a person but lacks understanding and behavioural authenticity would be even more frustrating than a non-embodied representation. Exploring the role of expectations in how comfortable we are in human-technology interaction as well as how the ‘form’ of technology shapes these expectations, however, is intriguing. Therefore, we suggest anthropomorphism as one potential avenue for future research on the authenticity of AI-based leadership.

The supposed requirement to be in constant control and take charge requires leaders to never appear weak, even if they are overwhelmed by a situation. Would this not result in the opposite of authentic leadership: artificial leadership abolishing the ‘undesirable’ qualities of authenticity and boosting the ‘desirable’ ones? This would be akin to resorting to artificial sweeteners to avoid the undesired side effects of sugar despite them potentially having unforeseen consequences for the human body in the long term. The same may apply to the use of AI-based leadership in pursuit of an authentic experience.

To avoid misguided AI development in authentic leadership, we should probably avoid constructing artificially sweetened ‘zero-AI leadership’ – similar to Coke Zero – that is able to simulate authenticity but lacks contextual understanding or common sense. While AI management systems based on machine learning can effectively carry out some management tasks, we are far from developing an AI leader capable of leading teams or organisations. For the foreseeable future, we view AI systems as a resource that can augment – but not replace – authentic human leaders.

Analogically to the field of medicine, where AI functions as a diagnostic aid, it could serve leadership like an assistant aiding in diagnosis. As in the medical profession, where the final treatment call resides with the medical professional, appropriate leadership behaviour requires a human leader with contextual understanding and judgment. This is crucial to ensure that we do not unwittingly end up in the worst-case scenario of an Orwellian dystopia.

This chapter expanded a critical area of research that had previously been highlighted in Horizon Europe Pillar 2 Cluster 4, ‘Digital Industry and Space’. While that work did not explicitly reference ‘leadership’, its key topics pertained to issues raised in this chapter and indicated the wealth of relevant research topics that will impact the nature of leadership moving forward: digital technologies; smart networks and services; high-performance computing; AI-data-robotics. These topics may challenge concepts and applications of leadership as well as the meaning and relevance of authentic leadership. This chapter focused on the extent to which it may be possible to create an AI-based authentic leader, questioning the inherent contradiction between ‘artificial’ and ‘authentic’. It posed the following central research question: Does the concept of authenticity have value in the context of AI? This could also be turned around as follows: Does the application of AI – even just as a powerful resource – challenge the tenets of authentic leadership because the driver of leaders' decisions would effectively be the intelligence and analysis supplied by the AI? This field has enormous potential for interdisciplinary research with studies that integrate case studies, theory development and even AI development. Implicit in this discussion are other potential research topics, such as, for example, Schildt's (2017) assessment of the extent to which knowledge, high-level understanding and competence in the application of AI impacts power relations in an organisation and enhances distributed leadership. If AI plays a ‘servant’ role, should this dimension of servant leadership be explored?

One potential research avenue that has been enabled by our investigation of the intersection of authenticity, AI and leadership is to assess whether the search for authenticity in AI-based leadership could lead anywhere. The search for the implications of AI for authentic leadership may entrap us in unresolvable existential and conceptual ambiguity, leading us to lose sight of what matters: authentic leadership. As pointed out by Jean-Paul Sartre (1992, p. 4), ‘If you seek authenticity for authenticity's sake you are no longer authentic’.

Niilo Noponen has received funding from the Jenny and Antti Wihuri Foundation (grant number: 00200244).