Introduction
As a new entity, "AI subordinates," begins to enter the workplace, what should we be mindful of? While we tend to focus solely on improving productivity, we need to seriously reconsider the underlying issues of accountability and risk management.
This article provides a clear explanation of the "responsibilities of management," "increased burden on middle managers," and "the ideal approach to risk management" that organizations implementing AI should be aware of.
The essence of the AI era: organizational structures where humans can take responsibility.
The most important question will be, "Can they fulfill their accountability?"
When utilizing AI, it's crucial not to rely solely on technological advancements, but to maintain a fundamental perspective: "Who is responsible?" Legally, only corporations and individuals can bear responsibility, and this is fundamental to any organization.
Three important points
Point 1: Thoroughly implement the "Human-in-the-loop" concept.
It is essential to have a system in place where humans are involved in every stage of the AI process, performing checks and making decisions. Instead of simply handing everything over to AI, humans must supervise and intervene within the loop, ensuring that people are present in the "decision loop" and the "responsibility loop."
Point 2: Management takes ultimate responsibility for the use of AI.
To lighten the burden on middle managers overseeing AI subordinates, it is desirable to establish an AI utilization system under senior management. We recommend setting up an AI committee directly under the board of directors and creating a system in which senior management is continuously involved.
Point 3: Existing"Three-line defense model"Review
Introducing AI subordinates makes the traditional defense system—which consists of "front-line (first line)," "compliance department (second line)," and "internal audit (third line)"—less visible. It is essential to create a system that allows for accurate understanding and visualization of the AI's status.
Practical methods for data management and risk control
If you're going to manage things with KPIs, it's meaningless unless you can maintain transparency about the situation.
AI subordinates are often described as being like "today's young employees," but just as humans make mistakes, AI can also make errors. When humans make mistakes, they are educated and supported to understand "what went wrong." The same approach is necessary for AI.
Examples of specific countermeasures
Design of a KPI measurement system
You will need to set your own company-specific evaluation metrics, such as the number of items processed and the error rate.3-line defense modelSimilarly, the question is whether "the current state of AI utilization can be accurately grasped as a prerequisite for KPI measurement."
Clarification of the department head's responsibilities
If all the employees in a department are replaced by AI subordinates, only the department head who makes the final decisions will remain, potentially leaving one or two people to bear all the responsibility. Remember that if the scale of the business doesn't change, the total amount of risk and responsibility will also remain the same.
Redesigning Performance Evaluation and Compensation
In organizations where the use of AI subordinates is a given, the ultimate responsibility rests with senior management. Therefore, it is necessary to consider appropriate evaluation and compensation commensurate with that responsibility.
If risk management is not working
In situations where visualization is impossible, conventional methods become ineffective.
The biggest problem with introducing AI subordinates is that it becomes difficult to see "what is happening on the ground." In a checking system that relies on two or three lines of monitoring, if the AI subordinates cannot be properly visualized, traditional risk management methods will cease to function.
Practical advice
Changes in the "sense of work"
Losing visibility into the field allows us to move away from the traditional mindset of "making decisions without knowing what's happening." Ultimately, management has an obligation to establish organizational structures to identify and address the potential for serious incidents.
A supplement that doesn't get too technical.
AI subordinates can be described as being like "fast-working young employees." Just like humans, they need to be managed using KPIs, but the most important thing is not to forget the fact that "transparency in understanding the situation" is crucial.
Summary
The shift to an "AI subordinate-centric" model requires a fundamental reassessment of management practices.
In today's AI era, corporate organizations have traditionally been based on the premise of "human activity." However, with the introduction of AI subordinates, their structure must also shift to one that is "based on AI subordinates." This requires not only improving productivity, but also raising fundamental questions such as "who is responsible" and "how can we make things visible?"
It is the management team that needs to take on this transformation from the top down. They must be aware that existing risk management methods may become ineffective, and it is essential that they accurately understand and increase transparency regarding the use of AI.
Considering the increased burden on middle managers at the operational level, senior management should take responsibility for establishing an AI utilization system. KPI measurement and compensation structures will also need to be redesigned.Putting the "Human-in-the-loop" concept into practiceBy doing so, it is important to ensure that humans are involved in the loop of judgment and responsibility.
Before introducing AI subordinates, it's crucial to be aware that significant changes to the organizational structure will be necessary. Understanding that legal responsibility rests solely with legal entities and individuals, a system must be established to ensure accountability.
Ultimately, as the foundation of management"The shift from human-centric organizations to AI-centric organizations"This is something we need to address. Only then will it be possible to create a corporate organization that forms the foundation for AI utilization.
