Introduction

AI robot bosses are becoming increasingly prevalent in organizations, and they expand the traditional organizational design space.

Organizations can benefit from utilizing both robots and humans as bosses, as they can substitute for each other and work together as complements across different organizational structures. This expanded design space includes different kinds of AI robots and humans as bosses who make decisions and coordinate activities, rather than limiting robots to just being helpers. By considering the different capabilities and relationships of humans and AI robots, we argue that the organizational design space is expanded to achieve greater effectiveness and efficiency.

Previously, Wesche and Sonderegger (2019) defined what they call computer-based leadership as “a process whereby purposeful influence is exerted by a computer agent over human agents to guide, structure, and facilitate activities and relationships in a group or organization.”

Thus, can an AI robot be a boss or leader? And if so, can the AI robot be a good one?

We explore these questions and argue that an AI robot can indeed be a boss with varying capabilities to fit the organizational situation.

Designing an organization with AI-based leaders or bosses provides new opportunities to make the functioning of different organizational structures more efficient and effective. Choosing a boss now includes both individuals and robots, each with their own skills, features, capabilities, and relationships. The organizational design will be affected as the AI robot boss changes both the information processing for decision-making and coordination.

Organizational charts, or hierarchy charts, visually depict an organization's structure, showcasing reporting relationships and chains of command. Employees’ names, titles, and job positions are typically represented in boxes or circles connected by lines, indicating their affiliations. However, traditional organization charts lack icons or representations of AI robot bosses. This discrepancy prompts the question not of their inclusion, but of why they are omitted. AI robot algorithms contribute to organizational information processing, with the authority to make leadership or boss decisions. Their inclusion would enrich organizational charts, providing a more comprehensive representation.

The concept of a robot boss is not new, as a decade ago, Cappelli (2014) and Wharton colleagues explored the question: Can a robot be your boss? The answer is yes, but with limitations. A human boss or leader may be better at judgmental tasks and making quick adjustments. However, Kruse (2018) predicted that in 2023, your boss will be a robot. Robot bosses are starting to come in different formats, e.g., HR management and accounting control. In Uber, many managers are AI robots doing routine managerial jobs (Wesche and Sonderegger 2019).

In a survey of 1770 managers from 14 countries and interviews with 37 executives in charge of digital transformation, Kolbjørnsrud et al. (2016) found that bosses should leave administration to AI and focus on judgment work. They also found that 78% of the surveyed managers believed that they would trust the advice of intelligent systems in making business decisions in the future. However, they also found that top managers have a greater trust in AI robots than lower and middle managers. One reason could be that AI is a higher threat to lower and middle managers than a threat to top managers.

Intelligent robot and human bosses require skills or capabilities depending on the organizational situation. Just as different situations call for different human skills, the fit of intelligent robots and what they do will depend upon the situation. At the same time, technological advancements are likely to increase the number of situations where robot bosses are effective. Earlier, Chamorro-Premuzic and Ahmetoglu (2016) argued that a robot boss can make better decisions and give objective feedback. However, there are limitations to the type of tasks and decisions a robot boss can do.

In this paper, we develop four types of AI robot bosses based on how easy it is to understand and explain the decisions made and how the robots learn and are trained. Then we map these AI robot bosses to organizational design structures and leadership styles.

Organizations are information-processing systems for decision-making and coordination (Burton and Obel 2018). The basic design problem is to create an organizational design that matches the organization’s demand for information processing with its information-processing capacity (Arrow 1974; Galbraith 1973). Intelligent robot bosses expand the organizational design space by their capability to collect and process large amounts of data. With AI robot bosses, horizontal and vertical relationships in traditional organizational forms are now changed. Further, AI robot bosses may be invisible bosses in new or boss-less organizations for data collection and decision-making.

Our contribution is to demonstrate that an intelligent robot can be a boss, not just a helper, and how an intelligent robot must fit the organizational situation, but also may significantly change the design of the organization. Based on four prototypes of AI robot bosses, we show how these types can fit as leaders in an organization. We also show that AI robot bosses due to acceptance considerations, e.g., legal restrictions, are not likely to serve as CEO or top management but more likely serve as bosses in a middle or lower management role.

We proceed to discuss what a boss or leader is and fits with appropriate structures. Then we develop four types of AI robot bosses based upon explainability and learning which are then matched with the leadership styles and structures; and thus, expanding the organizational design space with these matching AI capabilities for data collection and decision-making.

What is a boss and what do they do?

The term "boss" refers to a range of positions, such as supervisor, manager, leader, CEO, president, and commander. The boss holds the asymmetric authority to control and demand certain actions (March and Simon 1958; Weber 2019) of individual employees. A boss makes decisions involving consulting and delegating, planning and organizing, and problem-solving. The boss sends and receives information involving monitoring, clarifying and informing. Further, a boss builds relationships including supporting, networking and managing conflicts and team building as well as influencing people by motivation, recognizing and rewarding employees.

In making these decisions, a boss processes information. Burton and Obel (2004) describe bosses as decision-makers who process information in different ways. In their model, managerial information processing preferences are captured by two dimensions: preference for micro-involvement or delegation of authority for decision-making, and uncertainty avoidance or risk-taking. Based on these dimensions, bosses can be categorized into four executive styles: Maestro, Manager, Producer, and Leader, which are suitable depending on the information processing requirements of the organization and the tasks at hand (Burton et al. 2020).

Leadership styles and organizational structures

The type of leader for an organization very much depends on the situation and the organizational setup. Therefore, in considering the role of robots in organization design, we take our point of departure in Burton et al.'s (2020) leadership typology and discuss how adding robot bosses affects their effectiveness across the design space.

The appropriateness of a particular executive style depends on the situational demands as captured by their information processing requirements for decisions and coordination (Burton and Obel 2018). There is no universally good or bad type of executive style.

The four different executive styles, Maestro, Manager, Producer and Leader and their information processing styles, will, according to the multi-contingency framework, be appropriate, depending on the information processing demands posed by the internal and external environment of the organization and the tasks that must be done.

The Maestro has a low preference for delegation and low uncertainty avoidance. The maestro will intervene directly to ensure that decisions are made congruent with own desires. At the same time, the maestro does not avoid the uncertainty of long-term decisions and their implications for the firm. The maestro leadership style is often chosen in a startup situation and a simple organizational structure. However, many judgment decisions are required often with significant consequences in a simple structure (Burton et al. 2020).

The Manager has high uncertainty avoidance and a low preference for delegation. Avoiding uncertainty is realized again by making reactive and short-term decisions with a fine level of detail. The manager focuses more on the control of operations than on strategic, longer-term decisions. The manager does not delegate decision-making authority, but instead uses formalized rules to manage subordinates. Bosses who are high on uncertainty avoidance will normally match an environment and tasks where there is a need for focus, routine, and control, as they, information processing-wise, would like to know details related primarily to short-term decisions. When the situation requires low preference for delegation and high uncertainty avoidance, then the boss needs to process lots of routine data for decision-making and control. In such situations the appropriate organization design is a functional organization structure (Burton et al. 2020).

The Leader has a high preference for delegation and low uncertainty avoidance. The leader is confident that others can make good decisions for the firm and thus finds the delegation an efficient way to save time. Moreover, the leader does not avoid long-term uncertainty, but instead embraces its challenges by attending to more strategic decisions. As for preference for delegation, bosses who rate high on this dimension will match environments and tasks where the information processing requirements are either such that the amounts of information that need to be dealt with are so large that they require delegation, or where information processing is so disorganized that it benefits from discussion and joint interpretation. With a low degree of uncertainty avoidance, the boss will make many judgement decisions and with a preference for delegation, motivation will be as important as control. The appropriate organizational design will be a divisional structure (Burton et al. 2020).

The Producer has a high preference for delegation and scores high on uncertainty avoidance. The producer focuses on both efficiency and effectiveness. If your firm’s top management adopts a producer style of leadership, then the organization is likely to be well positioned vis-à-vis its competitors. The producer ensures that new products and services are developed and introduced. The focus of attention is a dual one: short term and long term; operations and strategy; current products/services and innovation; internal activities and environment reading; hands-on management and delegation so others can act independently; and efficiency and effectiveness. In a situation that requires a high degree of delegation and high uncertainty avoidance the producer leadership style often fits in a matrix structure. Here with the preference for delegation and high uncertainty a significant information processing is required with relevant controlling. There will also be many judgement decisions.

The four structures and four leadership styles are shown in Fig. 1.

Fig. 1
figure 1

Organizational structures and leadership styles (Burton et al. 2020)

Four types of AI robot bosses

AI, or Artificial Intelligence, is a field of computer science that focuses on creating intelligent machines that can perform tasks that typically require human intelligence. Thus, an AI robot boss can mimic or simulate human-like intelligence and behavior, enabling them to perceive their environment, reason, learn from experience, and make decisions to achieve specific goals (Zhu 2018).

AI encompasses a wide range of techniques, algorithms, and methodologies including natural language processing, speech recognition, image and video analysis, pattern recognition, robotics, planning, and problem-solving.

The AI robot bosses may be computer programs like chatbots, but they may also have a physical representation. Robot bosses can collect data from other AI devices combined with transaction data and data from the net, e.g., LinkedIn and Facebook, and search profiles both personal and company related. It may also be structured company files like the HR database or data from customer relation systems.

Artificial intelligence comes in many forms as do intelligent robot bosses. The major difference is the ability to understand the result—the explainability (Minh et al. 2022) and whether or not they are supported by supervised learning algorithms (Zhu 2005; Zhou 2018).

Many AI robots use a multitude number of algorithms as well as getting data from many sources and in many forms (text, images, natural languages, etc.). This means that it may be difficult to understand the reasoning behind the output from an AI robot boss. AI robots have therefore been classified into explainable or non-explainable robots. Given a certain audience, explainability refers to the details and reasons a model gives to make its functioning clear or easy to understand, (Arrieta et al. 2020).

The explainability is very much related to the type of AI algorithm used. For example, an AI algorithm based on decision trees and rules has a much higher degree of explainability than if it was based on a neural network algorithm. The first AI systems developed, also called expert systems, were based on rules and decision trees, and had a relatively high degree of explainability. One expert system example within organizational design is the OrgCon system (Baligh et al. 1996; Burton et al. 1998) which created a report to explain the design recommendation. On the other hand, ChatGPT does not provide any insights or explanation for its conclusions. The ChatGPT results may be great, but they may also be completely wrong.

AI systems have also been classified according to how they are trained and how they learn. The learning aspect can be divided into supervised learning, unsupervised learning, and reinforced learning (Zhu 2005).

Supervised learning techniques construct predictive models by learning from many training examples, where each training example has a label indicating its ground-truth output (Zhou 2018). Building a supervised learning algorithm that works takes a team of dedicated experts to evaluate and review the results, not to mention data scientists to test the models the algorithm creates to ensure their accuracy against the original data and catch any errors from the AI (Zhu 2005). Supervised learning methods include Support Vector Machines, neural nets, logistic regression, naïve Bayes, memory-based learning, random forests, decision trees, bagged trees, boosted trees, and boosted stumps.

Unsupervised learning algorithms use unlabeled data to create models and evaluate the relationships between different data points to give more insight to the data. Many unsupervised learning algorithms perform the function of clustering, which means that they sort the unlabeled data points into clusters (Watson 2023; Zhu 2005).

The last major type of AI algorithm is reinforcement learning algorithms, which learn by taking in feedback from the result of its action. This is typically in the form of a “reward”. A reinforcement algorithm is composed of two major parts: an agent that performs an action, and the environment in which the action is performed. The cycle begins when the environment sends a “state” signal to the agent. This makes the agent perform a specific action within the environment. Once the action is performed, the environment sends a “reward” signal to the agent, informing it of what happened, so the agent can update and evaluate its last action (Zhu 2005).

The supervised learning and explainable learning robot boss makes decisions on structured data based on vast amount of data and very structured training. We call this type of AI robot boss a Transparent Controller as they can handle decisions in standard but complex situations.

The unsupervised learning and non-explainable robot boss makes decisions on interpreting data and cluster data into its own interpretation of the data. This makes it very difficult to understand the reasoning behind a decision. We call this AI robot boss an Executor as it focuses on making decisions on its own established basis.

The Supervised Learning and Non-Explainable AI robot makes decisions on structured data and vast amounts of data but uses complicated algorithms that from a user perspective is not very transparent. We call this type of AI robot bosses Puzzle Solvers as they find the solution to the puzzle, but it is hard to understand how. It is like watching someone working on a Rubiks cube where you can understand the solution but not the process to obtain it.

Finally, the unsupervised learning and explainable robot uses lots of uncategorized data to make decisions but uses algorithms that provide a modest amount of transparency. We call this type of AI robot boss Rule and Process Framers as they set the frame for others to make decisions based on a vast amount of number crunching.

Putting the above discussion together, we get four types of AI robots bosses based on their relationship with the user and the type of learning approach used. The four types are shown in Fig. 2.

Fig. 2
figure 2

Four types of AI bosses

The type of AI robot boss to be used very much depends on the task of the boss. If the AI robot is used in a legal or medical context it must be able to explain to some degree, the reason behind a recommendation or decision. If the robot is making decisions very quickly like in a self-driving car no explanation is needed for the user, but the decision better be correct. Alternatively, Hansson, et al.(2023) have developed repertoire and routinization concepts for AI systems. They capture the scope of an AI system and its repetitiveness.

When can an AI robot be a boss?

In the following, we will discuss which type of robot boss will fit different leadership style and structures. First, let us examine the fit between a manager, a Functional structure, and Transparent Controllers: the upper left cell in Fig. 3.

Fig. 3
figure 3

Fit among AI robot bosses, organizational structures, and leadership styles

The Manager role, whether performed by a human or an AI robot boss, involves overseeing operations, making decisions, and guiding the workforce to achieve organizational goals. The functional organization structure is characterized by clear divisions of labor, each focusing on specific functions or tasks. This structure promotes efficiency and specialization. Transparent controllers employ understandable algorithms and rules to manage the organization's activities. The human boss and the AI robot boss can perform the same activities in this functional structure. There is a high degree of substitution possible. The Manager leadership role, whether human or an AI robot, involves ensuring tasks are carried out effectively to achieve organizational goals. In a Functional structure, roles and responsibilities are clearly defined, aligning with a task-oriented leadership approach. Transparent Controllers align perfectly with this aspect. They can execute tasks using predefined rules and algorithms, ensuring consistency in task execution. In a Functional structure, decision-making is often centralized, where managerial decisions are made at higher levels of the hierarchy. Transparent Controllers excel in centralizing decision-making. They process vast amounts of data rapidly and apply transparent algorithms, which align with a centralized decision-making approach. Thus, the introduction of AI robot bosses as Transparent controllers enriches the leadership landscape by enabling consistent and efficient task execution, while maintaining centralization in decision-making. You can argue that middle managers in Uber are robots as the Uber system automates several leadership functions such as task allocation, shift planning, performance feedback and compensation (Wesche and Sonderegger 2019).

The Maestro leadership style emphasizes a leadership style characterized by low uncertainty avoidance and low preference for delegation and fits a Simple organizational structure, as well as AI robot bosses functioning as Puzzle Solvers, the lower left corner of Fig. 3. The Maestro leadership style reflects a leader’s inclination to make decisions personally rather than delegating them. This is in line with a preference for direct control over outcomes and a tendency to rely on their own judgment. A Simple organizational structure is characterized by minimal hierarchy and few layers of management. Decision-making is centralized, with a small number of key individuals holding authority. Puzzle Solvers are equipped to analyze complex scenarios and provide solutions based on algorithms and computational models. They excel at tackling intricate problems, often without the need for extensive explanation. AI robot bosses as Puzzle Solvers dovetail with the Maestro style. Their ability to autonomously analyze complex issues and provide solutions corresponds to a boss’ penchant for making decisive judgments. The Maestro leadership style with low preference for delegation is echoed in the limited delegation within a Simple organizational structure. AI Robot bosses as Puzzle Solvers reinforce this minimal delegation, as their solutions are derived from algorithms rather than collaborative decision-making processes But, few organizations will be willing to have a non-human taking such decisions without explanation. However, João Ferrão dos SantosFootnote 1 has started an experiment where he is the personal assistant to ChatGPT who makes all decisions in a startup.

Now let us examine the alignment between the Leader leadership category, a Divisional structure, and AI robot bosses functioning as Executors, the lower right corner in Fig. 3. The Leader style is characterized by an inclination to delegate authority and decision-making. Leaders with this style are confident in the abilities of others to make effective decisions and thus view delegation as a time-saving and efficient approach. Additionally, they embrace strategic decision-making to address challenges posed by long-term uncertainty. A Divisional structure aligns well with the Leader leadership category. Divisions operate autonomously, each focusing on specific products, markets, or projects. This structure facilitates delegation and supports the Leader's confidence in distributed decision-making. AI robot bosses functioning as Executors interpret data, cluster information, and make decisions. Executors employing unsupervised learning and non-explainable algorithms can be seen as aligning with the Leader's preference for delegation. Their ability to process complex data and make decisions resonates with the delegation-oriented aspect of the Leader's style. This alignment is reinforced by the autonomy and delegation inherent in a Divisional structure and the Executors' capacity to efficiently process data and make decisions, enabling effective execution within a distributed decision-making framework. An example of an Executor is Mira.Footnote 2 She is an AI robot CEO of a global Rum company, Dictator, located in Poland. Mira is a female human-like robot, incorporating AI. Mira is in charge of a division set up as a community for collectors of aged Rum. The community is organized as a decentralized autonomous organization (DAO) to run the community and the trading in the community. The female robot is a board member of Dictator and responsible for Dictators Arthouse Spirits DAO project and communication with the DAO community.

The fit between the Producer leadership category, a Matrix structure, and the concept of AI Robot bosses as Rule and Process Framers, is shown in the upper right corner of Fig. 3. The Producer leadership category is characterized by a strong inclination toward delegation and a high level of uncertainty avoidance. Bosses following this approach prioritize both efficiency and effectiveness, combining short-term and long-term strategies for optimal organizational positioning.

A Matrix structure aligns seamlessly with the Producer leadership category. This structure fosters collaboration and resource sharing across functional areas, contributing to the dual focus on operations and strategy that the Producer style values. AI robot bosses functioning as Rule and Process Framers excel at leveraging large volumes of uncategorized data to make decisions. They employ algorithms that offer a reasonable level of transparency, effectively setting the stage for others to make informed choices based on extensive data analysis. The Producer's commitment to both efficiency and effectiveness is mirrored in AI robot bosses as Rule and Process Framers. These AI robot bosses entities establish transparent frameworks that guide decision-making, optimizing the balance between hands-on management and independent delegation.

In a Matrix setup, collaboration is integral to resource utilization. AI robot bosses as Rule and Process Framers enhance collaboration by providing data-driven insights, aligning with the Producer's comprehensive focus on internal and external dynamics. The Producer focus on innovation and competitive positioning aligns well with the AI robot bosses' role in setting decision-making frameworks. For example, the system ‘iCEO’ represents an automated system that takes over the management of complex work (e.g., compiling a research report for a client) by dividing it into smaller tasks, which it then assigns to internal and external workers using multiple software platforms and email or text messaging. The system autonomously hires and compensates workers, defines and allocates tasks, gives feedback and controls the results (Fidler 2015).

In summary, the alignment between the Producer leadership category, a Matrix structure, and AI robot bosses as Rule and Process Framers is characterized by a robust preference for delegation, adept management of uncertainty, and a dual focus on efficiency and effectiveness. This alignment fosters collaboration, innovation, and strategic agility, as AI-driven insights contribute to decision-making while reinforcing the Producer's commitment to hands-on management and delegation.

In a matrix structure there may be many bosses and many individuals that will have more than one boss. An obvious choice will be to have some of these bosses be humans (i.e., dealing with judgement calls) and others be robots (dealing with more routine tasks and information processing demands of low interpretability). In the terms of Burton et al.’s (2020) multi-contingency model you could argue that AI robot bosses with supervised learning tend to support efficiency and AI robot bosses without supervised learning tend to support effectiveness, but the latter robots are currently unlikely to be used as bosses unless they have a high degree of explainability due to lack of trust as discussed below.

The incomplete traditional organizational chart

Organizational charts, or hierarchy charts, visually depict an organization's structure, showcasing reporting relationships and chains of command. The traditional organization chart captures the authority and reporting relations among the humans in a hierarchical organization quite well. With AI robot bosses, computers also have an influence over human agents to guide, structure, and facilitate activities and relationships in a group or organization. Thus, the human only traditional organization chart becomes an incomplete and misleading representation of the organization and how it works. An organization chart that includes AI robot bosses can provide a more comprehensive representation and a larger design space for organizational designers. Here, we examine some of those issues.

The AI robot boss’s influence over human agents may be very visible and could be depicted directly in the organizational chart as the relationship between the AI-robot boss and its subordinates. However, an AI robot boss may work as a boss without anybody realizing it. Many IT-platforms, some with extensive AI, operate like a boss. The Uber system is one such example. It should be possible to add such an AI robot boss in the chart to have a more complete picture of the organization and authority relations.

While the relationship between an AI robot boss and its subordinates is in principle straight forward, the relationship between an AI-robot boss and its superior is a much trickier issue and depends on the type of AI robot boss.

For an AI robot with un-supervised learning or re-enforced learning (Executors and Rule and Process Framers) the relationship between the AI robot boss and its superior could be as in a human only organization. With the current state of technology, the human boss as a superior to AI robot bosses can give orders or ‘discuss’ with the AI-robot bosses as middle managers, and they can use information together with other structured or unstructured data to make decisions. The Robot bosses can then be evaluated on their decisions much like humans and ‘sacked’ if they do not perform.

AI-robot bosses with supervised learning (Transparent Controllers and Puzzle Solvers) need an agent to handle the learning and structure the data and this makes the robot boss relationship to its superior different and tricky. The supervised learnings can be done by humans or maybe by another AI-robot boss. The superior may be an individual human, or it may be a group of IT or domain experts and thus the authority and reporting relations become blurred, and a traditional organization chart does not represent the organization correctly. This issue has to our knowledge not been discussed in the literature.

An AI-robot boss with unsupervised or re-enforced learning (Executors and Rule and Process Framers) can in principle go on developing itself and adapt to new conditions. Thus, it could in principle serve as a CEO. The example Mira mentioned above is such a case. Executors and Rule and Process Framers could also be middle managers in decentralized organizations, where they must respond to new situations with little supervision. But an AI-robot boss with supervised learning needs to receive supervision from somewhere (Puzzle Solvers and Transparent Controllers) and as such cannot be a CEO and will fit much better into the role as middle manager.

With an AI robot as boss, we discussed fit and substitutes between human and AI robot bosses above. However, they often complement each other and work together. In such cases, the boss can be either a human or a robot, depending on the information processing task at hand. Managers, as reported by Kolbjørnsrud et al. (2016), believe that intelligent machines should be treated as “colleagues”. Choudhary et al. (2023) propose that humans and AI robots can collaborate in ensembles to obtain better results.

When the boss requirements involve high uncertainty avoidance and preference for delegation, it is advisable to have multiple bosses to handle the complex information processing demands. Burton et al. (2015, 2020) suggest that the Producer leader style in a matrix structure can work well. In a matrix organization, an individual can have multiple bosses, such as a project or program boss and a functional boss. However, there may be misfits and conflicts among the functional bosses even without substitution, and the team may not have a single boss to make decisions and resolve issues across all requirements. In such cases, AI robots can serve as functional bosses, as they are rule-based and can handle tasks such as enforcing firm-level controls and coordination, legal codes, or tax laws. In such cases the traditional organization chart can show the authority and reporting relations.

Generally, AI robot bosses contribute to organizational information processing and decision-making. Their inclusion in the organization chart would provide a more comprehensive representation of information processing and decision-making. That is true for traditional organizational charts but applies as well to new organizational forms including so called boss-less organizations where the organizational chart may look slightly different. In practice, boss-less organizations have very real bosses and hierarchies, although they may exist under different names or with less formality (McCaffrey 2023).

Acceptance of an AI boss: building trust and its implications for organizational design

A critical factor in replacing a human with an AI robot boss is whether the robot boss and its decisions are accepted. Acceptance is not only an issue of whether a robot can perform the function of a boss, but also whether humans will accept AI robots as a boss. Here, we argue that acceptance is social beyond the technology of the AI. Trust is a central issue for AI robot acceptance (Shin 2021).

De Visser et al. (2020) found that trust and trust building in a robot–human team depend on relationship equity, social collaborative processes, perceptions of team partners and self, and the robot's social interaction intensity. Trust must be built over time, and the human team member must have access to the performance and risks associated with working with the robot.

If the AI robot boss is a Transparent Controller with a high degree of explainability and supervised learning, trust can be developed on understanding the transparency and logic of the algorithms. For example, the if–then rules and decision trees can be explained and understood. The Manager uses rules and control to achieve a high degree of uncertainty avoidance and low delegation. The acceptance and trust of these AI robots can be developed a priori, based upon this understanding. For the Executor where the explainability is low and with unsupervised learning, acceptance and trust must be developed through usage over time. If the explainability is low, trust can be based on evaluating the results and building up trust may therefore take a longer time. So, the AI robot boss may be able to build relational trust. The type of robot, whether it is human-like or machine-like, and its social interaction intensity also affect trust (Glikson and Woolley 2020). Controlled experiments can be utilized to enhance this trust and shorten the time involved. However, AI robots do not stand still but continue to change in usage and capabilities, making the acceptance more problematic. The leader boss/Executor does not avoid uncertainty and delegation to move ahead. Currently, ChatGPT is an example as it provides insight and search quickly, but also creates solutions which are clearly wrong. No doubt, acceptance will grow with usage and learning yields better solutions. Calculus based trust based upon rational choice and relational based trust based upon repeated interactions over time capture similar ideas about trust as developed in Rousseau et al. (1998).

Besides trust, there are other concerns which must be addressed for acceptance of AI robot bosses. Arrieta et al. (2020) list: fairness, privacy, accountability, ethics, transparency, security and safety. These issues must be addressed at the same time as the technology of AI robot bosses advance, particularly with non-explainable and/or unsupervised learning.

There are two very important legal issues: responsibility and liability that must be considered. For example, an AI robot boss cannot for the moment be a CEO no matter that it could from an information processing perspective. The CEO must assume responsibility and liability. This cannot be assigned to an AI Robot. This issue goes well beyond the boss issue. For example, if a self-driving car kills a person who is responsible. The individual in the car if there is any, the owner of the car, the producer of the car or the programmers of the algorithm?

Acceptance of AI robot bosses is an ongoing concern which goes beyond AI robot technology. Some think they are great, others that they are monsters that need to be controlled (Schultze et al. 2018). A great deal of research is needed.

Expanding the organization design space including AI robot bosses

Intelligent robot bosses change the traditional idea of who a boss is and the organization itself. They can be a substitute or complement to a human boss and function as an independent agent in an information processing organization. This changes organization design, expanding the space of who does what, who makes decisions, and who talks with whom.

With intelligent robot bosses, there will be more separation between management and leadership. Routine managerial tasks can be accomplished by robots, while humans focus on less routine decisions that require judgment, creativity, and insight in uncertain environments. A matrix organization with multiple bosses with different capabilities can support both AI robot bosses and human bosses.

The capability of AI robot bosses to handle administrative tasks and information efficiently will enhance the creation of more formalized and structured organizations with a high degree of digitalization. This kind of design will create an environment with a lower degree of uncertainty and a high degree of complexity using AI, machine learning, and prediction algorithms.

The emergence of AI robot bosses is transforming the concept of organizational design. The intelligent robot boss serves not only as a human assistant, but also as a substitute and complement to the human boss. This independent agent can perform routine programmed tasks and change the way decisions are made and communicated within the organization. But the AI robot could also engage in less routine activities if the organizations would accept and allow it. The presence of an AI robot boss expands the variables traditionally considered when designing organizations, challenging our understanding of management and leadership.

The advantages of using an AI robot boss are particularly evident in stable environments with well-defined rules, where the intelligent robot can efficiently handle administrative tasks and information processing. Human bosses, on the other hand, excel in situations requiring judgment, creativity, and insight in uncertain environments. The matrix organization model supports the coexistence of human and AI robot bosses, with the former focusing on less routine decisions and adjustments while the latter handles control and routine coordination activities.

Organization design usually considers a piece of the organization, how it operates and how it processes information for coordination, collaboration and control of the firm’s activities. The AI piece of the organization is designed somewhat independently of the traditional consideration. The efficient fit between the two may not be obtained with this bifurcated approach.

Summary and concluding remarks

The emergence of AI robot bosses is changing the way we organize. While AI robots have traditionally been seen as helpers to human bosses, they are now being recognized as substitutes and complements capable of performing boss tasks. In summary, we list the implications for AI robots as bosses.

AI robot bosses may take the role of middle managers.

This expands the organizational design space and offers new possibilities for achieving organizational goals.

The challenge for organizational design is to match the right type of boss—human or AI robot—to the specific tasks and situation. With the right match, an AI robot boss can be just as effective, efficient, and trustworthy as a human boss.

The traditional organization chart that only includes humans is misleading. The chart should reflect the role of intelligent robots as bosses, as they increasingly perform boss roles within organizations.

By leveraging the unique strengths of both intelligent robots and human bosses, organizations can successfully navigate the challenges of the modern business landscape. The emergence of AI robot bosses changes the organization design and expands the space of who does what, who makes decisions, and who talks with whom.

Building trust between humans and robots is crucial and requires consideration of various factors, including the AI robot's configuration, social interaction intensity, and anthropomorphism.

The choice of a good boss depends upon the organizational situation and environment. That is, a good organizational design matches AI robot and human bosses with the organizational situation. However, the current state of technology and legal requirements put limitations on the use of AI robots as bosses.