‍

By Guest Autors:
Antonio de los Campos

Head of Strategic Planning at Bantotal and Founder at PMLA®

Shoham Adizes

Senior Associate at Adizes Institute 

‍

Many companies adopted AI expecting a simple result: higher productivity, lower costs, and better profits. Yet for many, the bottom line has barely moved.

MIT’s 2025 GenAI Divide report found that only a small minority of enterprise AI pilots are producing measurable P&L impact. McKinsey calls this the “gen AI paradox”: widespread adoption, but limited enterprise-level earnings impact. BCG similarly reports that regular AI use is now common, but value is concentrated among companies that redesign how work gets done, not merely those that deploy AI tools.

So the problem is not adoption. AI has been adopted. The problem is that many organizations are not designed to convert AI-enabled productivity into organizational results.

The mistake comes from a common but flawed assumption: if each employee becomes more productive, the organization becomes more productive.

That sounds logical, but it misunderstands what an organization is. An organization is not merely the sum of its individuals. It is a system of interdependencies. If employees are points, the organization is not the points; it is the lines connecting them.

Let's look at this with an illustrative example: During the 2021–2023 seasons, the football club Paris Saint-Germain brought together three of the world’s most talented players: Lionel Messi, Neymar Jr., and Kylian Mbappé. From an individual perspective, each of them represented a significant increase in “productivity,” greater goal-scoring ability, playmaking, and technical talent. However, the team failed to translate that potential into winning major international titles, particularly the UEFA Champions League.

This phenomenon was not due to a lack of quality among the players, but rather to the absence of a system that maximized their interaction.

In systemic terms, the team did not optimize the network of relationships, but simply increased the quality of the nodes.

What happens on a soccer field happens in organizations. Productivity is not a property of individuals. It is a property of the system.

AI may help individuals write faster, analyze faster, code faster, summarize faster, and decide faster. But faster individual output does not automatically become better organizational performance. If the capabilities change or the constraints are removed, one must modify the entire system to optimize it. It is not true that a system that is optimal without AI will also be optimal with AI, as we saw in the soccer example. This explains why AI has had a significant impact when evaluated at the individual level but a limited impact at the organizational level.

At the Adizes Institute, we study organizations as systems. From that perspective, the question is not whether AI has been integrated sufficiently. The question is whether the organizational system is being properly reconfigured to benefit from AI.

That system has at least three critical dimensions.

First is the managerial process: how the organization identifies problems and opportunities, discusses them, makes decisions, communicates decisions, implements them, and follows up. If this process is weak, AI will generate more information, more options, and more proposals — but not necessarily better decisions. The bottleneck is no longer access to information. The bottleneck is the organization’s ability to turn information into coordinated action.

Second is structure: who has authority over what, who is responsible for what, and how people are evaluated and rewarded. AI gives people access to more information and more analytical power. That means people can now make bigger recommendations, challenge more assumptions, and act with more confidence. But if authority and responsibility are unclear, AI can encourage people to overstep their roles. Decisions may be made by those with information but without authority, or blocked by those with authority but without understanding. In addition, it is very likely that changes will need to be made to the workflow, that some roles will be eliminated, others will be modified, and new roles will emerge.

In both cases, individual productivity does not translate into organizational results.

Third is common mission, vision, values, and strategy. AI increases individual capacity, but capacity must be aimed. If people are pulling in different directions, AI only helps them pull harder. One person becomes faster at selling, another faster at cutting costs, another faster at launching initiatives. But if there is no common direction, the organization does not accelerate. It spins.

Even if we agree with all the above, many organizations fail because their strategy is flawed. Every strategy is a set of assumptions about cause-and-effect relationships. If these assumptions are false, the strategy will be useless. So far, we have discussed the importance of adapting the system, but we never said that changes shouldn’t be made to the people. For organizations to become more productive, the people within them must adapt, changing both what they do and how they do it. The problem is that most organizations believe they must influence people to get them to modify the system, but from our perspective, the order is reversed. Organizations must modify their systems, and that will cause people to change.

AI can make people more productive. But only management can make the organization more productive. AI can strengthen the parts. But the bottom line changes only when the system enables the parts to work better together.

Guiding Principles:

1. Choose a cross-functional process that has a significant impact on the organization’s results, and redesign it from end to end, integrating AI as a native component.
2. Reconfigure the organization’s structure: roles, responsibilities, levels of authority, and compensation systems, taking into account which roles will disappear, which will be modified, and which need to be created.
3. Revise the organization’s vision, given that many constraints have been lifted with the availability of generative AI.

Practical recommendations:

1. Assign your best people to kickstart the change process: as long as the organization continues to operate, no one wants to bear the opportunity cost of assigning a valuable employee to this change process, and they end up delegating the task to a junior who is not equipped to handle the challenge.
2. Full-Time Dedication: What teams consider urgent ends up overshadowing what is important. Accept the cost of change—there is no such thing as a free lunch.
3. Decide what not to do: AI has the potential to drive productivity improvements across multiple areas. This can become a problem. Define a key metric you want to impact and assess the impact of your actions before you begin. Often, teams focus on areas that are important to a specific team but fail to make a meaningful difference at the corporate level.