AI- ready company

target readers-cv

By Mostafa Sayyadi and Luca Collina  

Having technical capacity does not always guarantee that the team will question outputs, collaborate to draw on expertise, and assume responsibility for the decision.  

The executive team can initiate an AI discussion with a simple question: “Where could this technology make our work faster?” The question is important but incomplete. Once the AI system starts to draft, classify, recommend, or make a decision, the manager has to know who will control the work, whose experience will be needed, how concerns could be raised, and who will be responsible for the decision. 

These are Organizational, not Technological, Questions. 

The point is often overlooked when managers approach the introduction of AI from the perspective of time savings. Time is an input to performance, not a measure of success. The work may be done faster while mistakes get transferred downstream or experienced workers spend time correcting the output that they did not participate in preparing. The proper question is which outcome was improved, for whom, and which additional work was done alongside this outcome. 

The AI tool may make the work of an individual faster but not improve the team’s ability to evaluate options, see the context, and make a decision. The increase in speed can coexist with a hidden decrease in the ability to question, explain, and own the decision. Bill Campbell’s career helps us understand this dilemma. Bill Campbell went from coaching the football team of an American university to technology and business leadership. Later on, he became known as an executive coach of Silicon Valley’s management teams. The value of the case is not in the attempt to emulate a legendary coach. It lies in the connection made by Campbell between the quality of relations and the quality of work. 

This connection becomes particularly crucial when AI enters the workplace. AI can allow people to produce more but does not replace the need to evaluate the context, challenge the assumptions and assume responsibility. Campbell’s approach can serve as a guideline for work related to AI implementation. 

The Team is the Unit of Performance. 

Campbell’s first piece of advice to managers in the era of AI is that the team, not the star, should be considered the unit of performance. AI can encourage a more individual perspective on productivity. A person can make a draft, analyze a set of documents, or prepare a presentation faster. The organization measures this advantage and assumes that the performance was improved. But a quicker individual task does not necessarily mean better organizational performance. 

The completed report may lack a history of the customer; the concise summary may not include an important exception to consider from a financial or operational perspective. The output can be distributed across different roles but lacks the expertise to do it. It is an issue of how the work is shared and evaluated rather than a technical mistake. 

The approach of Campbell suggests that a manager should consider how an introduction of AI will change the relationships among different roles. Who provides the context? Who controls the output of the system? Who identifies the inconsistencies with the customer’s case? Who has the authority to stop the process? If every person gets access to the same tool but continues working separately, the organization will have many users but no team that knows how to use AI. 

Every output has to be evaluated by a committee. But the review has to correspond to the task and consequences of the error: a routine output may be evaluated lightly, while a recommendation concerning the customer, employee, financial report, or the process covered by regulations should be controlled with different expertise and responsibility. The quickness should not hide the relationships on which the quality relies. 

A manager can look at the workflow: where an output is prepared, where it is used, what expertise is needed for the assessment, and who owns the result. Then the manager decides if some roles have to be connected, outputs have to be controlled differently, or the work has to be elevated. It is the issue of the team’s capability rather than access to a tool. 

Access alone is not capability. A team has to have shared criteria for when it is allowed to use AI, what has to be confirmed, and how learning from one user to another takes place. Otherwise, the organization will be dependent on its invisible individual practices. The manager may not know which tasks are supported by AI and which outputs require different control. 

Trust Allows for the Challenge. 

Trust was crucial for Campbell’s approach. When using AI, employees have to be able to tell that the output is incorrect, uncertain, or goes beyond the scope of their expertise, and disclose how they used a system. When admitting the problem means blame, employees will not reveal it until it is costly. 

The trust among people is different from the confidence in an AI system. A team may trust each other and have to control a system carefully; it may have a well-tested system and still lack the confidence to report its poor performance. Both points are important, and none should be taken for granted. 

Trust also requires transparency about the data. Employees have to know what systems are approved, what information should never be entered, and what records they create when they use a system. These measures are essential, but they have to be transparent. Monitoring without any purpose and limitations will push employees to conceal rather than report the usage. 

The trusted AI process describes the purpose of the process, the limitations, required reviews, and channels for concerns. It allows saying, “I am not sure if it is correct.” In tasks that require a decision, it can be the first step to a good decision. 

Campbell’s combination of support and candor is useful for managers: they can welcome the concerns of employees and still require the evidence. But the managers also have to be open about the fact that the tool was chosen badly or introduced inappropriately. The caring attitude without the criteria leads to indifference; the harsh criteria without trust lead to a lack of information. 

The Care also Involves the Work Itself. 

Campbell’s care about people is usually interpreted as a personality trait. But the managerial value of care is practical. A leader who knows the pressures and potential of his colleagues is able to make better decisions about their work. The adoption of AI increases the necessity of this kind of knowledge because the same change can mean assistance for one employee and a threat for another. 

Some employees may appreciate the assistance with routine tasks. Others may be afraid that AI will change the importance of their skills or reduce their control over the work. They may be asked to review the increased output without any additional time or have changing roles without any explanation of what comes next. Managers have to take these reactions as information about the change rather than assume that the silence means consent. 

The care does not mean a guarantee that every role will not change. It means the explanations about what is known, what is not, and involvement of people in reshaping the work. The managers can ask which tasks should be changed, where judgement is needed, and what skills employees want to develop. 

The development support has to be practical. The managers can give the time to practice, access to an expert, and the opportunity to discuss the future of the role. This indicates that adaptation is a shared responsibility, not a requirement for the employee after the tool is adopted. 

The output alone will not show if AI is increasing the capability or just adding work. If the system automates one of the steps, managers should ask if it is freeing time or shifting efforts to verification and correction. It is the issue of the work design, not the resistance to technology. 

Listening is a Way to Understand the Work. 

Listening was one of the main practices of Campbell. In the context of AI, listening is not only a demonstration of respect; it is a way to uncover the information that formal measures cannot show. Senior leaders can see adoption rates, speed, and costs of the process. The people working with the system can know that the output requires constant correction, that the questions of the customers are being incorrectly routed, or that an important exception is being omitted in the summary. 

The people who perform the work and the senior leaders see different aspects of the process. Listening brings these perspectives together. Managers can ask how the process has changed, where the system helps, where it generates additional work, and what happens when the recommendation contradicts the actual situation. The objective is to learn the workflow, not collect complaints. 

The specific questions will allow for collecting the evidence: what did you check, what was missing, when you rejected an output, and what prevented the escalation. These answers will show if training, access, controls, or role design have to be reconsidered. 

Listening also has to transform the metrics that the leaders use. Adoption rates show the access and usage of the system; they do not show if the performance got better. The teams have to have the metrics that will allow recording corrections, escalations, and cases when a human decision changed the output of AI. 

Listening has to continue after the launch. The system may be used for new purposes; employees may find workarounds, or the workload may reduce control. One-time consultations before the launch will hardly identify these changes. 

Managers Must Coach Judgement. 

As AI gets better at giving out the answers, managers might have to spend less time providing information and more time helping employees learn how to formulate the right questions, when to trust the outputs, and when to seek additional assistance. This is a coaching task. A manager who provides the right answer too fast will miss a great opportunity to develop a judgement of the team in situations when something goes wrong, or the context is different. 

Coaching starts with questions. What kind of decision do we have to make? What are the assumptions on which this output was based? What facts will change our decision? What will make this action unacceptable? Who has to check this decision before its approval? These are the questions and not a substitute for technical training. They link the technical capacity of a person with the understanding of his or her responsibility. 

These questions also make it clear that a manager is not supposed to know everything. A manager has to understand when specialist advice is required and to make sure that there is enough space for this advice to influence the decision. 

Coaching can help employees preserve and exchange experience. When AI standardizes some processes, managers should ask about the judgement that made these possible and how new employees will learn to manage exceptions. In such a case, the immediate output will become better, and the potential capacity of the team will become worse. 

The senior leaders also have to be coached. The decisions of these leaders are related to the investments, expectations, and risks, while the colleagues will probably be hesitant to criticize them. A trusted adviser will be able to look at the assumptions and team dynamics. Coaching does not mean that the capable leader needs some correction. 

Human Leadership is Necessary, but not Sufficient. 

It would be incorrect to interpret Campbell’s people-focused approach as the whole solution to the AI problems. Trust cannot provide information about the accuracy of a model. Listening is not a substitute for security controls. Care for the employees will not provide information on whether the client data can be used in the system. Even the best team can make the wrong decision if it does not possess the right data or technical skills. 

AI governance has to include both the operational controls and good relationships. Organizations have to understand which systems they use, what kind of information is processed, how it is tested, who can change settings, and how incidents. The suppliers’ claims, access rights, and data processing need to be checked; some of the decisions may require the review of specialists. 

Control works better when people understand it and can report their concerns. The useless policy will not protect the organization; the technical tests conducted without considering the work can miss the problem cases. Control sets the boundaries, while leadership helps people to understand when these boundaries are approached or violated. 

Campbell’s principles do not guarantee that the system is fair, safe, and accurate. They provide the conditions under which it is possible to ask about all these things and do something. Technical testing and legal review are necessary, especially if the consequences affect people who did not choose to use the system. 

It would also be incorrect to copy Campbell’s methods. His leadership style is based on his experience, personality, and relationships. Other leaders will be able to build trust in another way. A useful principle is not imitation but the discipline of taking relationships, team composition, candor, and decision ownership into account as the key aspects of organizational performance.

About the Authors

Mostafa SayyadiMostafa Sayyadi is an international management consultant who advises senior business leaders on developing innovation and improving leadership effectiveness. He works with organisations ranging from start-ups to Fortune 100 companies, helping leaders strengthen their capabilities and drive business success. 
 
lucaLuca Collina is a management and transformational consultant who has managed transformational projects and Automation internationally (Tunisia, China, Malaysia, Russia). He now helps companies understand how GEN-AI technology impacts business, use technology wisely, and avoid problems. He has an MBA in Consulting, has received academic awards for his research, and is a published author. Thinkers360 named him one of the Top Voices, Globally and in EMEA in 2023. Luca continuously upgrades his knowledge with experience and research to transfer it. He recently developed interactive courses on “AI & Business” and “Human Centric AI”. 

LEAVE A REPLY

Please enter your comment!
Please enter your name here