AI improves work only when leaders redesign organizations
When a system detects an anomaly and signals the need for action, who checks whether it’s accurate? Who decides how urgent the problem is? And who is accountable for the consequences if action is delayed? These are questions companies should always be asking, but they are even more critical when artificial intelligence enters into business processes.
When AI is introduced in organizations, its effects on work may take some time to emerge. First, companies might purchase licenses or launch digital assistants, but only later do people's decisions, responsibilities, and skills begin to change. Writing a report more quickly is one thing, but building an enterprise capable of making better decisions is something else entirely.
This distinction came up in discussions among the leaders of companies brought together by SDA Bocconi's Humans & Organizations Lab (Human.O Lab ). They have experience with practical AI applications—tools that classify documents, facilitate access to internal information, analyze data, generate forecasts, or flag anomalies—finding that some save time, and others help them identify opportunities they might otherwise have missed. But the most ambitious initiatives are beginning to transform work processes in their entirety, provided that the evolution is guided by informed leadership.
To explain, if an AI assistant can prepare an initial assessment of a complex case, someone must determine what information it can use, when to call in an expert, and how to document the final decision. If a model recommends a maintenance intervention, the relationship between automated predictions and the judgment of the people familiar with the equipment must be clearly defined. Technology makes new solutions possible; but the value of these solutions depends on these organizational choices.
For a long time, we imagined a good boss as someone who knows the answers, develops a plan, and ensures that it gets executed. With AI, the number of available answers multiplies, and plans can be continuously updated. This amplifies the need for leaders who can map out how to organize work and distribute it between people and intelligent systems. They must assign tasks, create opportunities for discussion, establish criteria for verification, and make room for experimentation. In addition to all this, they need to exercise oversight differently, drawing on new skills.
The first responsibility is to safeguard human judgment. An analysis produced in seconds can appear convincing even if it overlooks a crucial detail. So managers have to understand where AI can help generate hypotheses and where people need to examine the facts first. Leaders must ask their teams to explain their choices, evaluate alternatives, and challenge plausible recommendations, because a rapid fire answer does not necessarily equate to a high quality decision.
The second responsibility concerns learning. Companies are investing in training employees to use AI tools, often through internal communities and individuals tasked with disseminating expertise. This is an essential foundation, but beyond that, managers must be able to recognize system errors, understand what checks to run on AI- generated results, and revise the process when experimentation reveals limitations. To do all this, they need an environment where people feel free to voice their doubts and report any mistakes. If every uncertainty is concealed for fear of appearing unprepared, the organization won’t learn much, regardless of how many tools it has at its disposal.
Finally, there is a responsibility that ties into relationships. During a recent training program, AI agents were used to ask trainees questions and encourage them to consider different perspectives. Interaction increased, but participants tended to focus on the platform, and made less eye contact with their peers. Even a tool designed to help a group think can end up changing how its members listen and discuss ideas. Leaders must be attuned to both effects.
The widespread adoption of AI should not be mistaken for evidence of transformation. The real difference lies in better decisions, more time for higher-value activities, sharper skills, and greater clarity about individual responsibilities. Although these results are harder to measure than, say, the number of licenses activated, they are the outcomes that are worth investing in.
Artificial intelligence offers us an opportunity to rethink work. To fully exploit it, business leaders will need to place less faith in the mere availability of technology and pay more attention to how they design the organizations that use it.
