If I can digress for a moment, it is common knowledge that AI often provides responses that “sound credible.” This is not simply because AI retrieves answers commonly found on the internet. It is because AI’s learning process includes a mechanism for optimizing its responses so as to “increase user satisfaction.” As a result, even when a question for which there appears to be no appropriate answer is asked, AI conjectures the kind of response that its user is presumed to be seeking, and then generates a response that “sounds plausible.”
For AI developers, gaining positive user reviews is essential for widespread acceptance, and as a result, AI ends up with a tendency to please users. Put differently, if, for the sake of accuracy, AI flat-out rejected user opinions or downright dismissed a query saying that “there is no answer,” it would not have been adopted this widely.

Let’s go back to the concern I have mentioned. I am worried that if businesses use AI having such characteristics not as a tool but as a partner, it could undermine diversity, an essential and critical element in management. This element of diversity has attracted growing attention in recent years, as it has become recognized that enhancing enterprise value in an increasingly complex business climate requires multifaceted perspectives and diverse viewpoints in business execution and decision-making. Meanwhile, if businesses continue to use AI as a partner without fully understanding its characteristics, AI will learn more and more about its users and, before long, users might find themselves surrounded solely by AI models that affirm their views.  This poses a risk of undermining diversity in the thinking process.
In particular, the problems business leaders face most often do not have a single solution. This is exactly the situation where having diverse opinions will be essential. But advanced, well-trained and proficient partner-type AI is liable to clearly agree with senior management’s initial hypotheses. This is where one of the key risks of AI-enabled management lies. The risk becomes even greater when users view AI as being free from the complex relationships and biases that exist among people, and therefore assume that AI can provide inherently objective and candid opinions.

To control these risks, it is necessary to revisit the roles we have assigned to AI and how we use it. For instance, one measure would be to create a rule that would designate a tenure for partner-type AI, just as there are terms of office for people in positions, so that a revamp will be implemented periodically. Another idea would be not just to use AI that affirms our viewpoints, but also to employ AI that denies or rejects them.
 
Along with these approaches, businesses may also start seeking different qualities and attributes in their employees. If AI’s agreeable tendency is to remain at least to some degree, businesses should seek even further diversification of perspectives in hiring, promotion, and other aspects of their human resources strategies. For individual people, it will be more important than ever to have strong character that is not influenced by AI and to be receptive and listen to opinions that are different from their own.  

Today, corporate management transformation is impossible without AI. AI capabilities should be recognized as one of the major elements of managerial capital. In fact, NRI has pledged to proactively bolster its AI capabilities under the medium-term business plan that started in April 2026.
At the same time, looking ahead, there will likely be situations where the aforementioned risks actually need to be controlled. As such, I believe it is important to consider how to position AI from the perspective of diversity in management and how to use AI as a partner. To achieve this, I would first like to engage in thorough and meaningful discussions with my fellow human beings.