AI will not revolutionize business management but it could make it worse

It is no exaggeration to say that the democratization of new forms of artificial intelligence (AI), such as ChatGPT (OpenAI), Gemini/Bard (Google) and Copilot (Microsoft), is a societal revolution of the digital age.

The mainstream use of AI systems is a disruptive force in a number of areas including university education, the legal system and, of course, the work world.

These changes are taking place at such a bewildering pace that research is struggling to keep up. For example, in just a few months, the ChatGPT platform has improved so much that it now has the capacity to rank among the top 10 per cent of the best scores on the Uniform Bar Exam in the United States. These results are even encouraging some U.S. law firms to use AI software to replace the work of some paralegal workers in detecting a judge’s preferences to be able to personalize and automate pleading.

However, while the technological advances are remarkable, the promises of AI do not square with what we have learned in over 40 years of research in organizational psychology. Having worked for many years as an expert in strategic management, I will shed some distinct — but complementary — light on the sometimes dark side of organizations, i.e. behaviours and procedures that are irrational (or even stupid), and I will look at the impact that these have when AI is added to the package.

Stupid organizations

Have you ever found yourself in a professional situation where your idea was invalidated by the answer, “The rules are the rules,” even though your solution was more creative and/or less costly? Congratulations! You were (or still are) working in a stupid organization, according to science.

Organizational stupidity is inherent, to varying degrees, to all organizations. It is based on the principle that human interactions are, de facto, inefficient and that processes to control work (e.g. company policies), unless they are regularly updated, run the risk of making an organization, itself, stupid.

While some organizations work hard to update themselves, others, often for lack of time or in search of day-to-day convenience, maintain processes that no longer fit with the reality that the organization is facing — and they, then, become stupid. Two elements of organizational stupidity can be put forward: functional stupidity and organizational incompetence.

Functional stupidity

Functional stupidity occurs when the behaviour of managers in an organization imposes a discipline that constrains the relationship between employees, creativity and reflection. In such organizations, managers reject rational reasoning and new ideas and resist change, which has the effect of increasing organizational stupidity.

This results in a situation where employees avoid working as a team and devote their professional resources (e.g. their knowledge, expertise) to personal gain rather than that of the organization. For example, an employee might notice the warning signs of a machine failure in the workplace but decide not to say anything because “it’s not their job,” or because their manager will be more grateful to them for fixing the machine than for preventing it from breaking down in the first place.

In a context of functional stupidity, integrating AI into the workplace would only make this situation worse. Employees, being restricted in their relationships with their colleagues and trying to accumulate as many professional resources as possible (e.g. knowledge, expertise, etc.), will tend to multiply their requests to AI for information. These requests will often be made without contextualizing the results or without the expertise required for the analysis.

Take, for example, an organization that suffers from functional stupidity and that, traditionally, would assign an employee to analyzing market trends and then pass this information on to another team to set up advertising campaigns. The integration of AI would then run the risk of encouraging everyone in the organization (whether they have the expertise to contextualize the AI’s response or not) to look for new market trends in order to have the best idea in a meeting in front of the boss.

We already have some examples of functional stupidity cropping up in the news; for example, in a trial, a U.S. law firm cited (with help from ChatGPT) six jurisprudence cases that simply do not exist. Ultimately, this behaviour reduces the efficiency of the organization.

Incompetent organizations

Organizational incompetence lies in the structure of the company. It is the rules (often inappropriate or too strict) that prevent the organization from learning from its environment, its failures or its successes.

Imagine that you are given a task to complete at work. You can complete it in an hour, but your deadline is set for the end of the day. You may be tempted to stretch the time required to complete the task to the limit, because you have no advantage in completing it earlier, such as an additional task to complete or a reward for working quickly. As a result, you are practising the Parkinson’s principle.

In other words, your work (and the cognitive load required to execute it) will be modulated to meet the entire prescribed deadline. It is difficult to see to what extent the use of AI will increase work efficiency in an organization with a strong tendency towards the Parkinson’s principle.

The second element of organizational incompetence relevant to the integration of AI into the workplace is the principle of “kakistocracy,” or how individuals who appear to have the least competence to hold managerial positions nevertheless find themselves in those positions.

This situation arises when an organization favours promotions based on employees’ current performance rather than their ability to meet the requirements of new roles. In this way, promotions stop the day an employee is no longer competent in the role they currently perform. If all promotions in an organization are made this way, the result is a hierarchy of incompetent people. This is known as the Peter principle.

The Peter principle will have even more negative effects in organizations that integrate AI. For example, an employee who is able to master AI more quickly than their colleagues by writing programming code in record time to solve several time-consuming problems at work, will have an advantage over them. This skill will put them in good standing when it comes to their performance appraisal, and may even lead to promotion.

Incompetence and inefficiency

However, the employee’s AI expertise will not enable them to meet the conflict resolution and leadership challenges that new management positions bring. If the new manager does not have the necessary interpersonal skills (which is often the case), then he or she is likely to suffer from “injelitance” (a combination of incompetence and jealousy) when faced with these new challenges.

This is because when human abilities have to be brought to the forefront (creative thinking, the emotional aspect of all human relationships) and we reach the limits of AI, the new manager will be ineffective. Feeling incompetent, the manager will need more time to make a decision and will tend to find solutions to non-existent problems in order to put forward their technical skills and justify their expertise to the organization. For example, the new manager might decide that it is essential to monitor (using AI, naturally) the number of keystrokes made per minute by employees in their team. Of course, this is in no way an indicator of good performance at work.

In short, it would be wrong to think that a tool as rational as AI, in an environment as irrational as an organization, will automatically increase efficiency the way managers hope it will. Above all, before thinking about integrating AI, managers need to ensure that their organization is not stupid (in terms of both processes and behaviour).