Beyond “DX to AX”: The Role of AI That Management Must Define

The AI Wave Extends Beyond “Operational Efficiency”

“From DX to AX.” This phrase was highlighted at an educational IT conference hosted by Microsoft, arguing that AI Transformation (AX) is what comes after Digital Transformation (DX). Concurrently, the market is seeing a continuous stream of new solutions leveraging AI and cloud technology, such as “SaaS with advanced AI features for business automation” and “customizable B2B order management systems.”

At first glance, this seems like good news for executives and IT departments. If AI automates tasks and SaaS can be flexibly customized, it appears that the long-standing challenges of efficiency and digitalization could leap forward. However, this is precisely where the danger we have repeatedly pointed out—”the risk of management not defining IT”—lurks in a new form. In the face of a powerful tool like AI, is management once again being swept away by the “means” and losing sight of the “purpose”?

Two News Stories Reveal the Current State of AI & SaaS

First, let’s review the news that sparked this discussion. One is “KG Sekai Cart,” launched by the trading company Kanematsu. This is a B2B web-based order management system offered as a SaaS, yet it allows for customization to fit a company’s specific workflow. The other is “OmniSquare,” announced in beta by WhiteBox Inc. This SaaS boasts AI capabilities and promises to automate complex business processes.

Reading these news items superficially suggests a narrative: “Unprecedented, flexible, and powerful operational efficiency is achievable through AI and customization.” Indeed, compared to traditional packaged software or vendor-dependent systems, the agility of SaaS and the autonomy of AI are attractive. However, digging deeper from a management perspective reveals a different landscape. It is the recurrence of a long-standing structural problem: “the sophistication of means ends up masking the ambiguity of purpose.”

The New “Silos” Created by Customizable SaaS

The “customizable” feature emphasized by “KG Sekai Cart” requires particular caution. A system perfectly tailored to your company’s operations seems ideal. However, what management should question here is the fundamental reason: “Why does that workflow exist in the first place?” Customizing and solidifying inefficient rules or habitual processes born from poor inter-departmental coordination within a SaaS platform strays far from the essence of DX.

This risks worsening what we call the “split objective function.” The sales department customizes the tool for its “partial optimization” goal of “making order-taking easier,” while the procurement department does so to “reduce ordering errors.” As a result, the overall company’s data flow and decision-making structure are not considered, risking the creation of new, digitized silos. Precisely because SaaS is easy to implement, individual departments may adopt it without a management-level integration perspective, potentially accelerating the “SaaS spaghetti” phenomenon where the company’s IT landscape becomes even more complex.

The Pitfall of AI Business Automation: The Black-Boxing of Decision-Making

On the other hand, AI business automation tools, represented by “OmniSquare,” present a challenge of a different dimension. While dramatic efficiency gains are expected from AI making complex judgments autonomously, we must ask executives: Can you explain “on what criteria and why that AI is making a particular decision”?

For example, suppose an AI determines the timing and quantity of purchase orders. It would consider many factors: inventory data, sales forecasts, supplier reliability, etc. But what if its decision logic is a black box? If the AI suppresses orders during a sudden demand shift, leading to missed opportunities, it becomes extremely difficult for humans to verify and learn from the cause. This directly conflicts with what we value in “Management IT”: the “reproducibility of decisions.” Efficiency may increase, but the management decision-making process is delegated to and obscured by AI. This is a dangerous step that doesn’t redefine IT as a “decision-making apparatus” but rather externalizes and makes decision-making itself invisible.

The Essence of “AX” Lies Not in AI Adoption, But in Redesigning Decision-Making

So, how should we correctly define the shift “from DX to AX” as a management resource? The key is to stop viewing AI as merely an “operational efficiency tool.” The core of AX (AI Transformation) should be “designing new decision-making and operational architectures (structures) that leverage the characteristics of AI.”

When considering new AI tools or customizable SaaS, executives must ask themselves the following three questions.

Question 1: Which “IT” Does This Implementation Strengthen?

Evaluate it against the three IT classifications advocated by this media (Business IT, Management IT, Administrative IT).

• If strengthening Business IT (Growth & Speed), the purpose is likely speeding up customer response via AI or expanding sales channels via SaaS. In that case, ROI should be measurable through increased sales or reduced customer acquisition cost.
• If strengthening Management IT (Decision-Making & Reproducibility), the purpose is improving the quality of management decisions through AI data analysis or visualizing/integrating company-wide data flows with customizable SaaS. ROI is evaluated by faster decision-making speed or reduced risk.
• If strengthening Administrative IT (Stability & Cost), the focus is automating accounting tasks with AI or managing procurement costs with SaaS. ROI is clearly reflected in reduced man-hours or cost savings.

The problem is that rushing into “efficiency with AI anyway” without this distinction muddles objectives and makes investment effectiveness immeasurable. Introducing only Business IT and Administrative IT tools without a Management IT perspective will scatter data and make the company’s overall picture even less visible.

Question 2: Where is the Line Between Decisions Delegated to AI/Tools and Those Reserved for Humans?

This is a judgment concerning the very foundation of management. When introducing AI business automation, always create a “decision rule blueprint.” Under what conditions does the AI make autonomous judgments? What exceptions or cases above a certain threshold require human approval? This line-drawing is where management’s intent is reflected. The pattern to avoid most is being swept along by the tool’s functionality into an “entrust everything” mode, as it is tantamount to abandoning management.

Question 3: Does This Tool Repair or Worsen the Existing “IT Split”?

Before introducing new SaaS, review the company’s IT map (which tools each department uses). Will the new tool promote data linkage between departments and work towards integrating the previously disparate “objective functions”? Or will it fail to connect with existing tools, ending up as just a convenient tool for a specific department and deepening the split? Customizability should be exercised with the aim of achieving company-wide integration.

The Concrete First Step Executives Should Take in the AX Era

Theory alone won’t move things forward. So, what should be done concretely?

First, we recommend creating a simple checklist as an “AI/SaaS Evaluation Framework.” Discuss the following items with relevant departments during the review process for new tool adoption.

1. Clarify Purpose: Which of the three classifications does this tool strengthen? How will the expected ROI be measured?
2. Integration Potential: How will it link data with existing core systems or other SaaS? Is an API provided?
3. Decision Rule Transparency (for AI tools): To what extent are the AI’s decision criteria explainable? What is the human override process?
4. Customization Scope: Is customization for fixing department-local rules or for optimizing company-wide workflows?

Next, launch a small-scale “Management IT Integration Project.” Aim to create a “single source of truth” where company-wide data is aggregated and useful for management decisions. As a first step, if considering a customizable B2B order system like the one in the news, position it not merely as an order tool, but as a “platform for visualizing the entire supply chain and forecasting supply and demand.” This requires sales, procurement, production, and corporate planning to approach the implementation with a shared purpose.

Conclusion: AX is an Upgrade of Management, Not Just Tools

The phrase “from DX to AX” represents technological evolution. However, for executives, the essence of AX is not about acquiring the latest tool, AI. It is about redefining and integrating IT (the system of information and decision-making)—which management once neglected to define and allowed to split—into a form suitable for the AI era.

Customizable SaaS should not be a tool to indulge departmental whims, but “flexible building materials” for constructing company-wide optimal workflows. AI business automation should not be a black box replacing human judgment, but a “thinking extension device” that refines management’s decision criteria and enhances reproducibility.

Before being swayed by news of new tools, pause. Will it be a light that solves the old, deep-seated issue of your company’s “IT split”? Or will it merely paper over old problems with new tools, creating larger future repercussions? That discernment is the first, and most crucial, judgment for executives aiming to succeed in the AI era.