By Michel Pires de Araujo and Renata Biana da Silva
Throughout the history of management theory, the understanding of organizations has been shaped by various metaphors, each illuminating specific facets of their complexity while simultaneously obscuring others.
Gareth Morgan (2002), in his work “Organizational Images,” posits that all organizational and management theory and practice is based on images, or metaphors, that lead us to understand situations effectively, albeit partially. Metaphor, as a comparative figure of speech, acts as a force through which human beings create meaning, utilizing one element of their experience to understand another.
From mechanistic views that conceived of organizations as efficient and predictable machines, to the organic perspective that saw them as living systems adaptable to the environment, to the metaphor of the brain, which highlighted its capacity for learning and self-organization, the field of management has relentlessly sought new lenses to decipher organizational dynamics. In a scenario increasingly characterized by algorithmic decision-making and the proliferation of data, a new metaphor emerges as promising: that of “organizations as algorithms.”
While classic images of machines, organisms, brains, cultures, political systems, psychic prisons, flux and transformation, and instruments of domination continue to offer tools for analyzing organizational dynamics, the rise of artificial intelligence and algorithm-assisted decision-making demands a broadening of the interpretative lens.
In this sense, Glaser, Sloan, and Gehman (2024) propose that organizations can be viewed as algorithms, offering a nuanced understanding of the interconnections between data, algorithmic processes, and agency in an organizational landscape increasingly mediated by advanced technologies. This vision, the authors suggest, challenges traditional dichotomies by repositioning agency within sociotechnical arrangements, highlighting the transformative role of programming and prompting in the continuous reassessment of how organizations function in the digital age. However, we do not intend to delve into the merits of the use of the word “algorithm” without a detailed investigation of the origin of the construct, so that we can proceed with the proposed analysis.
Therefore, in 1937, Alan Turing introduced the famous “Turing machine,” considered the most robust and objective formalization of the concept of an “effectively computable function,” that is, the idea of a problem that can be solved through the execution of an algorithm. From this contribution, the investigation of which problems admit a computational solution became pertinent and systematic, as well as the relative degree of complexity of functions that, although computable, can differ considerably in the difficulty of their practical execution (Abreu, 1987).
In this sense, according to Donald Ervin Knuth, an American mathematician, renowned computer scientist, and professor emeritus at Stanford University, an algorithm is a finite sequence of well-defined rules that describes a step-by-step procedure for solving a specific type of problem.
Richardson (1919), in turn, explains that a system is called deterministic when its state, at any given moment, can be accurately predicted if we know sufficient information about it at previous moments and if we know the rule that links this information to the future. Thus, a deterministic system is a predictable system in which nothing happens by chance – everything follows a clear rule.
Thus, in the logical and operational sense, an algorithm is deterministic by nature, an intentionally designed deterministic system.
Artificial intelligence-based systems are trained using specific machine learning algorithms. For Shalev-Shwartz and Ben-David (2014), a machine learning algorithm is the computational mechanism that, given training data, produces a form of expertise—typically another program or model—capable of performing a specific task. This expertise is derived from the automated identification of relevant patterns in the data.
Kannegieter (2025) warns that non-determinism, in the context of Large Language Models (LLMs), means that the model can produce different outputs even when given the same input. For the author, a system based on generative artificial intelligence cannot be considered a deterministic system because, by technical definition, it is stochastic (incorporating an intrinsic degree of randomness in its output generation processes). This behavior is a byproduct of the complex neural networks and vast amounts of data used to train these models. Thus, while nondeterminism can lead to creative and diverse results, it can also cause inconsistency, which can be undesirable in certain business applications.
Calling organizations algorithms is an imprecise metaphor that ignores central premises of the very concept of an algorithm. Since the Turing machine in 1937, we have known that an algorithm is, by definition, a finite sequence of clear steps to solve a problem—that is, a deterministic procedure (Knuth, 1997; Richardson, 1919). It is predictable: given the same input, it always produces the same output.
Thus, saying that an organization “is an algorithm” ignores the fact that algorithms, in the technical sense, are intentionally designed to operate without surprise—something impossible in generative systems.
Of course, authors like Glaser et al. (2024) attempt to broaden the meaning of “algorithm” beyond classical computing, treating it as part of a living sociotechnical whole—an assemblage of routines, data, and decisions. But stretching the concept so far undermines its technical power. An algorithm, in Turing and Knuth’s sense, is neither dynamic nor subject to stochastic variation: it is a clear rule for solving a specific problem. Organizations are not like that.
Therefore, each of the metaphors discussed here sought, albeit unsuccessfully, to illuminate distinct aspects—from mechanical efficiency to organic adaptability, from the brain’s learning capacity to the symbolic complexity of culture, from the dynamics of political power to the depths of the unconscious, and the fluid nature of transformation. However, Morgan himself, by warning about the bias inherent in each metaphor, emphasizing the importance of a “diagnostic reading” that combines multiple perspectives to deal with complexity, issued a warning to researchers when conducting their research—to never claim that their metaphors—like that of the algorithm—are the ultimate when investigating organizational reality.
Therefore, the metaphor he proposes—that of the organization analyzed through the image of the algorithm—fails doubly: technically, because it equates deterministic systems with stochastic models; conceptually, because it transforms a precise notion into a vague label. Perhaps it would be more honest to speak of organizations as hybrid systems, in which algorithms, nondeterministic models, and human factors coexist—without reducing everything to a prescriptive step-by-step procedure. Vandekerckhove and Emmanuel (2025) reinforce that constant innovations in information technology and artificial intelligence have been transforming the way organizations are understood, which have come to be compared to “cyborgs” — hybrid entities, made up of human and technological elements, whose dynamics are largely driven by decisions mediated by algorithms.
REFERENCES
ABREU, Nair Maria Maia de. A teoria da complexidade computacional. Revista Militar de Ciência e Tecnologia, Rio de Janeiro, v. 4, n. 1, p. 90-95, jan./mar. 1987. Disponível em: https://rmct.ime.eb.br/arquivos/RMCT_1_tri_1987/teoria_complex_comput.pdf. Acesso em: 14 jun. 2025.
GLASER, V. L.; SLOAN, J.; GEHMAN, J. Organizations as Algorithms: A new metaphor for advancing management theory. Journal of Management Studies, v. 61: 2748-2769, 2024. https://doi.org/10.1111/joms.13033
KANNEGIETER, Trent. Nondeterministic Torts: LLM Stochasticity and Tort Liability. Available at SSRN 5208155, 2025.
KNUTH, Donald Ervin. Fundamental algorithms. The art of computer programming, v. 1, p. 261-268, 1997.
MORGAN, Gareth, 1943 – Imagens da organização: edição executiva/Gareth Morgan; tradução Geni G. Goldschmidt. – 2. ed. – 4a reimpressão – São Paulo: Atlas, 2002.
RICHARDSON, C. A. The notion of a deterministic system. The Philosophical Review, v. 28, n. 1, p. 47-68, 1919.
SHALEV-SHWARTZ, Shai; BEN-DAVID, Shai. Understanding machine learning: From theory to algorithms. Cambridge university press, 2014.
STANFORD UNIVERSITY. Donald Knuth. Disponível em: https://profiles.stanford.edu/donald-knuth. Acesso em: 14 jun. 2025.
VANDEKERCKHOVE, Wim; EMMANUEL, Myrtle. Inherent normativity of metaphors: ethics, organizations, and moral imagination. Philosophy of Management, v. 24, n. 2, p. 207-230, 2025.
