Governance in the AI Era: Designing to Overcome Fragmentation

What the Fragmentation of Global Financial Governance Reveals

At the Lujiazui Forum held in Shanghai in June 2025, the “fragmentation” of global financial governance was a key theme. There is growing momentum to call for unity in rebuilding multilateral mechanisms.

At first glance, this news may seem like an international finance story unrelated to small and medium-sized enterprises (SMEs). However, I believe it holds essential insights for governance design within SMEs.

Fragmentation refers to the phenomenon where once-unified rules and standards become disjointed across countries and regions. Isn’t a similar thing happening in the workplace of SMEs?

For example, different compliance standards across departments, varying interpretations of rules depending on the business partner, and ad-hoc data management left to individual teams. These are precisely examples of “governance fragmentation.”

In this article, we will explain the design philosophy needed to overcome this fragmentation and the specific form of governance required in the AI era, from the perspective of an SME manager.

Barriers and Breakthroughs in AI Governance Adoption

The “AI Governance Support Service” announced by Strategic Health Advisors (SHA) in June 2025 packages know-how cultivated at large corporations, offering lightweight plans that SMEs can start implementing in approximately 4 to 6 weeks.

The core of this news lies in addressing the common SME dilemma: “We don’t know where to start.” Many managers, when hearing “AI governance,” imagine complex frameworks intended for large enterprises and hesitate to adopt them.

However, what is truly needed is not a “perfect system,” but a “minimal design” that works on the ground.

Specifically, I recommend starting with the following three points:

First, clearly define the scope of AI usage. Which tasks will use AI, and which will not? If this boundary is ambiguous, it can lead to significant risks later.

Second, establish a verification process for data output by AI. AI must be designed on the premise that it can make mistakes. Especially when handling customer information or financial data, human confirmation is indispensable.

Third, appoint a person responsible for AI governance. Without clarifying who makes the final decisions, accountability becomes blurred when problems arise, and the organization’s overall governance ceases to function.

Transforming Data Processing and Automating Governance Settings

The theme “Dialogue with business data and automation of governance settings,” reported by Nikkei xTech Active, brings a new perspective to governance design in the AI era.

Traditional governance was based on the idea of setting rules and having humans follow them. However, with the advancement of AI, an era is dawning where “AI itself generates and updates the rules.”

This represents a significant opportunity for SMEs. Automation is essential for maintaining governance amidst labor shortages.

Specifically, the following applications are conceivable:

・AI automates contract reviews with business partners, instantly detecting risky clauses.

・AI periodically performs internal compliance checks and reports the results.

・AI automatically manages employee data access permissions, regularly revoking unnecessary ones.

However, a crucial point to note is that before introducing “automation of governance settings,” it is necessary to visualize current business processes.

Automation merely streamlines the “extension of the current state.” If the current processes themselves have problems, automation risks amplifying those issues.

The New Relationship Between Prediction Markets and Compliance

The “Enterprise Global Prediction Market Compliance Solution” announced by StarCompliance and Kalshi may seem irrelevant to SMEs at first glance. However, there is much to learn from this news.

A prediction market is a mechanism where market participants trade on the probability of future events to make predictions. The idea of applying this to compliance can also be adapted for governance design in SMEs.

For example, by running a small-scale “internal control prediction market” to predict internal fraud risks, you can quantify on-the-ground intuition. Even simply asking employees to anonymously answer “What is the probability of expense fraud occurring this quarter?” helps visualize risks.

The advantage of this method is that risk information is gathered bottom-up rather than top-down. In SMEs, information asymmetry often exists between management and the front line. A prediction market-like mechanism can be an effective tool to bridge this gap.

Practical Challenges Seen from Internal Control Reports

The internal control report (81st term) submitted by Taikisha Ltd. is a statutory disclosure document required for listed companies. While it may seem unrelated to SMEs, there is much to learn from this document.

The essence of an internal control report lies in management evaluating whether the company’s internal controls are functioning effectively and publishing the results. This “self-assessment” process is the fundamental governance practice that SMEs should adopt.

Even for SMEs, conducting a “mini internal control evaluation” once a year is sufficient. Try answering these three questions:

1. What are the significant risks for your company, and are they being addressed?

2. Are there any inefficiencies or redundancies in your business processes?

3. Is the work environment conducive to following rules, or does it make compliance difficult?

If you cannot answer these three questions, “fragmentation” in governance may be occurring. Risk perceptions may differ across departments, or there may be a gap between rules and actual practices on the ground.

Three Actions SMEs Should Take Right Now

So far, we have explored governance design for SMEs using the latest news as material. Finally, here are three actions you can start implementing today.

First, create a “Governance Fragmentation Map.” Visualize where rules and standards are inconsistent within your company. Identify “fragmentation” across departments, business partners, and processes.

Second, establish “Minimal AI Governance Rules.” As SHA’s service demonstrates, you don’t need to aim for perfection. Start by defining just three things: the scope of AI usage, the data verification process, and the person responsible.

Third, cultivate the “Habit of Self-Assessment.” Set aside time once a year to evaluate your company’s governance status. You don’t need a formal document like Taikisha’s internal control report. What matters is that management takes the time to ask, “Is our current governance adequate?”

Governance is not a one-time setup. It is a “living entity” that must be constantly updated in response to environmental changes. Why not start designing to overcome fragmentation today?